Ooh, this sounds very interesting; if they can pull this off, I can totally see Google taking a serious hit (especially with the quality already declined due to ads / SEO).
Could potentially be pretty bad for Google and I'm no MS/OpenAI fan. As a complete outsider to ML my hope is NLP style search with RL + LLM is going to disrupt the Old Way.
while competition is welcome, wouldn't they face the same exact problem as Google - filtering out tons of crap seo text and actually trying to separate out non-noise text from that?
And if "next gen SEO" (adversarial training data that cause something to be inappropriately or disproportionately represented in responses) lands in your chat search, it isn't a matter of just setting flags in a database to ignore or penalize a set of documents - you're gunnuh have to either retrain, or add a new set of layers (or similar) to filter out/penalize these results. If this starts happening at anything close to the rate that Google encounters spam, I can't see how they would keep up.
Language models are tested with hundreds of benchmarks including everything from bias to factuality and reasoning correctness. When they have a big deficiency it shows in the test scores.
Big deficiencies aren't monetizable by adversaries, tiny ones are (eg, impacting it's response to questions about one topic in particular).
In a very narrow niche there may not be many documents to pick from, either.
I don't think you can just automate this away in the context of generalized search. Search has to fulfill every niche; that seems like an indefensible position (strategically speaking, not in the moral sense). How can you benchmark bias in every niche? Your benchmarks show your reasoning is sound; what about the premises you're reasoning from? In the context of something narrowly scoped like a customer service bot, it makes sense to me how you could build expertise in constraining the model's output. But in terms of everything?
But I'll admit I don't have a crystal ball, happy to eat my words if they can operate with enough traffic and got long enough to attract the attention of spammers, and still keep them at bay. I think this space is stagnant and needs to be shaken up, and that chat interfaces have potential, so I'm not trying to be a hater. I just think this is gunnuh be a very difficult aspect.
would it? Many here are IT professionals and technology enthusiasts, naturally drawn to the next interesting thing, next disruption. Does the average user of Google care.
Google already has a AI chat some claim is more advanced than ChatGPT (Google Meena / Google LaMDA), but the there are no room for advertisement in a Chat (or voice) based search engine.
> Anything that has your attention can include ads.
It's funny because paper[1] from Google that started transformers and LLMs is called "Attention is all you need". And I agree that you can easily incorporate ads into chat, just ask ChatGPT to do it in a subtle way for you and you will see how it will incorporate ad into response.
There's a huge advertising opportunity in language model advertising. Think about anything a human salesperson does to advertise, except automatically and at scale.
Examples:
* When you ask for recommendations, it gives you x branded option or maybe the "best result" with an advertised alternative.
* When you search for symptoms, it recommends name brands instead of generics.
* When you ask for recipes, it directs you to <advertised cooking site>.
* When you ask for poetry in the style of Longfellow, it recommends an audiobook as "something you might enjoy".
* When you want to know about what happened to <recent celebrity that died>, it says "they died in a high speed chase according to <celebrity rag>. Here are some pictures and links to an article with more information"
* Platform integrations, like linking to the Google/windows store directly.
Many of these are already done by primitive assistants like Alexa. Going beyond that, you can copy pretty much everything done by influencer marketers in a chatbot.
When has a degraded product experience ever stopped advertisers? Even the famously UX-oriented Apple has inserted all sorts of upsell notifications, and nags into modern devices.
When I want to buy something like an air fryer, I either want something to either tell me the best overall air fryer, or the best one for me (“best air fryer for individuals with small kitchens”). I don’t want to do a whole bunch of research with just the raw specs to try to determine which is the best one for a person like me.
I could be nice to have an AI ask me questions about what I hope to use it for, what kind of space I have, and what my budget is, and then give me some recommendations.
GPT with ads wouldn't provide that service. All that would provide is a masterfully crafted, possibly false argument for why you should buy the product that paid for placement. You can already get that with google minus the conman pitch. So how is this even an improvement?
You wont get an honest answer from somebody taking money to sell one particular brand of air fryer. That would be like typing "air fryer" into Google Search and buying the first ad result.
I am not sure why this is different from search. Just show relevant ads on the right hand side of the page. If they are useful, people will click them too.
Google has the wrong leadership relative to Microsoft.
Nadella is the Gates/Jobs to Pichai's Ballmer/Sculley. Microsoft is playing 4D chess with their acquisitions and long-term vision; Google is floundering amidst perverse leadership incentives, inaction, and its non-diversified revenue stream.
I have prompted chatgpt enough over the past months to see how it would work. I have used my playground for it to get it to indicate what it is talking about, for instance and obviously, prefixing math content with /math but also /history, /media, /tech, /physics, /person etc.
Then you can send that to another service; pushing math to wolfram alpha (etc) makes chatgpt suddenly give perfect math answers and it works (it takes a bunch of hacks to ignore it’s confident lying part of answering equations). When we it names a person or location, it shows Wikipedia or maps or both etc.
In this way you use it’s bluffing powers without assuming (or even reading) it’s content; it returns articles etc (which can be wrong as well of course but at least you might know if it’s reputable or not).
I can easily see how bing or google would use something like this to search and return very relevant content, the queries might not be better than some people can create themselves with effort. However many people cannot create good queries and chatgpt creates good queries from extremely little input.
More importantly; it’s a conversation; you can fine tune the results. You can tell it that you meant the programming language, not oxidation when it shows completely bonkers results.
This will work and will work very well as far as I have been able to try. I cannot go further as I have no easy access to search engines without getting blocked immediately for automated querying.
Not sure how this is looking cost wise: chatgpt at scale will even hurt a company like MS for now. Having a million techies use it for some experimentation is a bit different than a billion+ users searching whatever.
Agreed, this seems like the immediate future. Large language model which has access to a bunch of "widgets" that it can use to extend itself. I can see two types of such widgets.
First type is info widgets which allow it access information outside of itself, like using a calculator by making it output something like {calc 1+1}, search wikipedia by {wiki something}, see the results of such a query and use that info in its output to the user.
Second type is UI widgets that allow it to display different types of things to the user. There is already a kind of widget in ChatGPT, which is display of Markdown. But it's easy to imagine widgets for say showing coordinates on a map, or visualizing something else.
Since there is a large number of such useful widgets, I'm guessing there will be an app store for them where developers can write their own widgets to extend it and make revenue based on how much they are used.
Since gpt can occasionally get little bits of code wrong that ruin the result, having some prepackaged functionality can prove useful. If you want to play around with letting GPT execute code, feel free to checkout my github repo that uses telegram as a frontend for a code-executing GPT: https://github.com/thornewolf/gpt-3-execution
ChatGPT fabricates lots of stuff, it's deceptive for common queries, but for programming related output, it's easily verifiable and delivers as an extremely valuable search tool. I can easly ask ChatGPT to explain stuff e.g. eBPF details without wasting time looking up the manuals. I hope Bing dominates Google and stackoverflow in this.
It's easily verifiable, but it may still waste time. I've had many cases where ChatGPT makes up functions that do exactly what I need, but then I find out these functions don't actually exist. This may not happen very often for super popular languages like Python or Javascript where training data is huge, but it happens all the time for the long-tail of languages. In those cases, it would've been faster for me to do a regular search.
I do agree with the overall point though. If you understand when to use it and when it's more likely to give you nonsensical answers, it can save a huge amount of time. But when I ask it about a topic that I don't know enough about to immediately verify the answer myself I'm forced to double check the answers for validity, which kind of defeats the purpose.
The best queries to ChatGPT are cases where I know what the answer should look like, I just forgot the syntax or some details. Bash scripts or Kubernetes manifests are examples here, I know them, I just keep forgetting the keywords because I only touch them every few weeks.
And don't get me started about asking ChatGPT about more general topics in e.g. economics or finance. What you get is a well-written summary of popular news and reddit opinions, which is dangerous if it's presented as "the truth" - The big mistake here is that the training procedure assumes that the amount of data correlates with correctness, which isn't true for many topics that involve politics or similar kinds of incentives where people and news spread what conveniently benefits them and gets clicks.
Wasting time and having to be constantly vigilant is exhausting and a slippery slope that makes it easier to fall for deceptive content and settling for "I don't know, it's probably close enough" instead of insisting on precision and accuracy.
Humans take a lot of shortcuts (such as believing more easily the same facts presented with a confident tone) and the "firehose of bs" exploits it: this was already the case before generative AI, but AI amplifies the industrial-scale imbalance between the time needed to generate partially incorrect data and the amount of time/energy required to validate.
Agreed that it is a slippery slope. Programming is understanding - like writing or teaching is understanding. To really understand something, we must construct it ourselves. We will be inclined to skip this step. This comment sums it up well:
> Salgat 8 days ago
> The problem with ML is that it's pattern recognition, it's an approximation. Code is absolute, it's logic that is interpreted very literally and very exactly. This is what makes it so dangerous for coding; it creates code that's convincing to humans but with deviations that allow for all sorts of bugs. And the worst part is, since you didn't write the code, you may not have the skills (or time) to figure out if those bugs exist
> To really understand something, we must construct it ourselves.
I think the real power of these bots will be to lead us down this path, as opposed to it doing everything for us. We can ask it to justify and explain its solution and it will do its best. If we're judicious with this we can use it to build our own understanding and just trash the AI's output.
How is that worse than having to look at every online post's date to estimate whether the solution is out of date? Or two StackOverflow results where one is incorrectly marked as duplicate and in the other the person posting the answer is convinced that the question is wrong.
ChatGPT can completely cut out the online search and give an answer directly about things like compiler errors, and elaborate further on any detail in the answer. I think that 2-3 further GPT generations down the line it will be worth the time for some applications.
The problem I see is less the overall quality of responses but people overestimating on where it can be used productively. But that will always be a problem with new tech, see Tesla drivers who regularly take a nap in the car because it didn't crash yet.
Unless the training of ChatGPT has a mechanism to excise the influence of now out-of-date training input, it will become increasingly more likely to give an outdated response as time goes by. Does its training have this capability?
The trick is to use it as an LLM and not a procedural, transactional data set.
For instance, “how do I create a new thread in Python”. Then ask “how do I create a new thread in Python 3.8”. The answers will (probably) be different.
Any interface to chatgpt or similar can help users craft good prompts this way. It just takes thinking about the problem a little differently.
One wildly inefficient but illustrative approach is to use chatgpt itself to optimize the queries. For the Python threading example, I just asked it “ A user is asking a search engine ‘how do I create threads in Python’. What additional information will help ensure the results are most useful to the user?”.
The results:
> The user's current level of programming experience and knowledge of Python
> The specific version of Python being used
> The desired use case for the threads (e.g. parallel processing, concurrent execution)
> Any specific libraries or modules the user wants to use for thread creation
> The operating system the user is running on (as this may affect the availability of certain threading options)
So if you imagine something like Google autocomplete, but running this kind of optimization advice while the user builds their query, the AI can help guide the user to being specific enough to get the most relevant results.
I understand this works well in many practical cases, but it seems to depend on a useful fraction of the training material making the version distinction explicit, which is particularly likely with Python questions since the advent of Python 3.
One concern I have goes like this: I seriously doubt that current LLMs are capable of anything that could really be called an understanding of the significance of the version number[1], but I would guess that it characterizes the various Python-with-versions strings it has seen as being close[2] so I can imagine it synthesizing an answer that is mostly built from facts about Python2.7. With a simple search engine, you can go directly to checking the source of the reply, and dig deeper from there if necessary, but with an LLM, that link is missing.
