Yes, that was obvious as soon as I saw it wasn’t live I clicked off.
You can train any LLM to perform a certain task(s) well and google engineers are not that dense.
This was obvious marketing PR as open AI has completely made google basically obsolete with 90% of my queries can be answered without wading through LLM generated text for a simple answer.
If LLMs can replace 90% of your queries, then you have very different search patterns from me. When I search on Kagi, much of the time I’m looking for the website of a business, a public figure’s social media page, a restaurant’s hours of operation, a software library’s official documentation, etc.
LLMs have been very useful, but regular search is still a big part of everyday life for me.
Sure we now have options, but before LLMs, most queries relied solely on search engines, often leading to sifting through multiple paragraphs on websites to find answers — a barrier for most users.
Today, LLMs excel in providing concise responses, addressing simple, curious questions like, "Do all bees live in colonies?"
Though because you don’t see the answers it doesn’t show you, it’s hard to really validate the quality, so I’m still wary, but when I look for specific stuff it tends to find it.
The video itself and the video description give a disclaimer to this effect. Agreed that some will walk away with an incorrect view of how Gemini functions, though.
Hopefully realtime interaction will be part of an app soon. Doesn’t seem like there would be too many technical hurdles there.
It can be realtime while still having more latency than depicted in the video (and the video clearly stated that Gemini does not respond that quickly).
A local model could send relevant still images from the camera feed to Gemini, along with the text transcript of the user’s speech. Then Gemini’s output could be read aloud with text-to-speech. Seems doable within the present cost and performance constraints.
I remember watching it and I was pretty impressed, but as I was walking around thinking to myself I came to the conclusion that there was something fishy about the demo. I didn't know exactly what they fudged, but it was far too polished to explain how well their current AI demos preform.
I'm not saying there have been no improvements in AI. There is and this includes Google. But the reason why ChatGPT has really taken over the world is that the demo is in your own hands and it does quite well there.
Indeed, and this is how Google used to be as a company. I remember when Google Maps & Earth launched, and how they felt like world-changing technology. I'm sure they're doing lots of innovative science and development still, but it's and advertising/services company now, and one that increasingly talks down to its users. Disappointing considering their early sense of mission.
Thinking back to the firm's early days, it strikes me that some HN users and perhaps even some Googlers have no memory of a time before Google Maps and simply can't imagine how disruptive and innovative things like that were at the time. Being able to browse satellite imagery for the whole world was something previously confined to the upper echelons of the military-industrial complex.
That's one reason I wish the firm (along with several other tech giants) were broken up; it's full of talented innovative people, but the advertising economics at the core of their business model warp everything else.
What I found impressive about it was the voice, the fast real-time response to video, and the succinct responses. So apparently all of that was fake. You got me, Google.
The entirety of the disclaimer is "sequences shortened throughout", in tiny text at the bottom for two seconds.
They do disclose most of the details elsewhere, but the video itself is produced and edited in such a way that it's extremely misleading. They really want you to think that it's responding in complex ways to simple voice prompts and a video feed, and it's just not.
Yea, of all the edits in the video, the editing for timing is the least of concern. My gripe is that the prompting was different and in order to get that information you have to watch the video only on YouTube, expand the description and click on a link to a different blog article. Linking a "making of" video where they show this and interview some of the minds behind Gemini would have been better PR.
They were just parroting this video on CNBC without any disclaimers, so the viewers who don't happen to also read hacker news will likely form a different opinion than those of us who do.
Yeah, and ads on Google search have the teeniest, tiniest little "ad" chip on them, a long progression of making ads more in-your-face and less well-distinguished.
In my estimation, given the context around AI-generated content and general fakery, this video was deceptive. The only impressive thing about the video (to me) was how snappy and fluid it seemed to be, presumably processing video in real time. None of that was real. It's borderline fraudulent.
The difference between “Hey, figure out a game based on what you see right now” vs “here is a description of a game with the only too possible outcomes as examples” cannot be explained by the disclaimer.
I suppose it's not false advertising, since they don't even claim to have a product released yet that can do this, since Trojans Ultra won't be available until an unspecified time next year
You're right, it's astroturfing a placeholder in the market in the absence of product. The difference is probably just the target audience - feels like this one is more aimed at share-holders and internal politics.
