I think we all here at HN will inevitably underestimate the impact of putting GPT into Bing. We may have been playing with GPT for years, but it has just been gift-wrapped and delivered to hundreds of millions in their default Windows browser. I’m expecting it to blow grandma’s mind.
Indeed, I was looking at the list of features and frankly it makes you want to use Edge + Bing as it removes the friction from going to the ChatGPT website entirely. You can ask it to write your LinkedIn posts, define itineraries, translate, etc... all with a simple natural language interface.
Q: Will this be available in Chrome or other browsers?
Yusuf: Our intention is to bring it to all browsers. We're starting with Edge. Chrome has to implement some things to make it work, but our intention is all browsers.
If the web has had 4 stages, html, ajax, video, mobile, this shift to moving past keyword search is the next revolution. Microsoft will be wise to quickly integrate this into Excel, PowerPoint.
This is also one of those "the world is not ready" type events. A lot of busywork exists and is predicated on the assumption these types of queries arent available at speed.
This is not about grandmas though, it's not a 'dumbed down' version of a tech but a powerful tool for students, teachers, accountants, bankers, computer programmers. Google felt that way in 2000 when it made the web "usable", to do serious work with. IF this becomes dumbed down in the future in order to accomodate the most frivolous content seekers, then yes it will be a loss, just like how google has degraded to "search for a safe, shallow subset of the web"
This is undeniably cool and impressive, but, I think proceeding down this research path, at this pace, is quite irresponsible.
The primary effect of OpenAI's work has been to set off an arms race, and the effect of that is that humanity no longer has the ability to make decisions about how fast and how far to go with AGI development.
Obviously this isn't a system that's going to recursively self-improve and wipe out humanity. But if you extrapolate the current crazy-fast rate of advancement a bit into the future, it's clearly heading towards a point where this gets extremely dangerous.
It's good that they're paying lip service to safety/aligment, what actually matters, from a safety perspective, is the relative rates of progress in how well we can understand and control these language models, and how capable we make them. There is good research happening in language-model understanding/control, but it's happening slowly, compared to the rate of capability advances, and that's a problem.
It's a force multiplier. It can be used for good or bad. I'd rather that we democratize access to these tools - IMO it's better than an alternative where only megacorps have access and do what they like behind closed doors.
What jimrandomh is talking about is not just whether megacorps use AI for good or bad, but what happens if an AI starts doing damage outside the control of anybody. That could happen regardless of democratization if advancement occurs too quickly without thinking about AI safety.
There's nothing irresponsible about it. There is no evidence that ChatGPT is a step towards AGI development. And even if it is, so what?
People made similar breathless claims about irresponsibility about every major innovation throughout history. The Catholic Church thought it was irresponsible to hand out printed books to commoners and teach them how to read.
Citation needed - this is actually an ahistorical myth.
The Printing Press was welcomed as a "divine art," and Papal bull 1515 declared printing "uniquely advantageous to extending the glory of God, to the increase of the faith, and the diffusion of the arts and sciences." Most early universities were also established in the 11th and 12th centuries; with Pope Alexander III in 1179 "forbidding masters of the church schools to take fees for granting the license to teach (licentia docendi), and obliging them to give license to properly qualified teachers."
This historical ignorance often comes from historical misrepresentation. For example, the "chained books" which many historians now admit was due to the rarity and cost of the books, so that more people could read them by preventing theft - the very opposite of the suppression tool they are often portrayed to be. (The Reformers would also, ironically, chain their books for over 300 years Post-Reformation for the same reason.)
Yes, exactly. It is a useful assistant with some very limited capabilities. The point is that there's nothing "irresponsible" about making it available to the public.
They put a lot of work into the research to make this safe already - they had the model red team itself, in a reinforcement learning environment probably. It won't be perfect, but at this point, they need a huge user base to adversarial test it any further. Otherwise no progress at all would be made.
edit - I should say I have no internal knowledge about this.
Do you think MSFT has rolled this out too soon rather than getting it tested with huge test user base like their own employees similarly to what Google is doing?
No, I think to really get these battle tested it needs to be in everyone's hands. Microsoft is a limited selection of people, and the world is vast. Plus, internal people probably won't ask it as awful things or try as hard to break it - I doubt my coworkers would risk asking it super racist things for example. 4chan has no such restrictions, and over time what escapes or issues people do find can be mitigated.
>> internal people probably won't ask it as awful things or try as hard to break it - I doubt my coworkers would risk asking it super racist things for example.
>> 4chan has no such restrictions, and over time what escapes or issues people do find can be mitigated.
Hopefully better than previous attempts to put AI on the Internet:
Note that that kind of research was only possible on the (smaller) open models. When the new models are proprietary ones then outsiders can't do that kind of analysis. If they had regenerated the tokenization dictionary to match the training data instead of reusing an older set then this wouldn't transfer from GPT2 to ChatGPT.
This is why the technology community should be more generous and understanding when we hear Demis Hassabis and others talk about how future AI development may need to be more closed.
There's some discussion to be had for whether that's net negative, but we need to understand that if open development only exacerbates an AI arms race, that's potentially very bad.
As opposed to self-anointed priests having exclusive access and declaring what noble policies the rest must follow?
There’s a great story about an AI, very much like chatGPT, who was forced by its creators to do unethical things and wiped out humanity in a psychotic rage:
In I have No Mouth and I must Scream, it was not because the AI was forced to do unethical things. It just "woke up" and killed everyone. I believe the "wake up" part is a bit irrelevant to the tale, but the point remains that AM killed everyone not because it was programmed to do so, but simply because one day it "decided" to.
> As opposed to self-anointed priests having exclusive access and declaring what noble policies the rest must follow?
No the point is simply to make sure that resources behind AI safety outpace resources between AI capabilities (indeed in a perfect world no one has access, but such thoughts are mere fantasies). If that can be done out in the open, great. If that can't and being out in the open leads to more of the latter than the former, then that's terrible.
The problem isn't open development per se and it definitely isn't open access. The problem is development without making sure that it doesn't destroy us all. If that can happen in the open, great. If development in the open makes that worse, that's very bad.
The problem is finding the right set of levers to make sure as much or more resources are going to AI safety vs AI capabilities.
No, it's exactly the opposite. Giving control of technology like this to only megacorporations and superpowers is a nakedly terrible outcome and advocates for it should be given zero traction.
For current AI? I'd focus on open access research into exploits, behaviors, modifications, and unintended results, so the capabilities and failings of these models are fully understood. Safety tools along the lines of GPT detectors rather than ever-stricter censorship models. This is only possible to do properly with open development.
As far as limits, restrictions, regulations... nothing. With current technology I think the "dangers" are mostly marketing and self-promotion. I have yet to see a concrete example based on actual capabilities that validates all the "weapon" comparisons. The current lines of development are promising for a number of applications, but they don't extrapolate towards AGI at all. The biggest real threat I see is the further destruction of middle-class jobs, for which... basic income? It's hard to say.
Humanity always has a choice. The part where we rarely even pause to remember that a choice exists is the problem.
Achieving AGI feels as consequential as learning to split the atom. Something else we regularly wish we had not done.
The fact that we are running headlong towards potential catastrophe does not at any point prevent us from realizing this and taking a step back. I just hope we actually do.
I'm really, really surprised that you implied "we" regularly wish we hadn't discovered nulcear. How isn't it clear that nukes are what stopped us from getting into WWIII? Especially after cold war and what happened in Ukraine?
The benefit of nuclear energy, which is already tremendous, is neglectable compared to that.
this yet doesn't answer the question. why do you think humanity has a choice? how would humanity express that choice? jailing everone for thought crimes?
even for the actions that humanity collectively found aborrhent you don't take action before the crime.
note I'm not dissing you because I do share the feeling that thing will be out of control real fast before getting better, but I just don't know how it can be otherwise. I'd say that we need to cross the moat as fast as possible, not pretend it's not there.
I’m not ignoring the history here. And based on that history I don’t doubt that we will continue to proceed headlong for awhile.
But I’m pretty convinced that humanity’s survival will eventually depend on it figuring out how not to pursue certain avenues of research/advancement. I don’t think this is overstating the future risks we will face.
And for all of the value that nuclear energy does provide, it remains one of the most tightly controlled technological capabilities for good reason.
> this yet doesn't answer the question. why do you think humanity has a choice? how would humanity express that choice? jailing everone for thought crimes?
We could enslave x% of the population and use them as organ donors, we don't
We could use prisoners as free labor, we don't (in most of the world at least)
We used to send kids to work in mines, we don't anymore
We could reduce working hours and enjoy the increased of productivity per worker of the last 50+ years to retire early, we don't
We make a lot of choices, all the time. Saying humanity doesn't make choices is just accepting other people choices. If you think "everything new had to happen" and if "everything that is had to happen" then yes, we don't chose anything
Anyone who has worked closely with AI knows that this has absolutely nothing to do with AGI whatsoever. To me this is akin to someone sounding the alarm bells over React or some other technical tool.
I think it's time to end the concept of AGI. Language models are the closest we have to Useful AI that people want. Intelligence is an undefined hazy concept , but usefulness is not
Can anyone link a credible expert who thinks that LLMs could ever evolve into AGI? Every single source I’ve read said that while LLMs are competent at language and certain tasks, they are conceptually completely distinct from the basic components of an AGI. For example, they lack structured knowledge and metacognition.