[1] The fact that it listed the version as being a factor in reply to your prompt does not establish that it does, as that can be explained simply by the frequency with which it has encountered sentences stating its importance.
[2] If only on account of the frequency with which they appear in similar sentences (though the whole issue might be complicated by how terms like 'Python3.8' are tokenized in the LLM's training input.)
It's all imperfect, for sure. For for instance see this old SO question [1], which does not specify python version. I pasted the text of the question and top answer into GPT-3 and prefaced it with the query "The following is programming advice. What is the langauge and version it is targeted at, and why?"
GPT-3's response:
> The language and version targeted here is Python 3, as indicated by the use of ThreadPoolExecutor from the concurrent.futures module. This is a module added in Python 3 and can be installed on earlier versions of Python via the backport in PyPi. The advice is tailored to Python 3 due to the use of this module.
That's imperfect, but I'm not trying to solve for Python specifically... just saying that the LLM itself holds the data a query engine needs to schematize a query correctly. We don't ChatGPT to understand the significance of version numbers in some kind of sentient way, we just need it to surface that "for a question like X, here is the additional information you should specify to get a good answer". And THAT, I am pretty sure, it can do. No understanding required.
I don't think the issue is whether current LLMs have sufficient data, but whether they will be able to use it sufficiently well to make an improvement.
The question you posed GPT-3 here is a rather leading one, unlikely to be asked except by an entity knowing that the version makes a significant difference in this context, and I am wondering how you envisage this being integrated into Bing.
One way I can imagine is that if the user's query specified a python version, a response like that given by GPT-3 in this case might be used in ranking the candidate replies for relevance: reject it if the user asked about python 2, promote it if python 3 was asked for.
Another way I can imagine for Bing integration is that perhaps the LLM can be prompted with something like "what are the relevant issues in answering <this question> accurately?" in order to interact with the user to strengthen the query.
In either case, Bing's response to the user's query would be a link to some 3rd-party work rather than an answer created by the LLM, so that would answer my biggest concern over being able to check its veracity, though its usefulness would depend on the quality of the LLM's reply to its prompts.
On the other hand, the article says "Microsoft is betting that the more conversational and contextual replies to users’ queries will win over search users by supplying better-quality answers beyond links", apparently saying that they envision giving the user a response created by the LLM, which brings the question of verifiability back to center stage. Did you have some other form of Bing-LLM interaction in mind?
Unless the responses in those old online forums where intentionally malicious, they might be reasonably helpful even if not 100%.
While ChatGPT spews out complete nonsense most of the time. And the dangerous part is that that nonsense looks very reasonable. It gets very frustrating after some time, because at first you are always happy that it gave you a nice solution, but then it's not usable at all.
I'm a glass-half-empty sort of person: in my experience, even perfectly good answers for a different version can be problematic, and sometimes harmful.
I am foreseeing a future in which programming language designers match the most sought after functions in google/bing/chatgpt and then implement those that do not yet exist because apparently there is a real need for those.
Yes, I had the same thought. LLM’s might be instrumental in new language design. If it can understand the most common structures being used, it makes sense to build libraries, macros, or language features.
I agree. ChatGPT is really really bad. It just makes up stuff and wraps its fabrications in an air of authority.
A "bullshit sandwich" if you will.
When one tells people this we get the reply "but so do random blogs! or reddit comments!". Well yes, but they're just random blogs and reddit comments, often peppered with syntactic and spelling mistakes, non sequiturs, and other absurdities. Nobody would take them seriously.
ChatGPT is very different. It doesn't say "this random redditor says this, and this other random redditor says the exact opposite, so IDK, I'm just a machine, please make up your mind".
What it says is "this is the absolute truth that I, a 'large language model', have been able to extract from the vast amount of information I have been trained on. You can rely on it with confidence."
I'm sorry to sound hyperbolic but this cannot end well.
ChatGPT doesn't say anything of the sort. In fact, it will vehemently insist that what it says is not necessarily true or accurate if you challenge it.
I'm sorry but this is demonstrably false. I have posted examples of this on HN before. Yes, if you tell ChatGPT that it's wrong, in some cases it says "I'm sorry" and tries again (and produces some other random guess). But if you ask it "are you sure?" it invariably affirms that yes, it's sure and it's in the right.
Hm, you're right. I'm pretty sure that it wasn't so gung-ho when I played with it earlier, but now even very explicit instructions along the lines of "you should only answer "yes" if it is absolutely certain that this is the correct answer" still give this response. Ditto for prompts like "is it possible that your answer was incorrect?"
I like bouncing my code problems off ChatGPT, it can give me an answer and I don't feel bad if I forgot something simple. The issue is I've had it give me completely wrong code only for it to be like "I'm sorry" and provide a second incorrect response.
Using a purpose built (or trained I guess) model for code generation would likely have better results. GitHub copilot is useful for this reason. I find ChatGPT for code is mainly useful if you want to instruct it in natural language to make subsequent changes to the output.
If you ask, there's a good chance ChatGPT can create that function for you. Just tell it: "That function `xyz()` doesn't exist in the library, can you write it for me?"
It does such a good job at giving answers that sound right, and are almost correct.
I could imagine losing many hours from a ChatGPT answer. And if you have to go through the trouble to verify everything it says to make sure it's not just making crap up, then imo it loses much value as a tool.
But so do people: I spent an hour yesterday trying regexps that multiple people on Stackoverflow confirmed would definitely do what I needed, and guess what? They did not do what I needed.
Same with copilot. Sometimes it's ludicrously wrong in ways that sound good. I still have to do my job and make sure they are right. But it's right or right enough to save me significant effort at least 75% of the time. Right enough to at least point me in the right direction or inspire me at least 90% of the time.
They released a zero day for a security hole in the human brain. That's what ChatGPT is. The security hole is well known and described perhaps the most understandable format is the book Thinking Fast And Slow which describes, ah, if I try to explain I will surely botch but perhaps put it this way: how things that appear more credible will be deemed credible because of the "fast" processes in our brains.
In this particular case, ChatGPT will write something nonsensical which people will accept more easily because of the way it is written. This is inevitable and extremely dangerous.
> Humans are still a lot better at writing something nonsensical that people will accept easily because of the way it's written.
Some are but not many. And then there's the amount. That's the crux of the matter. Have you seen that Aza Raskin interview where he posited one could ask the AI to write a thousand papers citing previous research against vaccines and then another thousand pro-vaccines? No human can do that.
> People are just as good at making up convincing sounding nonsense.
Perhaps as you just did, as I can find no one actually "injecting themselves with bleach."
The overall point stands: the difference between reading something dumb and doing that dumb thing is what it means to have agency. I personally don't think we should optimize the world 100% to prevent people who read something stupid from doing that stupid thing.
Or, if that's the path we're going to take, maybe we should first target things like the show Ridiculousness before we start talking about AI. After all, someone might do something dumb they see on TV!
People have absolutely injected themselves with what's known as "Miracle Mineral Solution", which is essentially bleach. It's more frequently drunk, of course.
Self Reply: I just now thought to use Copilot to get my regex and wow! I described it in a comment and it printed me one that was only two characters off, and now I have what I needed yesterday. I'd since solved the problem without a regex.
It's not perfect, but sometimes its amazing. In your case, not only did it provide the right solution, but it was about as fast as theoretically possible. About as fast as if you already knew the answer.
I had a similar experience with a shell command. Searched google, looked at a few posts, wasnt exactly what I needed but close. Modified it a few times and got it working. Went to save the command in a markdown file and when I explained what the command did, copilot made a suggestion for it. It was correct and also much simpler.
It went from taking 5-10 minutes to stumble through something just so I could do the thing I really wanted to do, to finding the answer instantly all from within the IDE. Can keep you in flow.
I dunno, verifying and adjusting an otherwise complete answer is a lot more rote than originating what that answer would be, and I think that has value.
It shows how form matters more than substance. Say real information in some poor structure and people will think you're wrong
Say incorrect stuff authoritatively and people will think you're right.
It happens to me all the time. I can't structure accurate information in a better way then some bullshit artist can spit off what they imagine to be real so everyone walks away believing in their haughty nonsense.
ChatGPT exploits that phenomena which is why it sounds like some overly confident oblivious dumb dumb all the time. That's the training set.
Almost once a week I'll go through a reddit thread and find someone deep in the negatives who has clearly done their homework and is extraordinarily more informed than anyone else but the problem is everyone else commenting is probably either drunk or a teenager or both so it doesn't matter.
Stuff is hard and people are mostly wrong. That's why PhDs take years and bars for important things are set so high
>It does such a good job at giving answers that sound right, and are almost correct.
For sure. But you have to compare against alternatives.
What would that be? Posting to stack overflow and maybe getting a helpful reply within 48 hours.
> I could imagine losing many hours from a ChatGPT answer.
Dont trust it. Verify it.
We expect to ask a question and get a good answer. In reality we should leverage how cheap the answers are.
The key is that it is way faster and has a broader set of knowledge than a human. Being an editor is often easier and more productive than being both a single generator and editor
ChatGPT can play an interesting role by separating duties in a process of productivity. ChatGPT can generate tons of true/false suggestions very fast and understandable by humans. Sometimes this helps a lot.
The downside is the risk of atrophying one's own mental ability to generate such suggestions if excessively relied upon. Given my druthers, I'd prefer to be a generator of text ChatGPT would want to absorb than to be a consumer of the mystery meat it is regurgitating.
Maybe its better in some programming languages, but my experience with verilog/systemVerilog output is that it generates a design with flaws almost every time (but very confidently). If you try to correct it with prompting it comes up with reasonable sounding responses about what its fixing then just creates more wild examples.
One pretty consistent way to see this is to ask for various very simple designs like a n-bit adder, it will almost always do something logically incorrect or syntactically incorrect with the carry in or carry out
ChatGPT has acted as an advanced rubber duck for me. It outputs a lot of bullshit but so often it gives me the prompt or way of thinking needed to move on.
And it’s so much faster than posting on stack overflow or some irc. It doesn’t abuse you for asking dumb questions either.
I tested chatgpt with some domain specific stuff and found it so wrong on the fundamentals that I immediately lost trust in any of its output for learning. I would not trust it to explain anything eBPF related reliably. You are more likely to get something that is extremely wrong or, worse, subtly wrong.
Add documentation to this method :
[paste a method in any language]
For me the results have been impressive. It’s even more impressive if you are not English speaking because it explains what the code does but also translates your domain terms in your own language.
More than code generation I see a really concrete application in having autogenerated and up to date documentation of public methods. It could be generated directly in your code or only by your IDE to help you in absence of human written documentation.
Other interesting things it can does is basic code review by proposing a « better » code and explaining what and why it changed something.
It can also try to rewrite a given code in another language. I haven’t tried a lot of things due to the limitations in response size but for what I tested, it looks like it is able to convert the bulk of the work.
While I’m not really convinced by code generation itself (a la copilot) I truly think that GPT can be a really powerful tool for IDE editors if used cleverly, especially to add meaning to unclear, decade old codebases from which original contributors are long gone.
And knowing that what is hard is not writing but reading code, I see GPT to be a lot more useful here than helping writing 10 lines in a keystroke.
> A common advice for documentation is "why not how", I'm not sure you can do "why" by looking at the "how".
You are right. It’s the rule when you write the doc.
But when you are let alone in an unknown codebase, having your IDE summarize the "what" in the auto completion popup could be really useful. Especially in codebases with wrong naming conventions.