This is common in all industries. Take gaming, for example. Game publishers love this kind of publicity, as it creates hype, which leads to sales. There have been numerous examples of this over the years: Watch Dogs, No Man's Sky, Cyberpunk 2077, etc. There's a period of controversy once consumers realize they've been duped, the company releases some fake apology and promises or doubles down, but they still walk out of it richer, and ready to do it again next time.
It's absolutely insidious, and should be heavily fined and regulated.
As a programmer, I'd say that all the demos of my code were honest and representative of what my code was doing.
But I recognize we're all different programmers in different circumstances. But at a minimum, I'd like to be honest with my work. My bosses seem to agree with me and I've never been pressured into hosting a fake demo or lie about the features.
In most cases, demos are needed because there's that dogfood problem. Its just not possible for me to know how my (prospective) customers will use my code. So I need to show off what has been coded, my progress, and my intentions for the feature set. In response, the (prospective) customer may walk away, they may have some comments that increases the odds of adoption, or they think its cool and amazing and take it on the spot. We can go back and forth with regards to feature changes or what is possible, but that's how things should work.
------------
I've done a few "I could do it like this" demos, where everyone in the room knew that I didn't finish the code yet and its just me projecting into the future of how code would work and/or how it'd be used. But everyone knew the code wasn't done yet (despite that, I've always delivered on what I've promised).
There is a degree of professional ethics I'd expect from my peers. Hosting honest demos is one of them, especially with technical audience members.
OP says that Gemini had still images as input, not video - and the dev blog post shows it was instructed to reply to each input in relevant terms. Needless to say, that's quite different from what's implied in the demo, and at least theoretically is already within GPT's abilities.
Google always does fake advertising. “Unlimited” google drive accounts for example. They just have such a beastly legal team no one is going to challenge them on anything like that.
What was fake about unlimited google drive? There were some people using petabytes.
The eventual removal of that tier and anything even close speaks to Google's general issues with cancelling services, but that doesn't mean it was less real while it existed.
What about when gmail was released and the storage was advertised as increasing forever, but at first they just increased it infinitesimally slower and then stopped increasing it all.
I don't remember the "increasing forever" ever being particularly fast. I found some results from 2007 and 2012 both saying it was 4 bytes per second, <130MB per year.
So it's true that the number hasn't increased in ten years, but that last increase was +5GB all by itself. They've done a reasonable job of keeping up.
Arguably they should have kept adding a gigabyte each year, based on the intermittent boosts they were giving, but by that metric they're only about 5GB behind.
I'll admit I was fooled. I didn't read the description of the video. The most impressive thing they showed was the real-time responses to watching a video. Everything else was about expected.
Very misleading and sad Google would so obviously fake a demo like this. Mentioning in the description that it's edited is not really in the realm of doing enough to make clear the fakery.
i too was excited and duped about the real-time implications. though i'm not surprised at all to find out it's false.
mea cupla i should have looked at the bottom of the description box on youtube where it probably says "this demonstration is based on an actual interaction with an LLM"
Good, that video was mostly annoying and creepy. The AI responses as shown in the linked Google dev blogpost are a lot more reasonable and helpful. BTW I agree that the way the original video was made seems quite misleading in retrospect. But that's also par for the course for AI "demos", it's an enduring tradition in that field and part of its history. You really have to look at production systems and ignore "demos" and pointless proofs of concept.
The GPT-4 demo early this year when it was released was a lot less.. fake, and in fact very much indicative of it's feature set. The same is true for what OpenAI showed during their dev days, so at the very least those demos don't have too much fakery going on, as far as I could tell.
A certain minimum level of jank always makes demos more believable. Watching Brockman wade through Discord during the napkin-to-website demo immediately made the whole thing convincing.
AI is in the "hold it together with hope and duct tape" phase, and marketing videos claiming otherwise are easy to spot and debunk.
>You really have to look at production systems and ignore "demos" and pointless proofs of concept.
While I agree, I wouldn't call proofs or concepts and demos pointless. They often illustrate a goal or target functionality you're working towards. In some cases it's really just a matter of allotting some time and resources to go from a concept to a product, no real engineering is needed, it all exists, but there's capital needed to get there.