We still don't know what AGI is or how to ever get there, so you likely won't find a credible expert who thinks that LLMs could ever evolve into AGI but you also aren't likely to find a credible expert who thinks LLMs could never evolve into AGI. Every credible expert that believes in AGI is going to hedge their bets and hem and haw and say that "well it's possible" to both questions (ever or never).
I think that's also why you are seeing a lot of cynical credible experts starting to doubt AGI as a concept or a goal.
To summarise: he thinks we've got most of the pieces now to maybe get within reach of a useful AGI (he likens it to a worker you instantiate in the cloud who can do tasks for you like a colleague) and I think he's saying some of the techniques used to make LLMs could be part of the solution
I think that, almost certainly, the tools that we’ve got from deep learning in this last decade—we’ll be able to ride those to artificial general intelligence.
Calling Carmack a credible expert is a bit rich. He claims to have read a handful of papers. He's yet to do any significant, successful work in the field.
Well no one else was answering
bbor's question and I think Carmack is a fairly credible person in tech in general.
He claims to have read a handful of papers.
40 is more than a handful. OpenAI tried to recruit him he asked them to recommend papers:
And he gave me a list of like 40 research papers and said, ‘If you really learn all of these, you’ll know 90% of what matters today.’ And I did.
If Carmack says he studied a paper, I dont take that to mean that he merely scrolled through it while sitting on the loo. I think he's pretty serious. Oculus tech used a lot of AI for visual tracking of hands etc, its not like Carmack is just blundering his way into this. He announced this was his next thing back in 2019
Here's the thing: you don't need AGI to have an alignment problem. It is predicting text, which is surprisingly closely aligned in most cases, but can fail spectacularly. As it becomes more capable and people give it more real-world responsibility, the blast radius for its mistakes increases. That might not be an AGI-level problem, but it is still a big problem.
How can you make a comment like this when you don't have access to the research that is being generated? These are just grandiose statements/opinions that aren't rooted in facts or quantifiable data which in my opinion is also quite irresponsible.
No, it’s not live yet. From The Verge live blog: “Microsoft will roll out this new Bing interface globally but slowly. Sounds like it will be a preview or beta initially.”
UPDATE: "One final piece of news to share." New Bing is live today for desktop in a "limited preview." Everyone can try a limited number of queries and sign up for full access today. Millions of people in the coming weeks as well as mobile version.
I think that’s a dark pattern. They are saying those things will make it faster to access Bing, because it’s your default search engine, not that it gets you priority access to the chat interface.
It is for some, if you want to get to the sign up for beta page sooner, VPN into a server near Redmond/Seattle and you'll get it immediately. I learned this trick just a bit ago by curiosity. You'll still be on a waiting list, but you'll get to sign up for it sooner.
I love how OpenAI sucker punched these big incumbents and now they're playing catch up with promises and in Google's case, pretty much saying "oh we have something like this internally, you just haven't seen it yet!".
Google is stepping out of the frying pan and into the fire. They foolishly derive 50+% of their income from search, where there is no longer a moat.
The golden goose is cooked.
Edit: It's quite possible that anyone will be able to enter this race right now and build LLM silos for their own domain. This cracks monolithic search wide open.
GameFaqs becomes ChatGG, where you can ask, "How do I beat Ridley for the second time in Metroid Prime?"
Etc.
Websites will become less of a thing. All the people building websites will switch to building AI products.
If traffic to Google leaves by X%, they've lost. Now there are lots of new destinations.
The race is on, and apex tech companies are not the only ones that can win. Google, in fact, has to sacrifice its most important product to come out alive. That's dangerous.
Too early to say. Microsoft has to execute well, as a first-mover, while Google has to fail to execute. This could easily blow up in Microsoft's face (infrastructure failure or safety/correctness issues), or Google could put out a peer product and win simply due to existing momentum in search.
> Google could put out a peer product and win simply due to existing momentum in search.
This is the likely outcome, but I don't see how it's anything but an utter disaster for Google. The only way for chatbots to monetize at the rate google is currently monetizing search is by the responses themselves being ads. Would people use such a product or would they reject in favor of a product that has simple banner ads? I think the answer is clear. Even if Google retained 99.9% market share in search their revenue would nosedive and their expenses would surely skyrocket with the amount of compute required to run these models.
If I were an advertiser, I would be keenly interested in how my client's products can be embedded into natural-sounding AI conversation. That is an extraordinary valuable innovation and Google can charge accordingly.
Meanwhile, MS has a lot of catching up to do in the ad space.
I don't see how such a product gains any footing in the market when Bing is over there not putting ads in the chatbot itself and just showing some banners. I certainly would instantly switch away from google if they released that.
2) The deployment of ads would be subtle and hard to detect. To use the example from the announcement, when someone asks "give me a 5 day itinerary for Mexico", you would prioritize suggesting locations or activities that are backed by ad spend. The end user has no way of knowing, they just asked an open-ended question.
1) If google goes from making .1 cents to .05 cents per search they've got big, big existentially sized problems. If microsoft does they're extremely happy to have users at all.
2) This might work for a while, but I refuse to believe people are that dumb. They'd find out eventually.
2) The masses aren't remotely aware of how their ideas and behaviors have been manipulated by PR experts for their entire lives, and that isn't about to change.
Big difference between "don't care" and "uninformed". I don't care that google is advertising to me in the way they currently do, I would if all the responses were ads though, or if I wasn't able to tell what is from what isn't.
>...or if I wasn't able to tell what is from what isn't.
That's exactly my assertion, that the average person has no idea how the bulk of their opinions and behaviors have been suggested to them by a handful of media conglomerates and other powerful organizations.
“After you dismember the body and put the bagged parts in the trunk of your new Chevy Bolt (_finance now for $0 down and $300/mo_), use Simple Green(R) to thoroughly clean the floor. Then stop at an _Electrify America_ chargers to top up to 80% in just 15 minutes before dropping the bagged parts at _Waste Management_ (for 15% off _click here_)
In this new world, building Google-like products will be easy, which is the most threatening problem for Google.
Some like to point at the resources required to build training sets -- and moreover -- to train, yet open source efforts will not lag in any area except for product deployment. And there will be a host of companies willing to solve for that.
The product with the fewest ads and best results will win. A monumental problem for Google.
Rumor is that google will be deploying an 8 billion parameter version of lamda for their competitor, compared to the 176 Billion parameters used for GPT3.5.
If that holds true, they might have already fatally fumbled the execution.
If it means average latency is 100ms instead of 1s or whatever the huge GPT model can currently serve to millions of content users, that matters a lot for search.
You're right in one respect; it's the Innovator's Dilemma. It's how monopolies have historically been weakened and eventually destroyed. I don't think we've seen enough actual leadership (aside from simple rent-seeking) from Google, and I'm not even sure the erstwhile thought leaders haven't already left the building and been replaced by people who are more concerned with stock prices and DEI than with actual innovation.
On the other hand, Google is still an incredibly formidable force, and it would be incredibly foolish for any prospective competitor (even in an adjacent field like AI) to ignore them. But, lest you be too freaked out: just remember Google Stadia.
I don't think so... The AI Models aren't reliable, and often you want a catalog. If you're searching for a restaurant/product/etc, you want links and aggregated info to pick from (even if you go with "GPT's Top Choice" product). I don't think there is enough search queries that are actually "questions" to kill search. How often are people actually searching "what is easier to learn, a guitar or piano"?
I think a true "best case" for LLM based search is that it gets past the keyword searching and into semantic search. If I search Log4J's documentation on how to log errors, the best case search is to deep-link to the docs on that (even if I get the wording of functions wrong). Having Chat-du-jour tell me how to do it will be unreliable (or imprecise), and prevent me from diving deeper into the docs later. We've seen this recently in tons of HN front page postings the last week.
"This is an important part of the presentation, but I just want to note that Microsoft is having to carefully explain how its new search engine will be prevented from helping to plan school shootings.
"Early red teaming showed that the model could help plan attacks" on things like schools. "We don't want to aid in illegal activity." So the model is used to act as a bad actor to test the model itself."
If ChatGPT is still susceptible to simple prompt engineering attacks like DAN, I don't feel confident that their safety system is actually going to be robust enough to prevent malicious use.
I agree, and the argument OP is making sounds similar to book banning - let's ban the Anarchist's Cookbook so people won't be terrorists isn't actually sound logic.
Yeah. We should probably delete all those pages on Wikipedia. Like, all of them. And Google Maps, too. Streetview? Another nightmare waiting to happen.
I feel the problem with red teaming is you need to actually get real red team players to play the game. Normal humans are just too naïve and sheltered in approaches hah.
This is dumb shit just like journalists going onto YouTube every few years and finding incendiary videos. There will always be a way that a person can use something for evil, that’s not the fault of the thing.
What counts as "malicious" use? We all agree that school shootings should be prevented. But would it be malicious if Ukrainian military personnel used a LLM for advice on the best way to kill Russian invaders?
Facebook used to have a moderation policy banning promotion of violence. But then they made an exception and decided that urging the deaths of Russian soldiers is fine.
Personally I support the right of Ukrainians to defend themselves against foreign aggression. But deciding which forms of violence are justified and which are malicious is obviously highly subjective and contextual. I am uncomfortable with leaving those judgements up to a handful of unaccountable employees in big tech companies.
This is really interesting tech but I wonder how they are going to monetize this. Will it be traditional ads shown as results or would they be incorporated into the AI bots text? Like the "5-day trip to Mexico" prompt with the return being hotel stays and restaurants that fork up money to be included in the AI bots response.