> A common advice for documentation is "why not how", I'm not sure you can do "why" by looking at the "how".
The "why" is important for inline comments, but for function and method comments I think the biggest is neither "why" nor "how", but "what". As in, "what does this method do?" especially with regards to edge cases.
I tried a few methods just now; it gives okay-ish docs. Lots of people don't write great comments in the first place, so it's about on-par. Sometimes it got some of those edge cases wrong though; e.g. a "listFiles()" which filters our directories and links isn't documented as such, but then again, many people wouldn't document it properly either.
I had a lot of fun with ChatGPT’s wholly fabricated but entirely legitimate-sounding descriptions of different Emacs packages (and their quite detailed elisp configuration options) for integrated cloud storage, none of which exist.
I’m not sure that fabricated nonsense would actually make Bing’s results any worse than they are today.
“It’s okay I don’t mind verifying all these answers myself” is an odd sort of sentiment, and also inevitably going to prove untrue in one sense or another.
If it generated the code, I would have to audit that code for correctness/safety/etc.
Or, more likely, I would just lazily assume everything is fine and use it anyway, until one day the unexamined flaws destroyed something costly in a manner difficult to diagnose because I didn't bother to actually understand what it was doing.
There really should be more horror at the imminent brief and temporary stint of humans as editors, code reviewers, whatever, over generative AI mechanisms (temporary because that will be either automated or rendered moot next). I'm unaware of any functional human societies that have actually reached the "no one actually has to work unless they want to do so, because technology" state, so this is an interesting transition, for sure.
> Or, more likely, I would just lazily assume everything is fine and use it anyway, until one day the unexamined flaws destroyed something costly in a manner difficult to diagnose because I didn't bother to actually understand what it was doing.
Well yeah, I'm right there with you. But that feels a lot like any software, open or closed source. Human programmers on average are better than AI programming today, but human programmers aren't improving as fast as AI is. Ten years from now, AI code will be able to destroy your data in far more unpredictable and baroque ways than some recent CS grad.
> I'm unaware of any functional human societies that have actually reached the "no one actually has to work unless they want to do so, because technology" state, so this is an interesting transition, for sure.
This is a really interesting thought. Are we seeing work evaporate, or just move up the stack? Is it still work if everyone is just issuing natural language instructions to AI? I think so, assuming you need the AI's output in order to get a paycheck which you need to live.
Then again, as a very long time product manager, I'm relatively unfazed by the current state of AI. The hundreds of requirements docs I've written over decades of work were all just prompt engineering for human developers. The exact mechanism for converting requirements to product is an implementation detail ;)
Knowing little about how ChatGPT actually works, is there perhaps a variable that could be exposed, something that would represent the model's confidence in the solution provided?
It's not available for ChatGPT but the other GPT models can expose the probability for each generated token, which can serve as a proxy for confidence.
Tuning the temperature and topP parameters you can also make the model avoid low probability completions (useful for less creative use cases where you need exact answers).
> It's not available for ChatGPT but the other GPT models can expose the probability for each generated token, which can serve as a proxy for confidence.
A proxy for confidence in what exactly?
Language models represent closeness of words, so a high probability would only express that those words are put together frequently in the corpus of text; not that their meanings are at all relevant to the problem at hand. Am I wrong?
In cases where you ask GPT-3 questions that have a clear correct answer, I think you can use the probability to judge how correct the answer is. For example, when asking "How tall is Mount Everest?" I would want the completion "Mount Everest is ____ meters above sea level." to have a very high probability for the ____ tokens.
This is because I'm operating under the assumption that sequences of words that appear often in the training set are more likely to represent something correct (otherwise you might as well train on random words). This only holds if the training set is big enough that you can estimate correctly (e.g. if the training set is small a very rare/wrong phrase may appear very often).
Maybe confidence was the wrong word, but for this kind of questions I would trust a high-probability answer way more than a low one. For questions belonging to very specific subjects, where training material is scarce, the model might have very skewed probabilities so they become less useful.
> In cases where you ask GPT-3 questions that have a clear correct answer, I think you can use the probability to judge how correct the answer is. For example, when asking "How tall is Mount Everest?" I would want the completion "Mount Everest is ____ meters above sea level." to have a very high probability for the ____ tokens.
Maybe, as long as you're aware that this is the same kind of correctness that you get from looking at Google's first search results (the old kind of organic pages, not the "knowledge graph", which uses an different process - precisely to avoid being spammed by SEO) i.e. "correctness by popularity".
This means that the content that is more replicated will be considered more true by the system, regardless of its connection to reality or its coherence with the rest of the knowledge in the system. And you know what they say about big enough lies that you keep repeating millions of times.
> This means that the content that is more replicated will be considered more true by the system, regardless of its connection to reality or its coherence with the rest of the knowledge in the system.
I understand the problem, but what better way do we currently have to measure its connection to reality? At least from a practical point of view it seems that LLMs have achieved way better performance than other methods in this regard, so repeatedness doesn't look like that bad a metric. Or rather, it's the best I think we currently have.
> I understand the problem, but what better way do we currently have to measure its connection to reality?
We can consider its responses to a broader range of questions than those having an unambiguous and well-known answer. Its propensity for making up 'facts', and for fabricating 'explanations' that are incoherent or even self-contradictory shows that any apparent understanding of the world being represented in the text is illusory.
I agree, and furthermore, a search engine is constrained to pick its responses from what's already out there.
This line of thought is a distraction, anyway. The likelehood that GPT-3 will do as well as a search engine on topics where there is an unambiguous and well-known answer does little to address the more general concern.
I don't think so. It doesn't understand what it says, it basically does interpolation between text it copy-pastes in a very impressive manner. Still it does not "understand" anything, so it cannot have any kind of confidence.
Take Stable Diffusion for instance: it can interpolate a painting from that huge dataset it has, and sometimes output a decent result that may look like what a good artist would do. But it doesn't have any kind of "creative process". If it tells you "I chose this theme because it reflects this deep societal problem", it will just be pretending.
It may not matter if all you want is a nice drawing, but when it's about, say, engineering, that's quite different.
I'd say you can't do that, because ChatGPT has no internal model for how the things it is explaining work; so there can't be any measure of closeness to the topic described, as would be the case for classification AIs.
ChatGPT models are language models; they represent closeness between text utterances. It works by looking for the chains of words most similar or usually connected to those indicated in the prompt, with no understanding of what those words mean.
As a metaphor, think of an intern who every morning is asked to buy all the newspapers in paper form, cut out the news sentence by sentence, and put all the pieces of paper in piles grouped according to the words they contain.
Then, the director requests to write a news item on the increase in interest rates. The intern goes to the pile where all the snippets about interest rates are placed, will randomly get a bunch of them, and write a piece by linking the fragments together.
The intern has a PhD in English, so it is easy for them to adjust the wording to ensure consistency; and the topics more talked about will appear more often in the snippets, so the ones chosen are more likely to deal with popular issues. Yet the ideas expressed are a collection of concepts that might have made sense in their original context, but have been decontextualized and put together pell-mell, so there's no guarantee that they're saying anything useful.
> ChatGPT models are language models; they represent closeness between text utterances. It works by looking for the chains of words most similar or usually connected to those indicated in the prompt, with no understanding of what those words mean.
No, it does not work that way. That’s how base GPT3 works. ChatGPT works via RLHF and so we don’t “know” how it decides to answer queries. That’s kind of the problem.
General Purpose Bullshitting Technology. I've always found LLMs most useful as assistants when working on things I'm already familiar with, or as don't-trust-always-verify high temperature creatives. I think that attempts to sanitize their outputs to be super safe and "reliable sources" will trend public models towards blandness.
SEO optimized sites can also be identified and avoided. There's various indicators of the quality of a site, to the point where I'm positive most people on HN can know to stay away or bail from one of those sites without even being consciouly aware of what gave them that sense of SEO.
This resonates with me. We have all worked with someone who is a superb bullshitter, 100% confident in their responses, yet they are completely wrong. Only now, we have codified that person into chatGPT.
I doubt it. Even if it was trained with 100% accurate information chatGPT would still prefer an incorrect decisive answer to admitting it doesn't know.
For now, but perhaps we are at a level where enough knowledge is there that future solutions can be inferred from the past ones and documentation/code of libraries available on the internet.
I believe the exact opposite. If one could prove that text has not been generated by an AI, that would have immense value. StackOverflow has a built-in validation process ("mark as the solution"), which says that some human found that it solved the problem. Doesn't mean it's correct, but still, that's something.
I really wonder what impact ChatGPT will have on search engines. I could imagine that the first 4 pages of Google/Bing results end up being autogenerated stuff, and it will just make it harder to find trustworthy information.
It doesn’t consistently know words have individual letters since it’s trained using byte pair encodings. This is one reason earlier versions of it couldn’t generate rhymes.
When it works it is great. I've been using it instead of Google a lot too, but when it makes mistakes it requires someone familiar with a subject to detect it. I'm not sure if it is ready to be used as as a search engine by everyone.
For example recently I asked it for the best way to search in an mbox file on arch Linux. It proceeded to recommend a number of tools including mboxgrep. When I asked how to install it on arch it gave me a standard response using the package manager, but mboxgrep is not an arch package. It isn't even an aur package. It requires fetching the source and building it by yourself(if I remember correctly one has to use an older version of gcc too). None if it was mentioned by chatgpt.
This is not the first time BTW, there was another software it recommended that Debian doesn't know about, when I asked it another time.
When ChatGPT serve me broken code, I would paste the errors back in and ChatGPT would try to make corrections. I don't see why ChatGPT couldn't do that itself with the right compiler, saving me from being a copy and paste clerk.
I found ChatGPTs answers relatively accurate for explaining programming related queries, feeding it documentation and asking questions related to that, etc. But I've also tried to use it for travel and health related queries. For travel queries, it confidently tells me the wrong information, "Do most restaurants in Chiang Mai accept credit cards?" got "Yes, most restaurants in Chiang Mai accept credit cards!", which is completely false. Also got wildly inaccurate information about the quality of drinking water. And for health related queries, it tells me the same weasel-worded BS that I get on health spam blogs. I tried to dig out more information regarding sources of both travel and health related information, but ChatCPT simply said it doesn't know the details of the sources of information.
I think a new implementation of ChatGPT is worth exploring though, one that cites sources and gives links to further information, and also one that has the ability to somehow validate it's responses for accuracy.
I think we should let this C era meme die, the manuals are often terrible. I'm currently working with the AWS SDK Python documentation and it's a hot pile of garbage from all points of view (UX, info architecture, technical detail, etc.).
Python lang docs are "kind-of-OK" but when someone raves about them I'm left scratching my head. Information is not always well-organized, examples are hit-and-miss, parameter and return types not always clear, etc.
Referencing docs as a programmer is generally a nightmare and a time sink, and it's the one use case where ChatGPT is slowly becoming indispensable crutch for me. I can ask for very specific examples that are not included in the docs, or that cannot be included in the docs, for example combinatorial in nature: "how can I mock this AWS SDK library by patching it with a context manager"? Occasionally it will hallucinate, but even if it gets it 8/10 times right - and it's higher than that in practice - it will prove revolutionary at least for this use case.
I'm amazed by how divisive it is. I've also been using it to significantly increase my productivity, be that documenting things or having it mutate code via natural language or various other tasks. I feel that if you keep in mind that hallucination is something that can happen, then you can somewhat mitigate that by prompting it in certain ways. E.g. asking for unit tests to verify generated functions, among other things.