Meanwhile some proof of concepts skip steps and show higher level function that needs some serious breakthrough work to get to, maybe multiple steps of that. Even this is useful because it illustrates a vision that may be possible so people can understand and internalize things you're trying to do or the real potential impact of something. That wasn't done here, it was embedded in a side note. That information needs to be before the demo to some degree without throwing a wet blanket on everything and needs to be in the same medium as the demo itself so it's very clear what you're seeing.
I have no problem with any of that. I have a lot of problems when people don't make it explicitly clear beforehand that it's a demo and explain earnestly what's needed. Is it really something that exists today in working systems someone just needs to invest money and wire it up without new research needed? Or is it missing some breakthroughs, how many/what are they, how long have these things been pursued, how many people are working on them... what does recent progress look like and so on (in a nice summarized fashion).
Any demo/poc should come up front with an earnest general feasibility assessment. When a breakthrough or two are needed then that should skyrocket. If it's just a lot of expensive engineering then that's also a challenge but tractable.
I've given a lot of scientific tech demonstrations over the years and the businesses behind me obviously want me to be as vague as possible to pull money in. I of course have some of those same incentives (I need to eat and pay my mortgage like everyone else). None-the-less the draw of science to me has always been pulling the veil from deception and mystery and I'm a firm believer in being as upfront as possible. If you don't lead with disclaimers, imaginations run wild into what can be done today. Adding disclaimers helps imaginations run wild about what can be done tomorrow, which I think is great.
I suppose this is a great example of how trust in authentic videos, audio, images, company marketing must be questioned and, until verified, assumed to be 'generated'.
I am curious, if the voice, email, chat, and shortly video can all be entirely generated in real or near real time, how can we be sure that remote employee is actually not a full or partially generated entity?
Shared secrets are great when verifying but when the bodies are fully remote - what is the solution?
I am traveling at the moment. How can my family validate that it is ME claiming lost luggage and requesting a Venmo request?
Fair, but that also assumes the recipients ("family") are in a mindset of constantly thinking about the threat model in this type of situation and will actually insist on hearing the passphrase.
You only know one piece of information about your family? I feel like I could reference many childhood facts or random things that happened years ago in social situations.
The question is if an attacker tells you they lost access can you please reset some credential, and your security process is getting on a video call because you're a fully remote company let's say.
I think it's also why we as a community should speak out when we catch them for doing this as they are discrediting tech demos. It won't be enough because a lie will be around the world before the truth gets out the starting gates but we can't just let this go unchecked.
If I demoed swype texting as it functions in my day to day life to someone used to a querty keyboard they would never adopt it
The rate at which it makes wrong assumptions about the word, or I have to fix it is probably 10% to 20% of the time
However because it’s so easy to fix this is not an issue and it doesn’t slow me down at all. So within the context of the different types of text Systems out there, I t’s the best thing going for me personally, but it takes some time to learn how to use it.
This is every product.
If you demonstrated to people how something will actually work after 100 hours of habituation and compensation for edge cases, nobody would ever adopt anything.
I’m not sure how to solve this because both are bad.
(Edit: I’m keeping all my typos as meta-comment on this given that I’m posting via swype on my phone :))
Does swype make editing easier somehow? iOS spellcheck has negative value. I turned it off years ago and it reduced errors but there are still typos to fix.
Unfortunately iOS text editing is also completely worthless. It forces strange selections and inserts edited text in awkward ways.
I’m a QWERTY texter but text entry on iOS is a complete disaster that has only gotten worse over time.
I'm an iOS user and prefer the swipe input implementation in GBoard over the one in the native keyboard. I'm not sure what the differences are, but GBoard just seems to overall make fewer mistakes and do a better job correcting itself from context.
As I was reading Andrew's comment to myself, I was trying to figure out when and why I stopped using swype typing on my phone. Then it hit me – I stopped after I switched from Android to iOS a few years ago. Something about the iOS implementation just doesn't feel right.
But you can install other keyboards like SwiftKey or Gboard which are closer to what you are used to on Android.
My only issue is that no keyboard implementation really supports more than two languages which makes me switch back to plain qwerty with autocomplete all the time.
Apple autocorrect has a tendency to replace technical terms with similar words, eg. rvm turns into rum or ram or something.
It's even worse on the watch somehow. I take care to hit every key exactly, the correct word is there, I hit space, boom replaced with a completely different word. On the watch it seems to replace almost every word with bullshit, not just technical terms.