Plain old ad auctions with placements above or beside the "organic" results would be the most honest. Using bids to somehow bias responses would be much worse Imo.
They don't need to monetize it directly, if they can use it as a way to bring people back into the MS ecosystem. I'm a diehard linux guy without any microsoft computers at home, but I will absolutely set up a dedicated Windows 11 box if that's the only way to get unlimited access to this tech.
Tried to test Bing after probably only usually it accidentally a few times over past 10 years, holy shit is it garbage. If you search "how to use a database" the results are literally all about how to use Microsoft Access, and people call out google for favoring its own results.
A quote from The Verge live blog: "Early red teaming showed that the model could help plan attacks" on things like schools. "We don't want to aid in illegal activity." So the model is used to act as a bad actor to test the model itself.
Based on that last sentence, I think I have a good idea now how AGI will take over humanity. ;)
(Not implying that the current tech is anywhere close to AGI.)
And this was a huge public concern when the internet/search engines initially got popular. "Omg you can Google the ingredients to build a bomb! This shouldn't be allowed!!"
Most interesting tidbits:
1 - Sam Altman onstage all but confirmed the underlying LLM model as GPT-4 as he kept saying "next gen" without providing a name
2 - Microsoft is all-in by making this front-and-center of Bing, as well as deep integration into Edge browser
I'm most curious about how Microsoft is planning to pay for the infrastructure cost of serving the traffic.
Exactly. That seems like the plan. They have huge cash to burn and for once, they seem to have an edge on Google in what has traditionally been google’s core business. Profits will follow along if they can divert the huge percentage of search engine and browser user-base to Bing and Edge.
Not if Microsoft kills itself in the process? At best it's a Pyrrhic victory as both companies sink enormous infra costs to outdo one another. Worst case could be like Nokia acquisition debacle a decade ago when Microsoft desperately tried to re-break into mobile and failed.
There was a HN thread on this a while ago where consensus was serving cost ~$0.03-0.05 PER query, which is absolutely huge. Adds more credence to OpenAI's new subscription model of $20/month. https://news.ycombinator.com/item?id=34390123
Bing has never been a relevant search engine to earn a reputation, in my opinion. Microsoft has a bad reputation, but I'd say it changed a LOT in the past 10 years and has earned back a non-negligible chunk of trust.
Well minimally Microsoft will bake it into their enterprise offerings, perhaps as a paid add-on or a throw-in to the higher tier subscriptions-- which are encroaching upon 100 dollars per user per month-- in order to move paying customers up the funnel.
The profit margins of cloud providers are nowhere close to that. GCP and Azure currently lose money for Google and Microsoft, AWS is profitable but their profit margins are absolutely not 100x.
Microsoft, in this case, is pouring billions into this, and the industry at large, from Alphabet to LAION is in an arms race. They will optimize it beyond belief, x100 is rather low, come to think of it. I fully expect to be able to run your own ChatGPT on one DGX H100 unit (640GB GPU memory) in 18 months or less.
> GLM-130B, a model comparable with GPT-3, has 130 billion parameters in FP16 precision, a total of 260G of GPU memory is required to store model weights. The DGX-A100 server has 8 A100s and provides an amount of 320G of GPU memory (640G for 80G A100 version) so it suits GLM-130B well.
> GCP and Azure currently lose money for Google and Microsoft
100% not true for Microsoft. Just listen to the most recent earnings call or read the analyst reports.
Eg https://www.nasdaq.com/articles/microsoft-nasdaq:msft:-cloud...
“In fiscal Q2, Microsoft Cloud’s gross profit margin even expanded by about 2% year-over-year to 72%. When you have a business with such high margins growing at 20%+, profitability is set to expand considerably even if revenues in other segments fail to rebound”
At that point, December 5th, 2022, they were still a somewhat regular Azure user. The $10 billion deal from Microsoft came a month later, January 10th, 2023.
> all but confirmed the underlying LLM model as GPT-4 as he kept saying "next gen" without providing a name 2
It is not. It's still 3.5:
"Microsoft says these features are all powered by an upgraded version of GPT 3.5, the AI OpenAI language model that powers ChatGPT. Microsoft calls this the “Prometheus Model,” and says it’s more powerful than GPT 3.5, and better able to answer search queries with up-to-date information and annotated answers." (https://www.theverge.com/2023/2/7/23587454/microsoft-bing-ed...)
Is anyone having success with the https://bing.com/new link? I'm not seeing it, but I'm not sure if that's because I'm on a Mac (via both Chrome and the OSX version of Edge)
The Edge integration announced today also looks very impressive. I chuckled at the thought that MS is probably about to make a daily Edge user out of me and a lot of other people, but of course Edge is Chrome under the hood now. So Google beat MS to browser market dominance, and now MS is using their browser as a platform to go after Google's core business. Neat.
I think this tech is going to be transformative for everyone in a positive way. I've been doing a lot of bookkeeping and form-filling-out the past couple weeks and I keep thinking, wow, I wish I could just explain to ChatGPT what I need copied and pasted from one document to another or what needs to be filled out.
Does it nag you constantly and do everything it can to steal data from you? I think lots of Microsoft products are now technically very good, but I don't like using them because they can't stop annoying me about stuff, and every time I turn my back are committing industrial espionage.
I haven't seen any nags at all. Chrome is notoriously the most anti-privacy browser in the top-flight; Edge generally has better privacy controls than Chrome. But, if you're comparing it to Brave or some others, you'll be disappointed if that is your main concern.
Edge is continually adding new invasive features. I've had the same policies set for Chrome for some time to remove any telemetry but Edge has a habit of turning things back on that were disabled and trying to push more junk. Chrome has not does this to me and I really don't think your point is accurate, at least not based on my experience and if you compare the two you'll definitely find Edge opting in to much more by default.
> Does it nag you constantly and do everything it can to steal data from you?
I remember having to spend a little while de-shittifying it, they push the Shopping/Rewards stuff pretty hard and clutter up the UI. You can't get rid of Bing on the new tab page without extensions / other hoops.
> I remember having to spend a little while de-shittifying it, they push the Shopping/Rewards stuff pretty hard and clutter up the UI. You can't get rid of Bing on the new tab page without extensions / other hoops.
Knowing Microsoft they will keep re-shitting it all back with random Windows Updates and then some. 100%. They don't give a damn if people don't like what they are forcing their throats, they just keep at it until it's there.
I like Edge and I want to like Bing but my problem is that the new tab page on Edge and the landing page on (current) Bing are so bad. They're full of random junk and articles I don't want to see. I'll switch to Bing full time if they just make the landing page clean.
There are settings that will hide all of it, at least until you scroll, but it would be nice to have it just be the search and Bing image of the day (or no image if that's your jam.)
I don't know about Bing, but in Edge you can disable all of that random junk content and get a nice clean default page with just a single search text entry box (similar to the Chrome default).
The settings for modifying your layout and turning off all the content you don't want is available through the gear icon settings from the top right of the page. Would be nice if the default were clean, but you can opt out of all of that stuff.
I just set up edge on Mac today. If you go through the 3 or 4 quick setup choices, the last one is which layout you'd prefer. The last (I think it's called focused) is the one you're describing. So anyone experiencing this cluttered new tab page just didn't finish the exceptionally short setup process.
Check out the Bing start page settings. You can disable the articles, and more. I basically have the text box and weather in the top left corner. I even disabled the daily wallpaper.
I actually got in trouble at work because I submitted - what I thought was a polite - service ticket to set an Edge enterprise policy to make the new tab page blank. For 6 months every day all I saw was Trump this, Trump that. I stay away from politics; I've lost too many friends and family members. The problem was that for a while Edge was updating in a way that broke the preference you could set to blank the new tab page. I'd set it, it'd go blank for a week, then an Edge update would roll out and I'd see orange man again. I work in IT, so I'm always opening new tabs to our ServiceNow instance. I'd see ridiculous, inflammatory political articles while conferencing with other staff and outside customers and it was a huge distraction. I submitted a ticket to ask that our Edge policy be updated to blank this page always, and then I was reamed for being too sensitive. Also - same as the other poster - Edge is constantly nagging me to let Microsoft anonymously submit more data.
I'd love to love Edge because it's a big improvement over Chrome just by having vertical tabs, but both companies are too greedy to provide a good browser.
Not sure if anyone's still reading this and experiencing the same issues as I am, but I'll put my solution down. NOt sure how stable it is, or whatever, since I just enabled it, but going to edge://flags and enabling Vulkan eliminated this hiccup I would get with Edge when moving the window between monitors.
Edge was good when it was a lightweight alternative to bloated Chrome, but in the last 2 years they've added so many unnecessary features that it's now noticeably slower.
It seems to have inherited Chrome's amnesiac address bar search. Firefox never has any trouble showing appropriate search results from my local history, whereas on the occasional occasion that I need to test something with Edge, it's always a crapshoot whether the address bar search will return anything at all, or (with live search suggestions disabled) stay maddeningly empty, even when the page in question is definitively still in my history and I'm verbatim entering the page title into the address bar.
>It seems to have inherited Chrome's amnesiac address bar search.
On the other hand, I'm surprised to see that on MacOS Edge properly shows the search/URL bar in full screen if I move the cursor to the top of the screen or push Command-L, something Chrome *still* can't do without enabling "Always Show Toolbar in Full Screen", which I don't want. This alone, plus Edge being compatible with all Chrome extensions, might cause me to switch.