I find this tool so useful, that I scratch my head when I read about how dismissive some people are of it.
I think one of the reasons why Python got such a reputation for good docs is because its primary competitors back in the day were Perl and Ruby. Ruby has horrible documentation to this day, and Perl has extensive docs that are difficult to navigate; in comparison with either, Python was definitely superior.
> I'm currently working with the AWS SDK Python documentation and it's a hot pile of garbage from all points of view (UX, info architecture, technical detail, etc.).
I agree that pretty much all AWS documentation is woeful, and it's a travesty that the service is so expensive yet its documentation is so poor. I would gladly dump AWS and never use it again, as I hate paying top-dollar to decipher the AWS doc team's mistakes (not to mention that they are unresponsive to bug reports and feedback).
My point was made more in jest, and supposed to point out the irony of the communities' changing expectations of what documentation should be like. I predict that in a few years we'll be circling back to prioritizing writing software documentation well. (Kind of like how everybody was hating on XML for the past 20 years and it's now having a renaissance because it actually does what it's supposed to well very well.)
where are they going to get a steady fresh firehose of data comparable to stackoverflows? who are the magical entities that will be feeding them all these inputs for bing to claim all the fame?
My concern is about this whole business model of adding ChatGPT to a search engine. A search engine is a free tool. Because it is free, the incentive to make money is via ads, referral etc. That means preferential treatment where the money is.
ChatGPT (or a similar product) focuses on solving users problem interactively. No ads, no going to another website etc. How would you make money from a search engine?
I was hoping a simple, paid model to start with. Over time, as the LLMs become commodity (200GB, runnable on Intel/AMD), ship it as part of the OS and other devices.
I would be somewhat wary of this idea. Because bots like ChatGPT will oversell and it is kinda creep.
Current search ads display options and let you filter through which is neat, but ChatGPT like agent will shred the original content and make it invisible to distinguish which is ads, and worse false advertising, which is not.
I think let ChatGPT to infer the attribution automatically and do a revenue sharing with original link owner will be a better option.
Are you by any chance referring to Goggles (with two 'g's) which is indeed a beta feature, but has nothing to do with Google? https://search.brave.com/goggles/discover
They could state "Some of the content in the following paragraph is sponsored by Apple:" and then weave it in. Edit: but still have vital content in the paragraph, like product placement in movies.
Your repository can be secured by enterprise grade post-modern Security-as-Code(tm) frameworks pioneered by SaCicorp(R) backed by YC '24-27 and other leading partners that enforces an indomitable security posture in any threat landscape.
You realize google probably has a 10x better AI then ChatGPT. They just arn’t publicizing it. Google has all the data. It's not even a competition. If bing adds it then google will simply add a better version of it.
> solving users problem interactively. No ads, no going to another website etc
I think there is no technical barrier to add Ads to a chat bot. It's even more deceptive when promotion campaigns are embbeded into text contexts implicitly. It would be much more dangerous and harder to block than, let's say a DIV on a webpage.
… now you can make money on the links directly, perhaps have sponsorships influence the recommendations, and have a strong signal of intent to purchase you can try to monetize later.
“Hey bing what’s a good hotel in downtown Montreal?” -> same
I see a lot of people saying this, but it doesn’t make sense to me.
As your conversation evolves with the bot, targeted ads could be shown with the same (or better) level of intent data available based on the human’s input.
the real “redpill” in all this is that chatgpt is a feature not a product; if microsoft work out a form factor that is cost efficient at search engine scale, google is more than capable of copying it. unlike the annual halfhearted google chat app, this needs no network effect and in fact google beats msft at the one network effect that does matter here: indexing the web.
I guess they decided Bing doesn't suck enough already, so they want to make it even worse.
Hey Microsoft, how about instead of adding chatgpt, filter the results so Bing actually works as a search engine? That is, select only the search results that contain the search terms, rather than returning the whole internet in an order somewhat influenced by the search terms.
There's a well known phenomenon that everyone appends "reddit" to their search results as otherwise there's too much SEO garbage. The thing with search engines is you can actually tell how good they are by comparing a few queries.
This has been going for many years already, with Google giving answers right in the search results and just using websites as their private content factories. The relationship has become more and more uneven every year, with no pushback from the websites (a collective action problem, some traffic is better than no traffic, or worse, traffic to your competitor) and it probably won't change as long as Google pays lip service and adds a few links once in a while. The quality of content and therefore the quality of Google's search results suffer in the long term because it makes no sense to invest in it, but as long as it's good for Google's quarterly results, they don't care. Their problem is new sites like Discord and mobile-only apps that never got addicted to traffic from Google in the first place.
I live from the content I write. It's not fluff. Some of it comes from weeks-long email conversations with government officials. It takes a lot of research and help from experts I have long-standing relationships with.
If search engines serve that information but deny me the traffic, the website dies, as does the source of the information.
I can deal with lazy copywriters just rephrasing my work because the original still outranks them, and I have legal options to deal with them.
I can't do anything if Google - over 80% of of my traffic - decides to proxy my content and starve me of my income.
That's one of the reasons more and more sites are starting to require sign-up to continue reading the article. So only provide a summary of your article to search engines and users can access the rest by the annoying sign up (or captcha). But that also means less of the content is searchable so the summary you provide for search engines has to be really good (maybe even AI can help to produce this summary).
And it's probably not just big companies we will have to worry about because at least they can be somewhat regulated and they are in the public eye. The other "threat" in the future is the "distributed" AI when people can run their own personal AI assistants that could collect information for them by any means (singing up to to websites, e-mailing, calling people, talking to other AI agents) and with filtering out ads and sponsored content. At that point probably everything worthwhile will be paywalled and the "SEO" game will be to convince/trick these AIs to sing up / pay for your content.
That's a very positive way to look at it: Use the AI for your own benefit to generate a summary but the full information is hidden behind a signin/paywall.
Of course whether this ends up being better for humanity is the question. On the other hand, maybe Google should be paying people for high quality, trainable content.
I have no paywall by design. It's a core principle behind the website I run. I can cover the bills through affiliate links for services I actually recommend. However those are stripped by whoever uses my content for their own benefit, including Google.
Google sometimes shows snippets from my website as a direct answer to a query (in a collapsible box). If the out-of-context answer satisfies the user, then I get no traffic.
In at least one case, Google gave a completely wrong answer (snippet for the wrong question), and credited me for it.
For the record, I run a pretty lean, annoyance-free website. It looks the same without an ad blocker. It uses Plausible for analytics. No newsletters, no annoyances. I still get shafted.
That's just too bad. I've been creating web services for free since 2000 and just gave up once I realized that the web is innately commercial and bad for users. No matter what I do is just at odds with the fact that I have to host the content myself when the users could easily just mirror it with something like bittorrent but all we lack is the 10 lines of code for that infrastructure to be usable in the common user's flow. Plus the moment that happens big corpo and govo will cry CP and copyright and there mere act of using a computer will become strictly regulated and file hosting will be illegal. Later on, I spent months creating high quality articles in niche technical subjects, but quickly stopped as I realized that I don't want to contribute to the web anymore. When a real medium for grownups (both because the regulation on the web is bogus and dystopic addressing childish concerns with no bearing in reality, and because the web is a terrible amateur protocol) appears I will publish on that.
Information should be retrievable without all kinds of nonsense personalization and ads, and in milliseconds, not 10 seconds. The 99.999% of web content made between 2000-now is not whatever virtuous content you claim to have struggled to create, but a bunch of bloat that just wastes the user's time, and most of the time it's not even a good answer to the question, but just the exact same paraphrased answer from several other commercialized blogs. Your doomsday scenario here would be the perfect justice, and you will be one of the _very few_ innocent victims of it. Of course it won't be so simple, anyone like Google would find a way to make the user experience insufferable. I don't see a place for monetization on the future web, it will just be a bunch of people exchanging information, like where are the bad guys with guns and should I avoid going there. AI can't answer that because you don't know its sources, rather you exchange information with your trusted peers and make judgements based on that. This isn't a money concerning thing, it's just people exchanging info for info as the internet was originally intended to be.
> Microsoft’s plans were reported earlier by The Information.
meta question but is it fair to report someone else’s scoop but bury that fact in the last sentence 4 paragraphs into the article? as a result of this burying Bloomberg gets to the top of HN rather than The Information which got the actual story. If I were that reporter i’d be pissed.
I think the main reason the Bloomberg article got to the top rather than the source (theinformation) is that The Information is beyond a paywall that you can't bypass.
There is a workaround link at the top of the thread. An important difference is that Bloomberg's paywall is intentionally semi-permeable - 'free' article limits, social media referrals, etc. The Information is (mostly?) an intentional 'hard' paywall - the content is only available to subscribers.
It really depends on the publication. In a past life, I was a newspaper reporter in a town with two papers. There were strict guidelines against doing this. If you lost the scoop, well, too bad.
I have seen this behavior more often in Europe but I guess it might be a thing in the US now too - I have been away from the day-to-day journalism scene for a long time.
One nice thing about this is that it might force google to finally do some real world open application of their AI.
Google very likely has the absolute best AI on Earth, and might even be a few steps ahead of Open.AI. However they are extremely coy about it and so far only use AI (public facing) to lightly augment their services rather than be the services. For instance, Google purposely makes assistant act more like a computer taking commands than a human having a conversation.
But we know that Google has at least parity with Open.AI, and it would be a fairly safe bet that they are ahead. We'll see if they have a "mic drop" moment when Bing comes out with this.
This is bad. This is a whole new level of disinformation. ChatGPT regularly produces wrong information that reads very convincingly. This is prematurely adopting bleeding edge technology just for the positive press it will bring. Microsoft, stop immediately.
I'm very interested in the UX decisions they make here. Is it just going to be ChatGPT with a Bing logo or will it be able to intelligently decide when a search engine is better? Will it give results in natural language?
If they sometimes do normal search instead at least that answers how they'll make money
I read earlier that Google has already declared ChatGpt as an existential threat. I am almost sure that Google will drop a huge update to search this year.
Yeah, I think that explainability and safety (e.g. racist, harmful, health related disinformation, etc) will be the huge issues here. Open AI can afford to play fast and loose, but big companies with reputations to protect like Microsoft and Google cannot.
Google will have to match that somehow. I guess the search engine wars are getting hot again.
I think the real challenge is internationalization - it would be a challenge to build GPT-N like models for all the other languages, that work as well as the original one.
Interesting if Google will roll it's own language model for that purpose. Is it possible, that we might get several language models, each one for a specific category of users, or would that approach lead to a loss in generality/quality?
Funny thing here is that in a blind test where the subject doesn‘t know where the result comes from, more than half like the Bing results better. So this seems to be mostly expectation …
ChatGPT already works suficiently well in my own language (Danish). It can probably be improved, but so far it has worked well enough for several news stations to report on ChatGPT while prompting it in Danish. And schools are also worried about its ability to write essays in Danish, among other things.
I had some fun with it trying out a few Dutch dialects. Not perfect but it seems to be able to translate what it knows rather than rely on knowledge in a particular language only. Likeswise, it can translate programs to different languages. Paste some code and let it translate to Rust, kotlin or whatever.
A major advantage Google has is youtube. Most of the content in web for Indian/Asian language is present inside youtube videos. If that content can be transcribed with decent precision youtube can be huge content booster for such language models.
For Polish, which is one of more difficult western languages, chatgpt works flawlessly. As in - it can write in a nicer style than an average person does.