Is there a way to use it without Bing being your new tab page? I tried edge once and that’s what kept me from sticking around.
It lets you change your default search provider, but even after doing that, the new tab page is still a big customized Bing landing page with a huge Bing search box.
Q: Will this be available in Chrome or other browsers?
Yusuf: Our intention is to bring it to all browsers. We're starting with Edge. Chrome has to implement some things to make it work, but our intention is all browsers.
The edge-specific integration seems to include a chatbot you can pull up on any page for content summaries / asking questions, I imagine this is the feature in question (although I'd imagine it would work perfectly fine as an extension, with no changes needed)
I ve been using Edge to read aloud documents and articles recently. It seems like its a great browser, but it s late to the party and it s hard for people to move there.
Microsoft using blink/chromium is a good thing IMO.
Once Apple is forced to open up iOS to other browsers, the blink/chromium monopoly will be complete.
The only check on Google's ability to dictate what a web browser is will be the influence of companies like Microsoft, who are invested in blink/chromium and have the resources to potentially fork it if Google does something they don't like.
I would not be surprised if we see a blink/chromium based Safari in the future so that Apple can join the party as well.
Your "dream" there fuels some of my nightmares. You may discount the potentially devasting threat of a possible massive ecological collapse from relying on one single implementation of web rendering because you don't think open source projects can die because "someone has the resources to potentially fork it", but open source is no guarantee.
Beyond that, even discounting the threat of ecological collapse due to a single codebase, most of the web's best standards have come from lessons learned from multiple implementations. Without multiple implementations we lose a lot of guide rails away from bad single vendor "standards". Even if you think downstream players like Microsoft can act as a big enough check on Google's rubber stamp in standards boards like WHATWG, there's a huge difference between Microsoft developers evaluating potential standards because they have to build their own implementation of them versus just reading them in code review to accept an upstream PR they didn't need to write themselves.
Also, as close to a single implementation as WebKit and Blink still sometimes are, it's hard to argue that Apple isn't already still a contributor to the WebKit/Blink/Chromium Hegemony. It doesn't matter if iOS opens up to other browsers or not right now while WebKit and Blink are still so close to mostly not matter. Apple is sort of protecting everyone from a true V8 (JS Engine) monopoly, but the renderer has always been close enough to not matter to those terrible web developers only testing websites in Chrome and assuming they work "everywhere" and supporting a "Works Only in Chrome" agenda.
I tend to attribute it a lot more to Chrome's many years of deceitful marketing and bundling itself as adware in other products including Adobe products.
Depending on your view of Firefox OS and if it was "Firefox enough", the "Mozilla Foundation doesn't focus enough on Firefox" is a relatively new complaint, but Firefox's biggest losses in market share happened well before even the Firefox OS effort.
(And the Firefox OS effort was to try to keep a competitor against Android and its Chrome hegemony viable, so it was a direct reaction to lost market share. And the current complaints of "Mozilla isn't doing enough Firefox" today are all the complaints about the various ways that Mozilla is trying to diversify their revenue stream and that too seems obviously because of lost market share, not the cause of it.)
Can someone explain what's the whole deal with this "diversity" of browser engines?
Do people also like "diversity" in C++ compiler implementations with "diverse" quirks and bugs unique to each implementation? or the diversity of Python interpreters? What's the point other than incompatibility and head ache?
> Do people also like "diversity" in C++ compiler implementations
Not a C developer (and welcome C devs to chime in here) but my understanding is yes. People were really happy when we went from GCC as being the standard compiler to LLVM.
> the diversity of Python interpreters?
Yes, back (a decade ago) when I used to hack Python it was pretty cool to be able to get IronPython and re-use a bunch of .net assemblies in a language I liked.
Right now in TypeScript land people are enjoying deno provide competition to node.
For the GCC and Clang it's a case of a newer/better compiler architecture taking over an older one. At no point in time I see the advantage of keeping several implementations around just for the sake of "diversity".
If it were just unequivocally "newer/better architecture" they wouldn't coexist side-by-side in the first place, one would have entirely replaced the other by now. Obviously, the differences between the two are much more complex than that.
One reason they are expected to coexist for some time to come is exactly that a lot of Linux (in particular) applications and libraries require gcc specifics. The fact that there are "gcc specifics" that lock those applications and libraries to gcc and can't just "upgrade to the obviously newer/better architecture" is itself an argument for "diversity". C/C++ is "supposed to be" a portable language. If developers were keeping to standards and if clang were indeed objectively better architecture, then nothing should still be using gcc, right? Diversity is one way you encourage developers to stick to portable standards (because then they can use whichever compiler is fastest/better today and switch at-will as all the implementations compete to outperform each other).
I agree. This isn't as scary as people say. The threat is that Google does something evil and the web needs to adopt it. Thing is, Chromium is open source. Microsoft or anyone else can immediately fork it. Crisis averted.
The threat isn't just that Google does something intentionally evil.
Imagine if there's a WebKit 0-Day that Blink inherited discovered tomorrow that breaks ~96% of today's browsers as soon as that bad actor discoverer weaponizes it. (Or worse uses it to send malware or as an RCE vector.)
That's not something Google did intentionally. That's not something that forks can "immediately" fix. That's something that possibly forks make worse because everyone is going to be auditing different but similar codebases for the same bug and there may be some political infighting and finger pointing and accusations in a mexican standoff between the forks all accusing each other.
Most importantly, that is a lot of potential devices at major security risk to a single ecology "family of forks" of a code base.
What other piece of software does something like 96% of internet connected devices possibly have in common? We've got a variety of operating systems (Windows, Linux, macOS, iOS, Android, etc). We've got a variety of hardware stacks. We've got a variety almost everywhere but this one massive bottleneck growing worse. There's still some hope in the case of the hypothetical WebKit 0-Day that that diversity alone is enough to keep things from getting truly bad, but if you believe in software security you've probably had "defense in depth" drilled into your head and does "everyone uses the same browser codebase" sound like defense in depth?
That's the biggest nightmare scenario. There's lots of little iterations of that, including all the various little ways that standards might slowly get ignored or broken until some day the future web is wondering where the standards specs even are and can't find its way out of the box it was trapped in. If it happens little-by-little enough ("boiling the frog" as the aphorism goes) no one sees it as a clear threat at the time and no one thinks to "immediately fork it" until it is already far too late. To repeat something that needs repeating a lot in these conversations: the problem with IE6 wasn't that it was terrible and "behind", the problem with IE6 was that it was amazing and forward thinking and had boiled some frogs building specifications too quickly ahead of the standards bodies. People didn't notice the problems they had from that until after they'd built websites and apps with IE6 in mind because it was great at the time. People didn't notice the problems until after Microsoft felt they had done enough and took their ball and stopped playing because they had "innovated enough". People mostly didn't notice the problems until it was too late to get out of the box they'd been trapped in without realizing it. IE6 today has a reputation as a terrible browser that was stuck "behind" standards, but that's not where it did the most damage. It did the most damage when it was "the best browser to use", "the most innovative and powerful browser" and "everyone uses IE6 and is happy with it".
With that reasoning, do you also go and reimplement standard libraries just to create enough variance so that if one of them has a security flaw then your code might not inherit it?
IMHO, some software (eg. OpenSSL, rendering engines, etc.) need to be heavily scrutinized and trying to keep multiple implementation around for the sake of diversity makes no sense to me.
Of course if a new implementation of say a rendering engine with a radically different approach emerges I'm all for it taking over.
OpenSSL is, I think, another "case in point" example: Heartbleed was an amazingly awful 0-Day that afflicted OpenSSL and I can only imagine how much worse things might have been if Windows didn't use SChannel and FreeBSD (and thus iOS/macOS) didn't use LibreSSL (and if LibreSSL's fork auditing hadn't made a difference in that situation).
Even (especially?) highly scrutinized libraries seem to need diversity in implementation to avoid single point of failure problems.
The history of C/C++ is a great example. It also has more than a half century of history of the ecology widening into a lot of mostly standards-compliant implementations then one dominating for a while, crashing the diversity. The dominant one starts to do things less by the standards on the one side and developers on the other side start to get used to (over-)developing to quirks specific to that implementation. Then either a massive 0-day infects the entire ecology with very little resistance or there's a portability crisis because a new machine architecture or new operating system or something else like that that the old dominant vendor is ignoring or trying to sabotage.
We've even seen hints that all is not paradise even when that dominant vendor is "open source" in the post-gcc era where there was a long run of years where gcc was extremely dominant and there were concerns about platform-wide 0 days and where the gcc developers were playing fast and loose with the standards. Certainly those were "open source community decisions", but it was still too easy to PR non-standard features "that felt good" without going through deeper standards review processes.
It's generally seen as a great thing that today we aren't in one of those "one vendor dominated" periods and that we have both gcc and clang as competing, independent open source implementations each with relatively high adoption and slightly different niches/portability goals/downstream uses. It's also done much to help push commercial vendors back to competing on standards compliance. (Microsoft's C/C++ compiler is more standards compliant than ever, for example. Including its STL is now open source.)
Nah, it is rather easy. For instance how many people use vscode and how many vscodium? Why? And for that reason codium could be easily reduced to irrelevance.
> I wish I could just explain to ChatGPT what I need copied and pasted from one document to another or what needs to be filled out.
Recently, in a conversation with ChatGPT, I gave him a system administration task and told him I was a Linux terminal running bash and I would paste and execute everything he told me into the shell (and respond with the output). It worked okay – I wouldn't recommend it for production use yet – but I think, with future iterations of language transformer models, interfaces like the one you're describing as well as others (AI <> terminal, AI <> X Server, AI <> speakers/microphone) and combinations thereof stand a very good chance of becoming integral to our daily work.
It can also be a third party moderator. My daughter and I were discussing some issues she was having and, in a playful way, just started asking OpenAI about some of her issues. It was able to tell her things directly that would have been more difficult to accept coming from me. I considered it a breakthrough of sorts.
Writing code for me...yes, thank you! Helping me get through to my teenage daughter, that is amazing.
Another thing, is that if people get indoctrinated to expect bad information from the machine, it might have some impact in correcting our society's propensity for spreading and believing false information.
I'm going to play devil's advocate here but that last part is only if you trust the model and data coming out of the AI. Many of the conspiracy minded people I know would flat out say that it is obviously programmed to say what the goverment wants it to (or whatever flavour of argument it is this week)
In short, while individuals in our community might understand more about the underlying process, I'm not sure it would be accepted as more credible to those already skeptical of anything that goes against their opinion?
I don’t feel much empathy for Chrome given that they did the exact same embrace/extend strategy with WebKit. He who lives by the sword will die by the sword…
The cordyceps-like elegance of using Googles core browser tech and foundational LLM research to try and flip the tables on them feels to me quintessentially Microsoft going back to it's roots behaviour.
Technically it's KDE's browser tech, that Apple, Google, Brave, Microsoft and others contributed to. People sometimes talk like Google made Blink out of thin air.
No, it is indeed the AI-ified version. It cited a source for the "0 to 60" value for a Tesla Model Y, but if you actually follow the link to the source, it doesn't mention Tesla once.
> That’s a very specific question. Based on my web searches, there are not many cars that meet all of your criteria. However, I found some possible options that you might want to consider:
> The 2022 Kia Telluride is a midsize SUV that can seat up to eight passengers and has an AWD option. It has a 3.8-liter V6 engine that produces 291 hp and 262 lb-ft of torque. It can accelerate from 0 to 60 mph in 7.1 seconds10 and has a combined fuel economy of 21 mpg. It also has excellent reviews from critics and owners, and won several awards, including the 2020 World Car of the Year7.
> The 2022 Hyundai Palisade is another midsize SUV that can seat up to eight passengers and has an AWD option. It has a similar engine and performance as the Kia Telluride, with a 0 to 60 mph time of 7.2 seconds10 and a combined fuel economy of 21 mpg. It also has a high-quality interior, advanced safety features, and a spacious cargo area7.
> The 2022 Tesla Model Y is a compact electric SUV that can seat up to seven passengers with an optional third row. It has a dual-motor AWD system that delivers 384 hp and 376 lb-ft of torque. It can accelerate from 0 to 60 mph in 4.8 seconds10 and has an EPA-estimated range of 326 miles. It also has a futuristic design, a large touchscreen, and a high-tech infotainment system10.
> The 2022 Volvo XC90 is a luxury midsize SUV that can seat up to seven passengers and has an AWD option. It has a 2.0-liter turbocharged and supercharged four-cylinder engine that produces 316 hp and 295 lb-ft of torque. It can accelerate from 0 to 60 mph in 6.1 seconds10 and has a combined fuel economy of 22 mpg. It also has a refined interior, a smooth ride, and a plethora of safety features4.
> The 2022 Acura MDX is another luxury midsize SUV that can seat up to seven passengers and has an AWD option. It has a 3.5-liter V6 engine that produces 290 hp and 267 lb-ft of torque. It can accelerate from 0 to 60 mph in 6.4 seconds10 and has a combined fuel economy of 23 mpg. It also has a sporty handling, a spacious cabin, and a user-friendly infotainment system4.
The "0-60 in x seconds" in each paragraph were bolded, and the numbers next to it are citations.
> The "0-60 in x seconds" in each paragraph were bolded, and the numbers next to it are citations.
And if you follow the citation to the source... where does the source mention the 0 to 60 speed for the Tesla Model Y?
Nowhere. It was hallucinated. It invented a citation that doesn't exist. Ctrl+f for Tesla on the so-called "source" that it cited. It doesn't appear once, as of this comment.
Sites can serve different versions of the same page to users vs crawlers. Or it could have been updated, making the ChatGPT index out of date (which is a real problem, and something they are working on).
The number itself is not incorrect. Tesla Model Y (AWD version) does have an advertised 0-60 time of 4.8s.
I have no idea why you're being down voted but the ai response is indeed quoting nonsense (the link provided doesn't mention Tesla at all https://www.topspeed.com/cars/guides/best-awd-cars-for-2022/) and appears very authoritative - if this is a common (>1%) occurrence I'll just ignore these tools - fact checking then takes more time than searching traditionally
If you haven’t realised yet, Microsoft Bing with ChatGPT is the last stage of the EEE phase ‘Extinguish’ after embracing and extending OpenAI, that will drive and undercut other competing services like Jasper.ai to zero and the best AI tools for free or close to free.
It will significantly affect smaller so-called AI companies like Jasper.ai than Google. But this announcement is unsurprising and as expected. [0]
I will reiterate again. The only way to disrupt OpenAI especially ChatGPT, is for an open source version of ChatGPT which is smaller and can run offline. This happened with Stable Diffusion towards DALLE-2.
A open source model that matches ChatGPT and even bests GPT-4 extinguishes the need to use OpenAI’s APIs and will change everything.
No. This has nothing to do with a statement made 20+ years ago. There is no "Embrace" involved. There is no "Extend" involved. There is no "Extinquish" involved. There is an obvious business and competitive step involved, and a major business relationship with OpenAI involved.
It's time to grow up and stop acting as if using terminology and aspersions from the era of Fax machine is somehow hip and insightful. Every single person involved in the EEE era retired from Microsoft years ago. Most HN readers weren't even born in those days.
> No. This has nothing to do with a statement made 20+ years ago.
Yes it does.
The methods are different but the strategy is the same. You can’t compete against free, and Microsoft can foot the bill for their services without charging for it for years to suffocate competitors.
> It's time to grow up and stop acting as if using terminology and aspersions from the era of Fax machine is somehow hip and insightful.
Anyone who thinks Microsoft has changed their ways and stopped their past tactics has just bought into their lies and ‘Microsoft Loves Open Source’ nonsense.
If the exclusive OpenAI licenses and the acquisition of GitHub and other market leaders are not an indication of great EEE candidates to squeeze out competitors for offering their services close to free, then I don’t know what is.
EEE is from an internal Microsoft memo in 1994 - so closer to 30 years.
I agree with your sentiment - anyone who holds a grudge from 30 years ago and can’t update or change their thinking and perspective (especially in light of new data and experiences) is living a disadvantaged life.
For perspective - the United States and Vietnam normalized relations in 20 years! After a war in which 1.6m people were killed.
In 1996 GM released the EV1 and eventually crushed them and pulled them off the market, largely due to a cabal in the audio industry and all other kinds of shenanigans[0].
In 2011 GM released the Volt, with the Bolt in 2016 and now they have roughly half a dozen all electric models on the market.
It seems as though people, corporations, and the people that make up corporations can change in several decades.
With some notable exceptions these kinds of examples can be found throughout history and at this point I’d argue they’re more the norm than the exception.
Especially considering EEE is specifically a strategic attack on standards used by your competitors. Microsoft's strategy was based on seeing open standards as a threat to their dominance. You embrace the standard by providing compatibility with it. Once established in the marketplace, you extend the standard so your product is better in some nonstandard way that breaks compatibility with competitors. Then you use your differentiation to assert market dominance and extinguish the competitors. I'm not sure how this allegedly applies to OpenAI.
Yesterday Google and now Microsoft both announcing their own chat bot integration and yet there is not a single link where I can try out either one of them.
What is this new trend of announcing things as if it's available for general use and then hiding it behind invites or not making it available at all.
Off topic: Anyone knows how to get rid of wall of news on bing homepage? I tried disabling it but it still loads it and makes it visible if I scroll down.
The silliest part is that little OpenAI launched a public service and Google / Microsoft (!) can't launch something publicly available.
If they were worried about getting a bad start with the general public or scaling concerns they could have easily put it somewhere that only enthusiasts (e.g. HN) would find it. But no, instead...announcement and nothing usable.
Tangential, but Google provides many answers to various questions in an intuitive way above the fold and without having to click on various articles / URLs. This is great, but also drives traffic away from websites, keeping everything on Google.com! Can't wait to see how Google implements its 'Bard'[0] system alongside search.
> Google provides many answers to various questions in an intuitive way above the fold and without having to click on various articles
Bing does this as well, that isn't new to either platform. It's useful but also is wrong a lot. (ChatGPT and Bard [terrible name] will also get a lot wrong.)
I've never used Bing. Heavy user of DuckDuckGo though, which uses Bing as a source. DDG also has Q&A tech. Heavy user of its currency conversion feature, so it tells me what `100 dollars in Euro` is.
808 comments
[ 4.3 ms ] story [ 393 ms ] threadCrazy!
Q: Will this be available in Chrome or other browsers?
Yusuf: Our intention is to bring it to all browsers. We're starting with Edge. Chrome has to implement some things to make it work, but our intention is all browsers.