Interestingly, it has no problem with throwing in words from my language when I forget english ones.
The only two issues I found:
- doesn't do good rhymes when I ask it to write poetry
- when I asked it to generate content that has mixed polish and english words (as if written by a pole who spent the last 20 years in US, and replaces some words with their english counterparts), it was unable to do so. It could only write either clear english or clear polish.
Lots of people are using a search engine as that partner in creativity at the moment and it's not that great. That's what makes it a viable alternative.
Pick a domain where you can't trust Google either, like finding which physical store in your area sells the widget you want. Verifying the presence is way cheaper than checking them all in the first place, and Google's results are poisoned by the billion websites trying to sell it to you and ship it.
Your comment reminds me of a comment made almost 2 decades ago by professor regarding wikipedia - it's just random people writing text, it'll never compete with encyclopedia, I will never trust it.
Don't ignore rate of change, recognise it can only improve with time, you're looking at very early system that will quickly be orders of magnitude better.
I think you were half right? I mean, I think it's a case people wanted something for free instead of paying, rather than a quality issue. I've come across a lot of really poor wiki articles, and in general to this day prefer the old encyclopedias or even something like Encarta to Wiki. Add to that the recent 'scandals' if you call them that, eroding trust by some.
Still, it's awesome how Wiki evolves and moves so fast. And most of the articles are pretty great. My wish is that the foundation would spend money on academics and researchers to enhance contributions, rather than a lot of the junk they spend money on today.
Your professor was and is correct in one respect: Wikipedia should not be cited whereas an encyclopedia in print can be cited (though it probably would still be much better to go to the source). At best Wikipedia is a tertiary source and as your professor correctly identified it is just random people writing text. It does compete with encyclopedia but not for that purpose and they are of course well within their rights not to trust it, especially not on subject matter that they are an expert in.
For day to day use it's fine though and the chances of your queries intersecting with a page that has errors on it and/or has been vandalized are relatively remote. But to be skeptical about what you read on WP isn't a bad thing per-se.
@hagbarth: > I still struggle to see chatgpt’s value as a search engine. It works great for me as a partner in creativity, so to say. Both for writing and coding. […]
@mirekrusin: >> Your comment reminds me of a comment made almost 2 decades ago by professor regarding wikipedia - it's just random people writing text, it'll never compete with encyclopedia, I will never trust it.
Don't ignore rate of change, recognise it can only improve with time, you're looking at very early system that will quickly be orders of magnitude better.
@jacquesm: >>> Your professor was and is correct in one respect: Wikipedia should not be cited whereas an encyclopedia in print can be cited. […]
===
The reason why @mikerusin was invoking the Wikipedia analogy was to point out that at a certain moment in time in the heady .com bubble days Wikipedia was an acorn and it was hard to imagine how it could ever grow to compete with the likes of Britannica. I remember the arguments at the time. Some people said "no way" and other said "huh, wouldn't be so sure, just you wait and see". Turns out the latter group were not only correct they were very correct. Wikipedia has entirely supplanted Brittanica and its ilk. I wouldn't even like to guess how more used Wikipedia is than its print rivals.
(And for the purposes for which Wikipedia is used people are aware of its limitations. It's not "At best [] a tertiary source", it's a secondary source for when it comes to citations but if one needs information in a hurry people the information they retrieve from it directly and do not go through the hassle of looking up the primary sources unless they have to. If you use it any other way I'd be very surprised. But this is by-the-by.)
As a response to @hagbarth's pessimism this perfectly echoes the debates we had around the time of the birth of Wikipedia. I wouldn't be so sure. Don't bet against it. Etc. Wikipedia scaled quickly because of crowd-sourcing but a ChatGPT turbo-charged Bing (or whatever) may not need a democratised version of ChatGPT, it may just need to harness the relentless pace of change in the hardware/software sector.
Long story short: I believe @mikerusin is correct and the analogy is a good one, I believe @hagbarth needs to try harder to see the potential here, and I reckon you're responding to a argument that was never put forward! (and I have no idea why I spent 20 minutes going through all this rather than taking a shower and starting my day :/ but such is life …)
> "For example, Wikipedia itself is a tertiary source."
As for the rest of the comment: it's a free world. And if you want to save some time there are 'vote' buttons which allow you to express the same sentiment in a less nuanced way (though I appreciate the effort) which would allow you to start your day on time ;)
“
Wikipedia is not a primary source
Wikipedia avoids describing topics that never have been described before — doing otherwise qualifies as performing original research. Unsourced eyewitness accounts or other unsourced information obtained from personal experience should not be added to articles, as this would cause Wikipedia to become a primary source for the added information (see Wikipedia:Verifiability).
Wikipedia is not a secondary source
Wikipedia does not offer interpretations or analyses that deviate from previously published interpretations and analyses — doing otherwise qualifies as performing original research.
Wikipedia is a tertiary source
Wikipedia summarizes descriptions, interpretations and analyses that are found in secondary sources, or bases such summaries on tertiary sources. Wikipedia illustrates such summaries and descriptions with material that is as close as possible to the primary source(s) on the described topic.
”
> As for the rest of the comment: it's a free world. And if you want to save some time there are 'vote' buttons which allow you to express the same sentiment in a less nuanced way (though I appreciate the effort) which would allow you to start your day on time ;)
What if your searches are to be "creative"? For instance I was testing it on "warehouse optimization". I tried googling that to find ideas, getting lots of SEO spam articles and things not relevant. Then I have to modify the query to hone in on what I actually want. Then wade through articles until I finally find some cool ideas to check out if are applicable to us.
With chatGPT I wrote a few sentences about our warehouse and asked it how to optimize some part of our process. Then it spat out 5 suggestions that were quite tailored. The same as I found when googling, but instantly in the correct context without me having to skim loads of google results and try various queries to avoid spam.
And the less experienced you are in IT, the better it's going to be to ask chatGPT rather than a search engine. Quickly parsing web pages and keyword optimisation are skills not everybody has.
Sure there's some prompt optimisation on the GPT side but it's nowhere as complex as navigating through Google and the results are just there.
I was just thinking about this the other day. We have built quite an enormous tacit skill & muscle memory for being able to search & skim articles and find something useful on Google. Our own neural nets have been trained for years on Google results, and that's why it feels so natural and easy to us. In reality, it's enormously difficult task, and also impossible to teach to someone.
I use it for coding every day now. I ask it for examples of such and such implementation, in a way that's impossible to get through documentation or even by reading sample code on the internet. For example, I'll ask:
"How can I code a mock for a boto3 dynamodb client, using unittest.mock, using a patching context manager"
And it will give me a highly relevant example, often right or close enough.
I will occasionally ask to come up with a function based on requirements but I don't find that nearly as useful as using it as super advanced search engine.
Irrespective of whether major search engines might use language models, fake web sites will use them.This could make it increasingly difficult to find valid information and maybe precipitate some sort of arms race between algorithms that detect algorithms
A pathetic scenario but somehow consistent with the rules (or lack thereof) of the game
It will be an interesting arms race, but I'm hopeful for truth there.
I can imagine good enough AI being able to spot truth even better than what humans do - by veryfing sites and commenters with sources of real information to estimate their credibility.
E.g. in a theme similar to Page Rank, you could have an AI that has some sites as a source of objective truth (Wikipedia, science journals, reputable sources of news etc), and then use that as a basis of estimating trustworthiness of a material.
Also, AI could find, for a given subject, opposing opinions, and estimate which ones are possibly fake, and which ones are real.
In essence - do what current fact-checkers do, but for every single website and comment in existence.
Sure, but humans face the same issue when fact checking. Even moreso because we can’t browse the whole of humanity’s knowledge as well as machines can.
What did you find fishy about them? They seem roughly the same, aside from the countries involved with the projects having a bit more local information.
The Dutch and German are a lot longer than the quite short English version. But ... that's just a matter of organisation: in the English the editors chose to make separate "Nord stream {1,2}" articles, in other languages they folded it in one article. On the German one in particular it's just two huge sections.
In short, it's fishy in the same way that bread tastes like fish: not at all.
I have absolutely no idea whether it applied to that case, too, but I've found that Wikipedia's language mapping sometimes breaks down when there's no easy 1:1 mapping between articles in differing languages.
> I can imagine good enough AI being able to spot truth even better than what humans do
The thing that makes me question that is the data that is used to train those models to begin with. To disect truth on the internet, can you use the internet as a source of truth to train it?
> some sites as a source of objective truth (Wikipedia, science journals, reputable sources of news etc)
the irony is that finding "objective truth" is a very non-trivial human game but in all cases costly. E.g journalism has been decimated after losing their traditional ad revenue. Wikipedia and science journals survive because they rely on informal and formal public funds etc.
It seems to me (or I hope) this isn't really much different from the current situation: we already have a ton of useless quasi-content made up only for search engines, just look for something like "best android phone" or whatever.
Search engines will have to rely more on signals outside the content, such as links from other authoritative sources, but it does not look like a qualitatively different world.
I agree it is a quantitative difference at its core but cheap automation can dramatically lower the signal to noise. Crossing a certain quality threshold may eventually precipitate binary behavioral changes for users (i.e. conclude that certain online tools are unusable / untrustworthy)
I also agree that authoritative sources become critical. Yet those typically rely on very human assessments (with their own pitfals and controversies) and in any case much more slow / costly to develop.
How exactly this all will play out is not clear (to me). But the naive technosolutionism of deploying "AI at scale" and believing that it will just work as advertised seems misplaced. The human condition is very reflexive.
936 comments
[ 3.6 ms ] story [ 322 ms ] threadIn a very narrow niche there may not be many documents to pick from, either.
I don't think you can just automate this away in the context of generalized search. Search has to fulfill every niche; that seems like an indefensible position (strategically speaking, not in the moral sense). How can you benchmark bias in every niche? Your benchmarks show your reasoning is sound; what about the premises you're reasoning from? In the context of something narrowly scoped like a customer service bot, it makes sense to me how you could build expertise in constraining the model's output. But in terms of everything?
But I'll admit I don't have a crystal ball, happy to eat my words if they can operate with enough traffic and got long enough to attract the attention of spammers, and still keep them at bay. I think this space is stagnant and needs to be shaken up, and that chat interfaces have potential, so I'm not trying to be a hater. I just think this is gunnuh be a very difficult aspect.
It's funny because paper[1] from Google that started transformers and LLMs is called "Attention is all you need". And I agree that you can easily incorporate ads into chat, just ask ChatGPT to do it in a subtle way for you and you will see how it will incorporate ad into response.
1. https://arxiv.org/abs/1706.03762
Examples:
* When you ask for recommendations, it gives you x branded option or maybe the "best result" with an advertised alternative.
* When you search for symptoms, it recommends name brands instead of generics.
* When you ask for recipes, it directs you to <advertised cooking site>.
* When you ask for poetry in the style of Longfellow, it recommends an audiobook as "something you might enjoy".
* When you want to know about what happened to <recent celebrity that died>, it says "they died in a high speed chase according to <celebrity rag>. Here are some pictures and links to an article with more information"
* Platform integrations, like linking to the Google/windows store directly.
Many of these are already done by primitive assistants like Alexa. Going beyond that, you can copy pretty much everything done by influencer marketers in a chatbot.
I need facts about products, not a salesman convincing me to buy their products.
I could be nice to have an AI ask me questions about what I hope to use it for, what kind of space I have, and what my budget is, and then give me some recommendations.