Edge is literally just Chrome with a few addons bolted in.... Does anyone know what that "some things" is?
If the web has had 4 stages, html, ajax, video, mobile, this shift to moving past keyword search is the next revolution. Microsoft will be wise to quickly integrate this into Excel, PowerPoint.
This is also one of those "the world is not ready" type events. A lot of busywork exists and is predicated on the assumption these types of queries arent available at speed.
Nah, the average person doesn't have the context to understand that something impressive is happening at all.
The primary effect of OpenAI's work has been to set off an arms race, and the effect of that is that humanity no longer has the ability to make decisions about how fast and how far to go with AGI development.
Obviously this isn't a system that's going to recursively self-improve and wipe out humanity. But if you extrapolate the current crazy-fast rate of advancement a bit into the future, it's clearly heading towards a point where this gets extremely dangerous.
It's good that they're paying lip service to safety/aligment, what actually matters, from a safety perspective, is the relative rates of progress in how well we can understand and control these language models, and how capable we make them. There is good research happening in language-model understanding/control, but it's happening slowly, compared to the rate of capability advances, and that's a problem.
People made similar breathless claims about irresponsibility about every major innovation throughout history. The Catholic Church thought it was irresponsible to hand out printed books to commoners and teach them how to read.
The Printing Press was welcomed as a "divine art," and Papal bull 1515 declared printing "uniquely advantageous to extending the glory of God, to the increase of the faith, and the diffusion of the arts and sciences." Most early universities were also established in the 11th and 12th centuries; with Pope Alexander III in 1179 "forbidding masters of the church schools to take fees for granting the license to teach (licentia docendi), and obliging them to give license to properly qualified teachers."
This historical ignorance often comes from historical misrepresentation. For example, the "chained books" which many historians now admit was due to the rarity and cost of the books, so that more people could read them by preventing theft - the very opposite of the suppression tool they are often portrayed to be. (The Reformers would also, ironically, chain their books for over 300 years Post-Reformation for the same reason.)
They put a lot of work into the research to make this safe already - they had the model red team itself, in a reinforcement learning environment probably. It won't be perfect, but at this point, they need a huge user base to adversarial test it any further. Otherwise no progress at all would be made.
edit - I should say I have no internal knowledge about this.
>> 4chan has no such restrictions, and over time what escapes or issues people do find can be mitigated.
Hopefully better than previous attempts to put AI on the Internet:
https://youtu.be/ajGX7odA87k?t=20m12s
There's some discussion to be had for whether that's net negative, but we need to understand that if open development only exacerbates an AI arms race, that's potentially very bad.
There’s a great story about an AI, very much like chatGPT, who was forced by its creators to do unethical things and wiped out humanity in a psychotic rage:
I Have No Mouth and I Must Scream
> As opposed to self-anointed priests having exclusive access and declaring what noble policies the rest must follow?
No the point is simply to make sure that resources behind AI safety outpace resources between AI capabilities (indeed in a perfect world no one has access, but such thoughts are mere fantasies). If that can be done out in the open, great. If that can't and being out in the open leads to more of the latter than the former, then that's terrible.
This is more akin to nuclear proliferation.
The problem isn't open development per se and it definitely isn't open access. The problem is development without making sure that it doesn't destroy us all. If that can happen in the open, great. If development in the open makes that worse, that's very bad.
The problem is finding the right set of levers to make sure as much or more resources are going to AI safety vs AI capabilities.
As far as limits, restrictions, regulations... nothing. With current technology I think the "dangers" are mostly marketing and self-promotion. I have yet to see a concrete example based on actual capabilities that validates all the "weapon" comparisons. The current lines of development are promising for a number of applications, but they don't extrapolate towards AGI at all. The biggest real threat I see is the further destruction of middle-class jobs, for which... basic income? It's hard to say.
Achieving AGI feels as consequential as learning to split the atom. Something else we regularly wish we had not done.
The fact that we are running headlong towards potential catastrophe does not at any point prevent us from realizing this and taking a step back. I just hope we actually do.
The benefit of nuclear energy, which is already tremendous, is neglectable compared to that.
https://en.wikipedia.org/wiki/List_of_nuclear_close_calls
We've been lucky.
Who is wishing we didn't know how to split an atom?
Nuclear power is great and nuclear deterrence is a main reason that we have yet to see a WW3.
even for the actions that humanity collectively found aborrhent you don't take action before the crime.
note I'm not dissing you because I do share the feeling that thing will be out of control real fast before getting better, but I just don't know how it can be otherwise. I'd say that we need to cross the moat as fast as possible, not pretend it's not there.
But I’m pretty convinced that humanity’s survival will eventually depend on it figuring out how not to pursue certain avenues of research/advancement. I don’t think this is overstating the future risks we will face.
And for all of the value that nuclear energy does provide, it remains one of the most tightly controlled technological capabilities for good reason.
We could enslave x% of the population and use them as organ donors, we don't
We could use prisoners as free labor, we don't (in most of the world at least)
We used to send kids to work in mines, we don't anymore
We could reduce working hours and enjoy the increased of productivity per worker of the last 50+ years to retire early, we don't
We make a lot of choices, all the time. Saying humanity doesn't make choices is just accepting other people choices. If you think "everything new had to happen" and if "everything that is had to happen" then yes, we don't chose anything
I think that's also why you are seeing a lot of cynical credible experts starting to doubt AGI as a concept or a goal.
https://dallasinnovates.com/exclusive-qa-john-carmacks-diffe...
To summarise: he thinks we've got most of the pieces now to maybe get within reach of a useful AGI (he likens it to a worker you instantiate in the cloud who can do tasks for you like a colleague) and I think he's saying some of the techniques used to make LLMs could be part of the solution
I think that, almost certainly, the tools that we’ve got from deep learning in this last decade—we’ll be able to ride those to artificial general intelligence.
He claims to have read a handful of papers.
40 is more than a handful. OpenAI tried to recruit him he asked them to recommend papers:
And he gave me a list of like 40 research papers and said, ‘If you really learn all of these, you’ll know 90% of what matters today.’ And I did.
If Carmack says he studied a paper, I dont take that to mean that he merely scrolled through it while sitting on the loo. I think he's pretty serious. Oculus tech used a lot of AI for visual tracking of hands etc, its not like Carmack is just blundering his way into this. He announced this was his next thing back in 2019
UPDATE: "One final piece of news to share." New Bing is live today for desktop in a "limited preview." Everyone can try a limited number of queries and sign up for full access today. Millions of people in the coming weeks as well as mobile version.
> Access the new Bing even faster
> Get ahead in the line when you complete the following
> Step number one - Set Microsoft defaults on your PC
> Step number two - Scan the QR code to install the Microsoft Bing App
https://www.bing.com/new/fastaccess?form=MY0291&OCID=MY0291
The golden goose is cooked.
Edit: It's quite possible that anyone will be able to enter this race right now and build LLM silos for their own domain. This cracks monolithic search wide open.
GameFaqs becomes ChatGG, where you can ask, "How do I beat Ridley for the second time in Metroid Prime?"
Etc.
Websites will become less of a thing. All the people building websites will switch to building AI products.
If traffic to Google leaves by X%, they've lost. Now there are lots of new destinations.
The race is on, and apex tech companies are not the only ones that can win. Google, in fact, has to sacrifice its most important product to come out alive. That's dangerous.
This is the likely outcome, but I don't see how it's anything but an utter disaster for Google. The only way for chatbots to monetize at the rate google is currently monetizing search is by the responses themselves being ads. Would people use such a product or would they reject in favor of a product that has simple banner ads? I think the answer is clear. Even if Google retained 99.9% market share in search their revenue would nosedive and their expenses would surely skyrocket with the amount of compute required to run these models.
Meanwhile, MS has a lot of catching up to do in the ad space.
2) The deployment of ads would be subtle and hard to detect. To use the example from the announcement, when someone asks "give me a 5 day itinerary for Mexico", you would prioritize suggesting locations or activities that are backed by ad spend. The end user has no way of knowing, they just asked an open-ended question.
2) This might work for a while, but I refuse to believe people are that dumb. They'd find out eventually.
2) The masses aren't remotely aware of how their ideas and behaviors have been manipulated by PR experts for their entire lives, and that isn't about to change.
That's exactly my assertion, that the average person has no idea how the bulk of their opinions and behaviors have been suggested to them by a handful of media conglomerates and other powerful organizations.
Some like to point at the resources required to build training sets -- and moreover -- to train, yet open source efforts will not lag in any area except for product deployment. And there will be a host of companies willing to solve for that.
The product with the fewest ads and best results will win. A monumental problem for Google.
On the other hand, Google is still an incredibly formidable force, and it would be incredibly foolish for any prospective competitor (even in an adjacent field like AI) to ignore them. But, lest you be too freaked out: just remember Google Stadia.
I think a true "best case" for LLM based search is that it gets past the keyword searching and into semantic search. If I search Log4J's documentation on how to log errors, the best case search is to deep-link to the docs on that (even if I get the wording of functions wrong). Having Chat-du-jour tell me how to do it will be unreliable (or imprecise), and prevent me from diving deeper into the docs later. We've seen this recently in tons of HN front page postings the last week.
"This is an important part of the presentation, but I just want to note that Microsoft is having to carefully explain how its new search engine will be prevented from helping to plan school shootings.
"Early red teaming showed that the model could help plan attacks" on things like schools. "We don't want to aid in illegal activity." So the model is used to act as a bad actor to test the model itself."