Nadella is the Gates/Jobs to Pichai's Ballmer/Sculley. Microsoft is playing 4D chess with their acquisitions and long-term vision; Google is floundering amidst perverse leadership incentives, inaction, and its non-diversified revenue stream.
Then you can send that to another service; pushing math to wolfram alpha (etc) makes chatgpt suddenly give perfect math answers and it works (it takes a bunch of hacks to ignore it’s confident lying part of answering equations). When we it names a person or location, it shows Wikipedia or maps or both etc.
In this way you use it’s bluffing powers without assuming (or even reading) it’s content; it returns articles etc (which can be wrong as well of course but at least you might know if it’s reputable or not).
I can easily see how bing or google would use something like this to search and return very relevant content, the queries might not be better than some people can create themselves with effort. However many people cannot create good queries and chatgpt creates good queries from extremely little input.
More importantly; it’s a conversation; you can fine tune the results. You can tell it that you meant the programming language, not oxidation when it shows completely bonkers results.
This will work and will work very well as far as I have been able to try. I cannot go further as I have no easy access to search engines without getting blocked immediately for automated querying.
Not sure how this is looking cost wise: chatgpt at scale will even hurt a company like MS for now. Having a million techies use it for some experimentation is a bit different than a billion+ users searching whatever.
First type is info widgets which allow it access information outside of itself, like using a calculator by making it output something like {calc 1+1}, search wikipedia by {wiki something}, see the results of such a query and use that info in its output to the user.
Second type is UI widgets that allow it to display different types of things to the user. There is already a kind of widget in ChatGPT, which is display of Markdown. But it's easy to imagine widgets for say showing coordinates on a map, or visualizing something else.
Since there is a large number of such useful widgets, I'm guessing there will be an app store for them where developers can write their own widgets to extend it and make revenue based on how much they are used.
I do agree with the overall point though. If you understand when to use it and when it's more likely to give you nonsensical answers, it can save a huge amount of time. But when I ask it about a topic that I don't know enough about to immediately verify the answer myself I'm forced to double check the answers for validity, which kind of defeats the purpose.
The best queries to ChatGPT are cases where I know what the answer should look like, I just forgot the syntax or some details. Bash scripts or Kubernetes manifests are examples here, I know them, I just keep forgetting the keywords because I only touch them every few weeks.
And don't get me started about asking ChatGPT about more general topics in e.g. economics or finance. What you get is a well-written summary of popular news and reddit opinions, which is dangerous if it's presented as "the truth" - The big mistake here is that the training procedure assumes that the amount of data correlates with correctness, which isn't true for many topics that involve politics or similar kinds of incentives where people and news spread what conveniently benefits them and gets clicks.
Humans take a lot of shortcuts (such as believing more easily the same facts presented with a confident tone) and the "firehose of bs" exploits it: this was already the case before generative AI, but AI amplifies the industrial-scale imbalance between the time needed to generate partially incorrect data and the amount of time/energy required to validate.
> Salgat 8 days ago
> The problem with ML is that it's pattern recognition, it's an approximation. Code is absolute, it's logic that is interpreted very literally and very exactly. This is what makes it so dangerous for coding; it creates code that's convincing to humans but with deviations that allow for all sorts of bugs. And the worst part is, since you didn't write the code, you may not have the skills (or time) to figure out if those bugs exist
https://news.ycombinator.com/item?id=34140585
I think the real power of these bots will be to lead us down this path, as opposed to it doing everything for us. We can ask it to justify and explain its solution and it will do its best. If we're judicious with this we can use it to build our own understanding and just trash the AI's output.
ChatGPT can completely cut out the online search and give an answer directly about things like compiler errors, and elaborate further on any detail in the answer. I think that 2-3 further GPT generations down the line it will be worth the time for some applications.
The problem I see is less the overall quality of responses but people overestimating on where it can be used productively. But that will always be a problem with new tech, see Tesla drivers who regularly take a nap in the car because it didn't crash yet.
The trick is to use it as an LLM and not a procedural, transactional data set.
For instance, “how do I create a new thread in Python”. Then ask “how do I create a new thread in Python 3.8”. The answers will (probably) be different.
Any interface to chatgpt or similar can help users craft good prompts this way. It just takes thinking about the problem a little differently.
One wildly inefficient but illustrative approach is to use chatgpt itself to optimize the queries. For the Python threading example, I just asked it “ A user is asking a search engine ‘how do I create threads in Python’. What additional information will help ensure the results are most useful to the user?”.
The results:
> The user's current level of programming experience and knowledge of Python
> The specific version of Python being used
> The desired use case for the threads (e.g. parallel processing, concurrent execution)
> Any specific libraries or modules the user wants to use for thread creation
> The operating system the user is running on (as this may affect the availability of certain threading options)
So if you imagine something like Google autocomplete, but running this kind of optimization advice while the user builds their query, the AI can help guide the user to being specific enough to get the most relevant results.
One concern I have goes like this: I seriously doubt that current LLMs are capable of anything that could really be called an understanding of the significance of the version number[1], but I would guess that it characterizes the various Python-with-versions strings it has seen as being close[2] so I can imagine it synthesizing an answer that is mostly built from facts about Python2.7. With a simple search engine, you can go directly to checking the source of the reply, and dig deeper from there if necessary, but with an LLM, that link is missing.
[1] The fact that it listed the version as being a factor in reply to your prompt does not establish that it does, as that can be explained simply by the frequency with which it has encountered sentences stating its importance.
[2] If only on account of the frequency with which they appear in similar sentences (though the whole issue might be complicated by how terms like 'Python3.8' are tokenized in the LLM's training input.)
GPT-3's response:
> The language and version targeted here is Python 3, as indicated by the use of ThreadPoolExecutor from the concurrent.futures module. This is a module added in Python 3 and can be installed on earlier versions of Python via the backport in PyPi. The advice is tailored to Python 3 due to the use of this module.
That's imperfect, but I'm not trying to solve for Python specifically... just saying that the LLM itself holds the data a query engine needs to schematize a query correctly. We don't ChatGPT to understand the significance of version numbers in some kind of sentient way, we just need it to surface that "for a question like X, here is the additional information you should specify to get a good answer". And THAT, I am pretty sure, it can do. No understanding required.
1. https://stackoverflow.com/questions/30812747/python-threadin...
The question you posed GPT-3 here is a rather leading one, unlikely to be asked except by an entity knowing that the version makes a significant difference in this context, and I am wondering how you envisage this being integrated into Bing.
One way I can imagine is that if the user's query specified a python version, a response like that given by GPT-3 in this case might be used in ranking the candidate replies for relevance: reject it if the user asked about python 2, promote it if python 3 was asked for.
Another way I can imagine for Bing integration is that perhaps the LLM can be prompted with something like "what are the relevant issues in answering <this question> accurately?" in order to interact with the user to strengthen the query.
In either case, Bing's response to the user's query would be a link to some 3rd-party work rather than an answer created by the LLM, so that would answer my biggest concern over being able to check its veracity, though its usefulness would depend on the quality of the LLM's reply to its prompts.
On the other hand, the article says "Microsoft is betting that the more conversational and contextual replies to users’ queries will win over search users by supplying better-quality answers beyond links", apparently saying that they envision giving the user a response created by the LLM, which brings the question of verifiability back to center stage. Did you have some other form of Bing-LLM interaction in mind?
While ChatGPT spews out complete nonsense most of the time. And the dangerous part is that that nonsense looks very reasonable. It gets very frustrating after some time, because at first you are always happy that it gave you a nice solution, but then it's not usable at all.
A "bullshit sandwich" if you will.
When one tells people this we get the reply "but so do random blogs! or reddit comments!". Well yes, but they're just random blogs and reddit comments, often peppered with syntactic and spelling mistakes, non sequiturs, and other absurdities. Nobody would take them seriously.
ChatGPT is very different. It doesn't say "this random redditor says this, and this other random redditor says the exact opposite, so IDK, I'm just a machine, please make up your mind".
What it says is "this is the absolute truth that I, a 'large language model', have been able to extract from the vast amount of information I have been trained on. You can rely on it with confidence."
I'm sorry to sound hyperbolic but this cannot end well.
The other thing that I've had success with is asking for references for the information, which will often link you to the relevant docs.
I could imagine losing many hours from a ChatGPT answer. And if you have to go through the trouble to verify everything it says to make sure it's not just making crap up, then imo it loses much value as a tool.
Same with copilot. Sometimes it's ludicrously wrong in ways that sound good. I still have to do my job and make sure they are right. But it's right or right enough to save me significant effort at least 75% of the time. Right enough to at least point me in the right direction or inspire me at least 90% of the time.
How is ChatGPT enabling this? All of that is very possible without ChatGPT. The damaging part is deciding to do it.
In this particular case, ChatGPT will write something nonsensical which people will accept more easily because of the way it is written. This is inevitable and extremely dangerous.
Conversely, I just ask ChatGPT to extol the virtues of leaded gasoline, and instead I got a lecture on exactly why and how it's extremely harmful.
Some are but not many. And then there's the amount. That's the crux of the matter. Have you seen that Aza Raskin interview where he posited one could ask the AI to write a thousand papers citing previous research against vaccines and then another thousand pro-vaccines? No human can do that.
People are just as good at making up convincing sounding nonsense.
Perhaps as you just did, as I can find no one actually "injecting themselves with bleach."
The overall point stands: the difference between reading something dumb and doing that dumb thing is what it means to have agency. I personally don't think we should optimize the world 100% to prevent people who read something stupid from doing that stupid thing.
Or, if that's the path we're going to take, maybe we should first target things like the show Ridiculousness before we start talking about AI. After all, someone might do something dumb they see on TV!
Ingesting, injecting, that’s pretty similar. Nobody needs to make anything up there.
https://www.justice.gov/usao-sdfl/pr/leader-genesis-ii-churc...
I had a similar experience with a shell command. Searched google, looked at a few posts, wasnt exactly what I needed but close. Modified it a few times and got it working. Went to save the command in a markdown file and when I explained what the command did, copilot made a suggestion for it. It was correct and also much simpler.
It went from taking 5-10 minutes to stumble through something just so I could do the thing I really wanted to do, to finding the answer instantly all from within the IDE. Can keep you in flow.
Say incorrect stuff authoritatively and people will think you're right.
It happens to me all the time. I can't structure accurate information in a better way then some bullshit artist can spit off what they imagine to be real so everyone walks away believing in their haughty nonsense.
ChatGPT exploits that phenomena which is why it sounds like some overly confident oblivious dumb dumb all the time. That's the training set.
Almost once a week I'll go through a reddit thread and find someone deep in the negatives who has clearly done their homework and is extraordinarily more informed than anyone else but the problem is everyone else commenting is probably either drunk or a teenager or both so it doesn't matter.
Stuff is hard and people are mostly wrong. That's why PhDs take years and bars for important things are set so high
For sure. But you have to compare against alternatives. What would that be? Posting to stack overflow and maybe getting a helpful reply within 48 hours.
> I could imagine losing many hours from a ChatGPT answer.
Dont trust it. Verify it.
We expect to ask a question and get a good answer. In reality we should leverage how cheap the answers are.
This 100%.
ChatGPT can play an interesting role by separating duties in a process of productivity. ChatGPT can generate tons of true/false suggestions very fast and understandable by humans. Sometimes this helps a lot.