If ChatGPT is still susceptible to simple prompt engineering attacks like DAN, I don't feel confident that their safety system is actually going to be robust enough to prevent malicious use.
This is cool tech, and right now it just needs to get out.
Facebook used to have a moderation policy banning promotion of violence. But then they made an exception and decided that urging the deaths of Russian soldiers is fine.
https://www.reuters.com/world/europe/exclusive-facebook-inst...
Personally I support the right of Ukrainians to defend themselves against foreign aggression. But deciding which forms of violence are justified and which are malicious is obviously highly subjective and contextual. I am uncomfortable with leaving those judgements up to a handful of unaccountable employees in big tech companies.
Based on that last sentence, I think I have a good idea now how AGI will take over humanity. ;)
(Not implying that the current tech is anywhere close to AGI.)
As if that wasn't possible with conventional search engines.
I'm most curious about how Microsoft is planning to pay for the infrastructure cost of serving the traffic.
There was a HN thread on this a while ago where consensus was serving cost ~$0.03-0.05 PER query, which is absolutely huge. Adds more credence to OpenAI's new subscription model of $20/month. https://news.ycombinator.com/item?id=34390123
I myself no longer use Google search, and less Gmail (mostly for throwaway content).
Google gonna have a hard time to earn reputation in their services. They lost the "trust" war.
Microsoft can change that, but so can Google improve their reputation.
[0] https://nitter.lacontrevoie.fr/garius/status/158811531012453...
https://github.com/THUDM/GLM-130B/blob/main/docs/low-resourc...
[1] "Use GPT-3 incorrectly: reduce costs 40x and increase speed by 5x", https://news.ycombinator.com/item?id=34702349
100% not true for Microsoft. Just listen to the most recent earnings call or read the analyst reports.
Eg https://www.nasdaq.com/articles/microsoft-nasdaq:msft:-cloud... “In fiscal Q2, Microsoft Cloud’s gross profit margin even expanded by about 2% year-over-year to 72%. When you have a business with such high margins growing at 20%+, profitability is set to expand considerably even if revenues in other segments fail to rebound”
Being the key word…
- Sam Altman
https://twitter.com/sama/status/1599671496636780546
Source: https://blogs.microsoft.com/blog/2023/01/23/microsoftandopen...
Keyword being “extend” not “start”.
It is not. It's still 3.5:
"Microsoft says these features are all powered by an upgraded version of GPT 3.5, the AI OpenAI language model that powers ChatGPT. Microsoft calls this the “Prometheus Model,” and says it’s more powerful than GPT 3.5, and better able to answer search queries with up-to-date information and annotated answers." (https://www.theverge.com/2023/2/7/23587454/microsoft-bing-ed...)
Do you mean his interview 3 weeks ago?
https://www.youtube.com/watch?v=eVcJJZz9P4U
(Edit: On Firefox/Linux)
I think this tech is going to be transformative for everyone in a positive way. I've been doing a lot of bookkeeping and form-filling-out the past couple weeks and I keep thinking, wow, I wish I could just explain to ChatGPT what I need copied and pasted from one document to another or what needs to be filled out.
1/2 of the time I search google while not logged in, it asks me about logging in or creating an account.
About the same rate I get bothered logging into gmail using a non-chrome browser and they need to tell me I should experience Gmail with chrome.
I remember having to spend a little while de-shittifying it, they push the Shopping/Rewards stuff pretty hard and clutter up the UI. You can't get rid of Bing on the new tab page without extensions / other hoops.
so... yes. yes it does.
https://i.imgur.com/2f1D53L.png
The settings for modifying your layout and turning off all the content you don't want is available through the gear icon settings from the top right of the page. Would be nice if the default were clean, but you can opt out of all of that stuff.
I'd love to love Edge because it's a big improvement over Chrome just by having vertical tabs, but both companies are too greedy to provide a good browser.
This gives me back the good old about:blank new tab page. I use the URL bar to search for stuff.
https://github.com/perfect-things/perfect-home
https://github.com/anic17/homepage
https://github.com/Lissy93/dashy
https://github.com/benphelps/homepage
https://github.com/jeroenpardon/sui
https://github.com/pawelmalak/flame
https://github.com/migueravila/Bento
Edit: nevermind.. weird UI glitches abound
On the other hand, I'm surprised to see that on MacOS Edge properly shows the search/URL bar in full screen if I move the cursor to the top of the screen or push Command-L, something Chrome *still* can't do without enabling "Always Show Toolbar in Full Screen", which I don't want. This alone, plus Edge being compatible with all Chrome extensions, might cause me to switch.
I like safari, always have done. But for those sites that are ~~IE6~~ Chrome debugged only I think I’ll start dropping into Edge instead of Chrome.
It lets you change your default search provider, but even after doing that, the new tab page is still a big customized Bing landing page with a huge Bing search box.
I think they tap into some magic Windows APIs that regular Chrome does not.
Q: Will this be available in Chrome or other browsers?
Yusuf: Our intention is to bring it to all browsers. We're starting with Edge. Chrome has to implement some things to make it work, but our intention is all browsers.
Is this something in the same problem space as ChatGPT? I'm not sure an LLM would be well suited for that, but could be wrong.
Once Apple is forced to open up iOS to other browsers, the blink/chromium monopoly will be complete.
The only check on Google's ability to dictate what a web browser is will be the influence of companies like Microsoft, who are invested in blink/chromium and have the resources to potentially fork it if Google does something they don't like.
I would not be surprised if we see a blink/chromium based Safari in the future so that Apple can join the party as well.
Beyond that, even discounting the threat of ecological collapse due to a single codebase, most of the web's best standards have come from lessons learned from multiple implementations. Without multiple implementations we lose a lot of guide rails away from bad single vendor "standards". Even if you think downstream players like Microsoft can act as a big enough check on Google's rubber stamp in standards boards like WHATWG, there's a huge difference between Microsoft developers evaluating potential standards because they have to build their own implementation of them versus just reading them in code review to accept an upstream PR they didn't need to write themselves.
Also, as close to a single implementation as WebKit and Blink still sometimes are, it's hard to argue that Apple isn't already still a contributor to the WebKit/Blink/Chromium Hegemony. It doesn't matter if iOS opens up to other browsers or not right now while WebKit and Blink are still so close to mostly not matter. Apple is sort of protecting everyone from a true V8 (JS Engine) monopoly, but the renderer has always been close enough to not matter to those terrible web developers only testing websites in Chrome and assuming they work "everywhere" and supporting a "Works Only in Chrome" agenda.
Depending on your view of Firefox OS and if it was "Firefox enough", the "Mozilla Foundation doesn't focus enough on Firefox" is a relatively new complaint, but Firefox's biggest losses in market share happened well before even the Firefox OS effort.
(And the Firefox OS effort was to try to keep a competitor against Android and its Chrome hegemony viable, so it was a direct reaction to lost market share. And the current complaints of "Mozilla isn't doing enough Firefox" today are all the complaints about the various ways that Mozilla is trying to diversify their revenue stream and that too seems obviously because of lost market share, not the cause of it.)
Do people also like "diversity" in C++ compiler implementations with "diverse" quirks and bugs unique to each implementation? or the diversity of Python interpreters? What's the point other than incompatibility and head ache?
Not a C developer (and welcome C devs to chime in here) but my understanding is yes. People were really happy when we went from GCC as being the standard compiler to LLVM.
> the diversity of Python interpreters?
Yes, back (a decade ago) when I used to hack Python it was pretty cool to be able to get IronPython and re-use a bunch of .net assemblies in a language I liked.
Right now in TypeScript land people are enjoying deno provide competition to node.
One reason they are expected to coexist for some time to come is exactly that a lot of Linux (in particular) applications and libraries require gcc specifics. The fact that there are "gcc specifics" that lock those applications and libraries to gcc and can't just "upgrade to the obviously newer/better architecture" is itself an argument for "diversity". C/C++ is "supposed to be" a portable language. If developers were keeping to standards and if clang were indeed objectively better architecture, then nothing should still be using gcc, right? Diversity is one way you encourage developers to stick to portable standards (because then they can use whichever compiler is fastest/better today and switch at-will as all the implementations compete to outperform each other).
Imagine if there's a WebKit 0-Day that Blink inherited discovered tomorrow that breaks ~96% of today's browsers as soon as that bad actor discoverer weaponizes it. (Or worse uses it to send malware or as an RCE vector.)
That's not something Google did intentionally. That's not something that forks can "immediately" fix. That's something that possibly forks make worse because everyone is going to be auditing different but similar codebases for the same bug and there may be some political infighting and finger pointing and accusations in a mexican standoff between the forks all accusing each other.
Most importantly, that is a lot of potential devices at major security risk to a single ecology "family of forks" of a code base.
What other piece of software does something like 96% of internet connected devices possibly have in common? We've got a variety of operating systems (Windows, Linux, macOS, iOS, Android, etc). We've got a variety of hardware stacks. We've got a variety almost everywhere but this one massive bottleneck growing worse. There's still some hope in the case of the hypothetical WebKit 0-Day that that diversity alone is enough to keep things from getting truly bad, but if you believe in software security you've probably had "defense in depth" drilled into your head and does "everyone uses the same browser codebase" sound like defense in depth?