1. Generating (or simply repeating) obvious ideas in a domain that I am not an expert in
2. (With some prompting) Generating creative ideas in a domain that I am familiar with
3. Generating obvious ideas in a domain I'm familiar with when I'm too tired to think or preoccupied
Not only do you get a productivity boost by being an editor but it also complements human energy cycles
One pretty consistent way to see this is to ask for various very simple designs like a n-bit adder, it will almost always do something logically incorrect or syntactically incorrect with the carry in or carry out
And it’s so much faster than posting on stack overflow or some irc. It doesn’t abuse you for asking dumb questions either.
Add documentation to this method : [paste a method in any language]
For me the results have been impressive. It’s even more impressive if you are not English speaking because it explains what the code does but also translates your domain terms in your own language.
More than code generation I see a really concrete application in having autogenerated and up to date documentation of public methods. It could be generated directly in your code or only by your IDE to help you in absence of human written documentation.
Other interesting things it can does is basic code review by proposing a « better » code and explaining what and why it changed something.
It can also try to rewrite a given code in another language. I haven’t tried a lot of things due to the limitations in response size but for what I tested, it looks like it is able to convert the bulk of the work.
While I’m not really convinced by code generation itself (a la copilot) I truly think that GPT can be a really powerful tool for IDE editors if used cleverly, especially to add meaning to unclear, decade old codebases from which original contributors are long gone.
And knowing that what is hard is not writing but reading code, I see GPT to be a lot more useful here than helping writing 10 lines in a keystroke.
A common advice for documentation is "why not how", I'm not sure you can do "why" by looking at the "how".
You can do javadoc style params, and there's some value there,but not much.
You are right. It’s the rule when you write the doc.
But when you are let alone in an unknown codebase, having your IDE summarize the "what" in the auto completion popup could be really useful. Especially in codebases with wrong naming conventions.
The "why" is important for inline comments, but for function and method comments I think the biggest is neither "why" nor "how", but "what". As in, "what does this method do?" especially with regards to edge cases.
I tried a few methods just now; it gives okay-ish docs. Lots of people don't write great comments in the first place, so it's about on-par. Sometimes it got some of those edge cases wrong though; e.g. a "listFiles()" which filters our directories and links isn't documented as such, but then again, many people wouldn't document it properly either.
I’m not sure that fabricated nonsense would actually make Bing’s results any worse than they are today.
“It’s okay I don’t mind verifying all these answers myself” is an odd sort of sentiment, and also inevitably going to prove untrue in one sense or another.
Or, more likely, I would just lazily assume everything is fine and use it anyway, until one day the unexamined flaws destroyed something costly in a manner difficult to diagnose because I didn't bother to actually understand what it was doing.
There really should be more horror at the imminent brief and temporary stint of humans as editors, code reviewers, whatever, over generative AI mechanisms (temporary because that will be either automated or rendered moot next). I'm unaware of any functional human societies that have actually reached the "no one actually has to work unless they want to do so, because technology" state, so this is an interesting transition, for sure.
Well yeah, I'm right there with you. But that feels a lot like any software, open or closed source. Human programmers on average are better than AI programming today, but human programmers aren't improving as fast as AI is. Ten years from now, AI code will be able to destroy your data in far more unpredictable and baroque ways than some recent CS grad.
> I'm unaware of any functional human societies that have actually reached the "no one actually has to work unless they want to do so, because technology" state, so this is an interesting transition, for sure.
This is a really interesting thought. Are we seeing work evaporate, or just move up the stack? Is it still work if everyone is just issuing natural language instructions to AI? I think so, assuming you need the AI's output in order to get a paycheck which you need to live.
Then again, as a very long time product manager, I'm relatively unfazed by the current state of AI. The hundreds of requirements docs I've written over decades of work were all just prompt engineering for human developers. The exact mechanism for converting requirements to product is an implementation detail ;)
It gives you answers with 100% confidence and believable explanations. But sometimes the answers are still completely wrong.
Tuning the temperature and topP parameters you can also make the model avoid low probability completions (useful for less creative use cases where you need exact answers).
A proxy for confidence in what exactly?
Language models represent closeness of words, so a high probability would only express that those words are put together frequently in the corpus of text; not that their meanings are at all relevant to the problem at hand. Am I wrong?
This is because I'm operating under the assumption that sequences of words that appear often in the training set are more likely to represent something correct (otherwise you might as well train on random words). This only holds if the training set is big enough that you can estimate correctly (e.g. if the training set is small a very rare/wrong phrase may appear very often).
Maybe confidence was the wrong word, but for this kind of questions I would trust a high-probability answer way more than a low one. For questions belonging to very specific subjects, where training material is scarce, the model might have very skewed probabilities so they become less useful.
Maybe, as long as you're aware that this is the same kind of correctness that you get from looking at Google's first search results (the old kind of organic pages, not the "knowledge graph", which uses an different process - precisely to avoid being spammed by SEO) i.e. "correctness by popularity".
This means that the content that is more replicated will be considered more true by the system, regardless of its connection to reality or its coherence with the rest of the knowledge in the system. And you know what they say about big enough lies that you keep repeating millions of times.
I understand the problem, but what better way do we currently have to measure its connection to reality? At least from a practical point of view it seems that LLMs have achieved way better performance than other methods in this regard, so repeatedness doesn't look like that bad a metric. Or rather, it's the best I think we currently have.
We can consider its responses to a broader range of questions than those having an unambiguous and well-known answer. Its propensity for making up 'facts', and for fabricating 'explanations' that are incoherent or even self-contradictory shows that any apparent understanding of the world being represented in the text is illusory.
This line of thought is a distraction, anyway. The likelehood that GPT-3 will do as well as a search engine on topics where there is an unambiguous and well-known answer does little to address the more general concern.
Take Stable Diffusion for instance: it can interpolate a painting from that huge dataset it has, and sometimes output a decent result that may look like what a good artist would do. But it doesn't have any kind of "creative process". If it tells you "I chose this theme because it reflects this deep societal problem", it will just be pretending.
It may not matter if all you want is a nice drawing, but when it's about, say, engineering, that's quite different.
ChatGPT models are language models; they represent closeness between text utterances. It works by looking for the chains of words most similar or usually connected to those indicated in the prompt, with no understanding of what those words mean.
As a metaphor, think of an intern who every morning is asked to buy all the newspapers in paper form, cut out the news sentence by sentence, and put all the pieces of paper in piles grouped according to the words they contain.
Then, the director requests to write a news item on the increase in interest rates. The intern goes to the pile where all the snippets about interest rates are placed, will randomly get a bunch of them, and write a piece by linking the fragments together.
The intern has a PhD in English, so it is easy for them to adjust the wording to ensure consistency; and the topics more talked about will appear more often in the snippets, so the ones chosen are more likely to deal with popular issues. Yet the ideas expressed are a collection of concepts that might have made sense in their original context, but have been decontextualized and put together pell-mell, so there's no guarantee that they're saying anything useful.
No, it does not work that way. That’s how base GPT3 works. ChatGPT works via RLHF and so we don’t “know” how it decides to answer queries. That’s kind of the problem.
Google will probably build the same thing. Stackoverflow can suffer though...
I really wonder what impact ChatGPT will have on search engines. I could imagine that the first 4 pages of Google/Bing results end up being autogenerated stuff, and it will just make it harder to find trustworthy information.
It could generate a python script that counted the days of the week with the letter T, but still insisted that Sunday had a T
Edit: Scratch that. I just tried again and now it says that saturday doesn't have a T in it
For example recently I asked it for the best way to search in an mbox file on arch Linux. It proceeded to recommend a number of tools including mboxgrep. When I asked how to install it on arch it gave me a standard response using the package manager, but mboxgrep is not an arch package. It isn't even an aur package. It requires fetching the source and building it by yourself(if I remember correctly one has to use an older version of gcc too). None if it was mentioned by chatgpt.
This is not the first time BTW, there was another software it recommended that Debian doesn't know about, when I asked it another time.
I think a new implementation of ChatGPT is worth exploring though, one that cites sources and gives links to further information, and also one that has the ability to somehow validate it's responses for accuracy.
God forbid!
Python lang docs are "kind-of-OK" but when someone raves about them I'm left scratching my head. Information is not always well-organized, examples are hit-and-miss, parameter and return types not always clear, etc.
Referencing docs as a programmer is generally a nightmare and a time sink, and it's the one use case where ChatGPT is slowly becoming indispensable crutch for me. I can ask for very specific examples that are not included in the docs, or that cannot be included in the docs, for example combinatorial in nature: "how can I mock this AWS SDK library by patching it with a context manager"? Occasionally it will hallucinate, but even if it gets it 8/10 times right - and it's higher than that in practice - it will prove revolutionary at least for this use case.
I find this tool so useful, that I scratch my head when I read about how dismissive some people are of it.
I agree that pretty much all AWS documentation is woeful, and it's a travesty that the service is so expensive yet its documentation is so poor. I would gladly dump AWS and never use it again, as I hate paying top-dollar to decipher the AWS doc team's mistakes (not to mention that they are unresponsive to bug reports and feedback).
My point was made more in jest, and supposed to point out the irony of the communities' changing expectations of what documentation should be like. I predict that in a few years we'll be circling back to prioritizing writing software documentation well. (Kind of like how everybody was hating on XML for the past 20 years and it's now having a renaissance because it actually does what it's supposed to well very well.)
The function does not exist, it's entirely made up.
What's weirder is that when I told it that the answer was wrong it provided a different solution that was correct.
Where do you think it got the information?
But the holy grail would be if it could write all my unit tests...
ChatGPT (or a similar product) focuses on solving users problem interactively. No ads, no going to another website etc. How would you make money from a search engine?
I was hoping a simple, paid model to start with. Over time, as the LLMs become commodity (200GB, runnable on Intel/AMD), ship it as part of the OS and other devices.
I’m joking of course but somebody is going to this
Current search ads display options and let you filter through which is neat, but ChatGPT like agent will shred the original content and make it invisible to distinguish which is ads, and worse false advertising, which is not.
I think let ChatGPT to infer the attribution automatically and do a revenue sharing with original link owner will be a better option.
Already done :)
https://future.attejuvonen.fi/
1 https://kagi.com/
It’s not ChatGPT as they judged that too low quality (with the lying and everything) and instead something that actually gives sources.
I have not used it yet, so I don’t know what the quality is.
Correction: it goggle (beta), not google (beta).
(Disclaimer: I work at Brave)
I just checked...
I'm confused.
Your repository can be secured by enterprise grade post-modern Security-as-Code(tm) frameworks pioneered by SaCicorp(R) backed by YC '24-27 and other leading partners that enforces an indomitable security posture in any threat landscape.
"Wow!! What is that?"
regular expression
Microsoft makes money from services so it doesn't have to make money from a search engine.
Google, on the other hand, makes money from ads so it doesn't have to make money from services.
If Microsoft can take Google's ad revenue away they no longer have to compete against Google's free services.
The converse also holds but it may be easier for Microsoft to weaken Google.
I think there is no technical barrier to add Ads to a chat bot. It's even more deceptive when promotion campaigns are embbeded into text contexts implicitly. It would be much more dangerous and harder to block than, let's say a DIV on a webpage.
What prevents search engines from doing this nowadays is culture and law, not technical aspects.
<chatgpt summary of cards>
Links to cards mentioned in the summary.
… now you can make money on the links directly, perhaps have sponsorships influence the recommendations, and have a strong signal of intent to purchase you can try to monetize later.
“Hey bing what’s a good hotel in downtown Montreal?” -> same
As your conversation evolves with the bot, targeted ads could be shown with the same (or better) level of intent data available based on the human’s input.