That's the biggest nightmare scenario. There's lots of little iterations of that, including all the various little ways that standards might slowly get ignored or broken until some day the future web is wondering where the standards specs even are and can't find its way out of the box it was trapped in. If it happens little-by-little enough ("boiling the frog" as the aphorism goes) no one sees it as a clear threat at the time and no one thinks to "immediately fork it" until it is already far too late. To repeat something that needs repeating a lot in these conversations: the problem with IE6 wasn't that it was terrible and "behind", the problem with IE6 was that it was amazing and forward thinking and had boiled some frogs building specifications too quickly ahead of the standards bodies. People didn't notice the problems they had from that until after they'd built websites and apps with IE6 in mind because it was great at the time. People didn't notice the problems until after Microsoft felt they had done enough and took their ball and stopped playing because they had "innovated enough". People mostly didn't notice the problems until it was too late to get out of the box they'd been trapped in without realizing it. IE6 today has a reputation as a terrible browser that was stuck "behind" standards, but that's not where it did the most damage. It did the most damage when it was "the best browser to use", "the most innovative and powerful browser" and "everyone uses IE6 and is happy with it".
IMHO, some software (eg. OpenSSL, rendering engines, etc.) need to be heavily scrutinized and trying to keep multiple implementation around for the sake of diversity makes no sense to me.
Of course if a new implementation of say a rendering engine with a radically different approach emerges I'm all for it taking over.
Even (especially?) highly scrutinized libraries seem to need diversity in implementation to avoid single point of failure problems.
We've even seen hints that all is not paradise even when that dominant vendor is "open source" in the post-gcc era where there was a long run of years where gcc was extremely dominant and there were concerns about platform-wide 0 days and where the gcc developers were playing fast and loose with the standards. Certainly those were "open source community decisions", but it was still too easy to PR non-standard features "that felt good" without going through deeper standards review processes.
It's generally seen as a great thing that today we aren't in one of those "one vendor dominated" periods and that we have both gcc and clang as competing, independent open source implementations each with relatively high adoption and slightly different niches/portability goals/downstream uses. It's also done much to help push commercial vendors back to competing on standards compliance. (Microsoft's C/C++ compiler is more standards compliant than ever, for example. Including its STL is now open source.)
Embrace, Extend. That's the beauty of open source, right? If development and innovation slows, somebody can fork.
Wonder what comes next from Microsoft after Embrace and Extend ...
Mozilla/5.0 (Macintosh; Intel Mac OS X 13_2) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/110.0.0.0 Safari/537.36 Edg/109.0.1518.78
Recently, in a conversation with ChatGPT, I gave him a system administration task and told him I was a Linux terminal running bash and I would paste and execute everything he told me into the shell (and respond with the output). It worked okay – I wouldn't recommend it for production use yet – but I think, with future iterations of language transformer models, interfaces like the one you're describing as well as others (AI <> terminal, AI <> X Server, AI <> speakers/microphone) and combinations thereof stand a very good chance of becoming integral to our daily work.
Writing code for me...yes, thank you! Helping me get through to my teenage daughter, that is amazing.
Another thing, is that if people get indoctrinated to expect bad information from the machine, it might have some impact in correcting our society's propensity for spreading and believing false information.
In short, while individuals in our community might understand more about the underlying process, I'm not sure it would be accepted as more credible to those already skeptical of anything that goes against their opinion?
I mainly use Duckduckgo and Startpage but sometimes Bing, especially in Edge.
The answer I got here: https://www.bing.com/search?q=What%20cars%20should%20I%20con...
Says the Tesla Model Y can go from 0 to 60 in 4.8 seconds, and has a citation link to a page that... doesn't even mention Tesla once.
Did they not roll this out in the EU yet or something?
> That’s a very specific question. Based on my web searches, there are not many cars that meet all of your criteria. However, I found some possible options that you might want to consider:
> The 2022 Kia Telluride is a midsize SUV that can seat up to eight passengers and has an AWD option. It has a 3.8-liter V6 engine that produces 291 hp and 262 lb-ft of torque. It can accelerate from 0 to 60 mph in 7.1 seconds10 and has a combined fuel economy of 21 mpg. It also has excellent reviews from critics and owners, and won several awards, including the 2020 World Car of the Year7.
> The 2022 Hyundai Palisade is another midsize SUV that can seat up to eight passengers and has an AWD option. It has a similar engine and performance as the Kia Telluride, with a 0 to 60 mph time of 7.2 seconds10 and a combined fuel economy of 21 mpg. It also has a high-quality interior, advanced safety features, and a spacious cargo area7.
> The 2022 Tesla Model Y is a compact electric SUV that can seat up to seven passengers with an optional third row. It has a dual-motor AWD system that delivers 384 hp and 376 lb-ft of torque. It can accelerate from 0 to 60 mph in 4.8 seconds10 and has an EPA-estimated range of 326 miles. It also has a futuristic design, a large touchscreen, and a high-tech infotainment system10.
> The 2022 Volvo XC90 is a luxury midsize SUV that can seat up to seven passengers and has an AWD option. It has a 2.0-liter turbocharged and supercharged four-cylinder engine that produces 316 hp and 295 lb-ft of torque. It can accelerate from 0 to 60 mph in 6.1 seconds10 and has a combined fuel economy of 22 mpg. It also has a refined interior, a smooth ride, and a plethora of safety features4.
> The 2022 Acura MDX is another luxury midsize SUV that can seat up to seven passengers and has an AWD option. It has a 3.5-liter V6 engine that produces 290 hp and 267 lb-ft of torque. It can accelerate from 0 to 60 mph in 6.4 seconds10 and has a combined fuel economy of 23 mpg. It also has a sporty handling, a spacious cabin, and a user-friendly infotainment system4.
The "0-60 in x seconds" in each paragraph were bolded, and the numbers next to it are citations.
> The "0-60 in x seconds" in each paragraph were bolded, and the numbers next to it are citations.
And if you follow the citation to the source... where does the source mention the 0 to 60 speed for the Tesla Model Y?
Nowhere. It was hallucinated. It invented a citation that doesn't exist. Ctrl+f for Tesla on the so-called "source" that it cited. It doesn't appear once, as of this comment.
The number itself is not incorrect. Tesla Model Y (AWD version) does have an advertised 0-60 time of 4.8s.
It will significantly affect smaller so-called AI companies like Jasper.ai than Google. But this announcement is unsurprising and as expected. [0]
I will reiterate again. The only way to disrupt OpenAI especially ChatGPT, is for an open source version of ChatGPT which is smaller and can run offline. This happened with Stable Diffusion towards DALLE-2.
A open source model that matches ChatGPT and even bests GPT-4 extinguishes the need to use OpenAI’s APIs and will change everything.
[0] https://news.ycombinator.com/item?id=34684700
No. This has nothing to do with a statement made 20+ years ago. There is no "Embrace" involved. There is no "Extend" involved. There is no "Extinquish" involved. There is an obvious business and competitive step involved, and a major business relationship with OpenAI involved.
It's time to grow up and stop acting as if using terminology and aspersions from the era of Fax machine is somehow hip and insightful. Every single person involved in the EEE era retired from Microsoft years ago. Most HN readers weren't even born in those days.
Yes it does.
The methods are different but the strategy is the same. You can’t compete against free, and Microsoft can foot the bill for their services without charging for it for years to suffocate competitors.
> It's time to grow up and stop acting as if using terminology and aspersions from the era of Fax machine is somehow hip and insightful.
Anyone who thinks Microsoft has changed their ways and stopped their past tactics has just bought into their lies and ‘Microsoft Loves Open Source’ nonsense.
If the exclusive OpenAI licenses and the acquisition of GitHub and other market leaders are not an indication of great EEE candidates to squeeze out competitors for offering their services close to free, then I don’t know what is.
I agree with your sentiment - anyone who holds a grudge from 30 years ago and can’t update or change their thinking and perspective (especially in light of new data and experiences) is living a disadvantaged life.
For perspective - the United States and Vietnam normalized relations in 20 years! After a war in which 1.6m people were killed.
In 1996 GM released the EV1 and eventually crushed them and pulled them off the market, largely due to a cabal in the audio industry and all other kinds of shenanigans[0].
In 2011 GM released the Volt, with the Bolt in 2016 and now they have roughly half a dozen all electric models on the market.
It seems as though people, corporations, and the people that make up corporations can change in several decades.
With some notable exceptions these kinds of examples can be found throughout history and at this point I’d argue they’re more the norm than the exception.
[0] - https://en.m.wikipedia.org/wiki/General_Motors_EV1
I'm not sure if it's not live yet or if I'm not seeing it because I'm checking from a Mac (via both Chrome and the OSX version of Edge)
Edit - Well, it gives me the waitlist at least. Possibly because I am in the MS corp net - changes may not have propagated? IDK.
Now I get the "new" homepage but can only do pre-set queries.
Firefox on Linux, India.
What is this new trend of announcing things as if it's available for general use and then hiding it behind invites or not making it available at all.
Off topic: Anyone knows how to get rid of wall of news on bing homepage? I tried disabling it but it still loads it and makes it visible if I scroll down.
If they were worried about getting a bad start with the general public or scaling concerns they could have easily put it somewhere that only enthusiasts (e.g. HN) would find it. But no, instead...announcement and nothing usable.
Call me back when I can try this, Bing.
[0] https://blog.google/technology/ai/bard-google-ai-search-upda...
Bing does this as well, that isn't new to either platform. It's useful but also is wrong a lot. (ChatGPT and Bard [terrible name] will also get a lot wrong.)