I, for one, welcome our new robotic overlords: https://future.attejuvonen.fi/
Hey Microsoft, how about instead of adding chatgpt, filter the results so Bing actually works as a search engine? That is, select only the search results that contain the search terms, rather than returning the whole internet in an order somewhat influenced by the search terms.
Bing is just catering to their core users here, turns out this is an excellent idea for consuming and serving adult content.
Try out Bing vs Google for this purpose, and prepare to be amazed.
Anyway, Bing being a pr0n search engine is equal parts meme and reality.
I like how Microsoft/Bing don't take themselves too seriously since usurping Google in this regard is a fool's errand at this point.
> What is the world record for crossing the English Channel entirely on foot?
It replies with swimming and crossing on ferries.
https://www.google.com/search?q=What+is+the+world+record+for...
If the endpoint is not the user clicking in a link then why should these companies give away all that value?
Stackoverflow has value for these models, the others will just make the nonsense responses worse.
I live from the content I write. It's not fluff. Some of it comes from weeks-long email conversations with government officials. It takes a lot of research and help from experts I have long-standing relationships with.
If search engines serve that information but deny me the traffic, the website dies, as does the source of the information.
I can deal with lazy copywriters just rephrasing my work because the original still outranks them, and I have legal options to deal with them.
I can't do anything if Google - over 80% of of my traffic - decides to proxy my content and starve me of my income.
And it's probably not just big companies we will have to worry about because at least they can be somewhat regulated and they are in the public eye. The other "threat" in the future is the "distributed" AI when people can run their own personal AI assistants that could collect information for them by any means (singing up to to websites, e-mailing, calling people, talking to other AI agents) and with filtering out ads and sponsored content. At that point probably everything worthwhile will be paywalled and the "SEO" game will be to convince/trick these AIs to sing up / pay for your content.
I agree this worsens it.
Google sometimes shows snippets from my website as a direct answer to a query (in a collapsible box). If the out-of-context answer satisfies the user, then I get no traffic.
It gets worse with current state of common websites full of mailing, ads and cookie popup, moreover account-walled content, even with adblocker.
For the record, I run a pretty lean, annoyance-free website. It looks the same without an ad blocker. It uses Plausible for analytics. No newsletters, no annoyances. I still get shafted.
Information should be retrievable without all kinds of nonsense personalization and ads, and in milliseconds, not 10 seconds. The 99.999% of web content made between 2000-now is not whatever virtuous content you claim to have struggled to create, but a bunch of bloat that just wastes the user's time, and most of the time it's not even a good answer to the question, but just the exact same paraphrased answer from several other commercialized blogs. Your doomsday scenario here would be the perfect justice, and you will be one of the _very few_ innocent victims of it. Of course it won't be so simple, anyone like Google would find a way to make the user experience insufferable. I don't see a place for monetization on the future web, it will just be a bunch of people exchanging information, like where are the bad guys with guns and should I avoid going there. AI can't answer that because you don't know its sources, rather you exchange information with your trusted peers and make judgements based on that. This isn't a money concerning thing, it's just people exchanging info for info as the internet was originally intended to be.
meta question but is it fair to report someone else’s scoop but bury that fact in the last sentence 4 paragraphs into the article? as a result of this burying Bloomberg gets to the top of HN rather than The Information which got the actual story. If I were that reporter i’d be pissed.
link to real source https://archive.ph/o/1ChFk/https://www.theinformation.com/ar...
Good question about journalistic ethics.
I have seen this behavior more often in Europe but I guess it might be a thing in the US now too - I have been away from the day-to-day journalism scene for a long time.
2023 is going to be an interesting year.
PaLM (540B parameters): https://ai.googleblog.com/2022/04/pathways-language-model-pa...
Med-PaLM: https://archive.is/oC5Gj (https://twitter.com/vivnat/status/1607609299894947841)
Google very likely has the absolute best AI on Earth, and might even be a few steps ahead of Open.AI. However they are extremely coy about it and so far only use AI (public facing) to lightly augment their services rather than be the services. For instance, Google purposely makes assistant act more like a computer taking commands than a human having a conversation.
But we know that Google has at least parity with Open.AI, and it would be a fairly safe bet that they are ahead. We'll see if they have a "mic drop" moment when Bing comes out with this.
I'm very interested in the UX decisions they make here. Is it just going to be ChatGPT with a Bing logo or will it be able to intelligently decide when a search engine is better? Will it give results in natural language?
If they sometimes do normal search instead at least that answers how they'll make money
As long as I ALWAYS know when a GPT derived match is shown to me.
Preferably with a diagnostic akin to sql EXPLAIN.
Useful to be able to flag incorrect info although this is open to trivial subversion.
I think the real challenge is internationalization - it would be a challenge to build GPT-N like models for all the other languages, that work as well as the original one.
Interesting if Google will roll it's own language model for that purpose. Is it possible, that we might get several language models, each one for a specific category of users, or would that approach lead to a loss in generality/quality?
So small languages are not necessarily a problem.
Interestingly, it has no problem with throwing in words from my language when I forget english ones.
The only two issues I found: - doesn't do good rhymes when I ask it to write poetry - when I asked it to generate content that has mixed polish and english words (as if written by a pole who spent the last 20 years in US, and replaces some words with their english counterparts), it was unable to do so. It could only write either clear english or clear polish.
And it works for generating text that would be tedious to do manually.
But for searching information, I can’t bring myself to trust it.
Don't ignore rate of change, recognise it can only improve with time, you're looking at very early system that will quickly be orders of magnitude better.
Still, it's awesome how Wiki evolves and moves so fast. And most of the articles are pretty great. My wish is that the foundation would spend money on academics and researchers to enhance contributions, rather than a lot of the junk they spend money on today.
https://en.wikipedia.org/wiki/Wikipedia:Academic_use
For day to day use it's fine though and the chances of your queries intersecting with a page that has errors on it and/or has been vandalized are relatively remote. But to be skeptical about what you read on WP isn't a bad thing per-se.
@mirekrusin: >> Your comment reminds me of a comment made almost 2 decades ago by professor regarding wikipedia - it's just random people writing text, it'll never compete with encyclopedia, I will never trust it.
Don't ignore rate of change, recognise it can only improve with time, you're looking at very early system that will quickly be orders of magnitude better.
@jacquesm: >>> Your professor was and is correct in one respect: Wikipedia should not be cited whereas an encyclopedia in print can be cited. […]
===
The reason why @mikerusin was invoking the Wikipedia analogy was to point out that at a certain moment in time in the heady .com bubble days Wikipedia was an acorn and it was hard to imagine how it could ever grow to compete with the likes of Britannica. I remember the arguments at the time. Some people said "no way" and other said "huh, wouldn't be so sure, just you wait and see". Turns out the latter group were not only correct they were very correct. Wikipedia has entirely supplanted Brittanica and its ilk. I wouldn't even like to guess how more used Wikipedia is than its print rivals.
(And for the purposes for which Wikipedia is used people are aware of its limitations. It's not "At best [] a tertiary source", it's a secondary source for when it comes to citations but if one needs information in a hurry people the information they retrieve from it directly and do not go through the hassle of looking up the primary sources unless they have to. If you use it any other way I'd be very surprised. But this is by-the-by.)
As a response to @hagbarth's pessimism this perfectly echoes the debates we had around the time of the birth of Wikipedia. I wouldn't be so sure. Don't bet against it. Etc. Wikipedia scaled quickly because of crowd-sourcing but a ChatGPT turbo-charged Bing (or whatever) may not need a democratised version of ChatGPT, it may just need to harness the relentless pace of change in the hardware/software sector.
Long story short: I believe @mikerusin is correct and the analogy is a good one, I believe @hagbarth needs to try harder to see the potential here, and I reckon you're responding to a argument that was never put forward! (and I have no idea why I spent 20 minutes going through all this rather than taking a shower and starting my day :/ but such is life …)
Sorry, but that's incorrect. Wikipedia mostly uses secondary sources, rarely primary ones that makes it for the most part a tertiary source.
https://en.wikipedia.org/wiki/Wikipedia:Primary_Secondary_an...
> "For example, Wikipedia itself is a tertiary source."
As for the rest of the comment: it's a free world. And if you want to save some time there are 'vote' buttons which allow you to express the same sentiment in a less nuanced way (though I appreciate the effort) which would allow you to start your day on time ;)
I'd like to see empirical analysis substantiating this claim. Not saying you're wrong, just saying I'd like some hard data.
Hmm, having said that, look at what I found: https://en.wikipedia.org/wiki/Wikipedia:Wikipedia_is_a_terti...
> As for the rest of the comment: it's a free world. And if you want to save some time there are 'vote' buttons which allow you to express the same sentiment in a less nuanced way (though I appreciate the effort) which would allow you to start your day on time ;)fair point :)
With chatGPT I wrote a few sentences about our warehouse and asked it how to optimize some part of our process. Then it spat out 5 suggestions that were quite tailored. The same as I found when googling, but instantly in the correct context without me having to skim loads of google results and try various queries to avoid spam.
Sure there's some prompt optimisation on the GPT side but it's nowhere as complex as navigating through Google and the results are just there.
"How can I code a mock for a boto3 dynamodb client, using unittest.mock, using a patching context manager"
And it will give me a highly relevant example, often right or close enough.
I will occasionally ask to come up with a function based on requirements but I don't find that nearly as useful as using it as super advanced search engine.
A pathetic scenario but somehow consistent with the rules (or lack thereof) of the game
I can imagine good enough AI being able to spot truth even better than what humans do - by veryfing sites and commenters with sources of real information to estimate their credibility.
E.g. in a theme similar to Page Rank, you could have an AI that has some sites as a source of objective truth (Wikipedia, science journals, reputable sources of news etc), and then use that as a basis of estimating trustworthiness of a material.
Also, AI could find, for a given subject, opposing opinions, and estimate which ones are possibly fake, and which ones are real.
In essence - do what current fact-checkers do, but for every single website and comment in existence.
Look for english version of article about Nord Stream... Compare with any other langage (no need to know these other languages).
There is something fishy going on here.
https://en.wikipedia.org/wiki/Nord_Stream
https://nl.wikipedia.org/wiki/Nord_Stream
https://de.wikipedia.org/wiki/Nord_Stream
The Dutch and German are a lot longer than the quite short English version. But ... that's just a matter of organisation: in the English the editors chose to make separate "Nord stream {1,2}" articles, in other languages they folded it in one article. On the German one in particular it's just two huge sections.
In short, it's fishy in the same way that bread tastes like fish: not at all.
One month ago, they were no english version available from the french page on the article, only a 3 lines 'simplified english' version were linked.
The thing that makes me question that is the data that is used to train those models to begin with. To disect truth on the internet, can you use the internet as a source of truth to train it?
the irony is that finding "objective truth" is a very non-trivial human game but in all cases costly. E.g journalism has been decimated after losing their traditional ad revenue. Wikipedia and science journals survive because they rely on informal and formal public funds etc.
Search engines will have to rely more on signals outside the content, such as links from other authoritative sources, but it does not look like a qualitatively different world.
I also agree that authoritative sources become critical. Yet those typically rely on very human assessments (with their own pitfals and controversies) and in any case much more slow / costly to develop.
How exactly this all will play out is not clear (to me). But the naive technosolutionism of deploying "AI at scale" and believing that it will just work as advertised seems misplaced. The human condition is very reflexive.