> seems to replace almost every word with bullshit
Sort of related, it also doesn't let you cuss. It will insist on replacing fuck with pretty much anything else. I had to add fuck to the custom replacement dictionary so it would let me be. What language I choose to use is mine and mine alone, I don't want Nanny to clean it up.
They've pretty much solved this with iOS 17. You can even use naughty words now, provided you use it for a day or so to have it get used to your vocabulary.
Its honestly pretty mind boggling that we’d even use querty on a smartphone. The entire point of the layout is to keep your fingers on the home row. Meanwhile people text with a single or two thumbs 100% of the time.
"The entire point of the layout is to keep your fingers on the home row."
No, that is how you're told to type. You have to be told to type that way precisely because QWERTY is not designed to keep your fingers on the home row. If you type in a layout that is designed to do that, you don't need to be told to keep your fingers on the home row, because you naturally will.
Nobody really knows what the designers were thinking, which I do not mean as sarcasm, I mean it straight. History lost that information. But whatever they were thinking that is clearly not it because it is plainly obvious just by looking at it how bad it is at that. Nobody trying to design a layout for "keeping your fingers on the home row" would leave hjkl(semicolon) under the resting position of the dominant hand for ~90% of the people.
This, perhaps in one of technical history's great ironies, makes it a fairly good keyboard for swype-like technologies! A keyboard layout like Dvorak that has "aoeui" all right next to each other and "dhtns" on the other would be constantly having trouble figuring out which one you meant between "hat" and "ten" to name just one example. "uio" on qwerty could probably stand a bit more separation, but "a" and "e" are generally far enough apart that at least for me they don't end up confused, and pushing the most common consonants towards the outer part of the keyboard rather than clustering them next to each other in the center (on the home row) helps them be distinguishable too. "fghjkl" is almost a probability dead zone, and the "asd" on the left are generally reasonably distinct even if you kinda miss one of them badly.
I don't know what an optimal swype keyboard would be, and there's probably still a good 10% gain to be made if someone tried to make one, but it wouldn't be enough to justify learning a new layout.
> Nobody really knows what the designers were thinking, which I do not mean as sarcasm, I mean it straight. History lost that information.
My understanding of QWERTY layout is that it was designed so that characters frequently used in succession should not be able to be typed in rapid succession, so that typewriter hammers had less chance of colliding. Or is this an urban myth?
My understanding (which is my recollections of a dive into typewriter history decades ago) is that avoiding typebar collisions was a real concern, but that the general consensus was that the exact final layout was strongly influenced by allowing salesmen to quickly type out 'typewriter' on the top row of letters.
You have to be taught to use the home row because the natural inclination for most people is to peck and hunt with their two index fingers. Watch how old people or young kids type. That being said staying on the home row is how you type fast and make the most of the layout. Everything is comfortably reachable for the most part unless you are a windows user ime.
If you learn a keyboard layout where the home row is actually the most common keys you use, you will not have to be encouraged to use the home row. You just will. I know, because I have, and I never "tried" to use the home row.
People don't hunt and peck after years of keyboard use because of the keyboard; they do it because of the keyboard layout.
If you want to prove I'm wrong, go learn Dvorak or Colemak and show me that once you're comfortable you still hunt and peck. You won't be, because it wouldn't even make sense. Or, less effort, find a hunt & peck Dvorak or Colemak user who is definitely at the "comfortable" phase.
People will actually avoid using their nondominant fingers like their ring or pinky. Its an issue with typing irrespective of layout. Its an issue with guitarplaying even. I am no pianist but I wouldn’t be surprised if new players have an aversion towards using those fingers as well.
I understand how dvorak is designed. I am still not convinced people will be using all their fingers especially their pinkys in a consistent manner that without learning that this is what you should work towards.
I really shouldn't be astonished at the number of people who will prioritize their own theorizing over the field report of someone who actually did the thing in question, but yet still somehow I am.
I did it. It happened. Theorizing about why it didn't happen is not terribly productive.
Hold up young one. The reason for QWERTYs design has absolutely not been lost to history yet.
The design was to spread out the hammers of the most frequently used letters to reduce the frequency of hammer jamming back when people actually used typewriters and not computers.
The problem it attempted to improve upon, and which is was pretty effective at, is just a problem that no longer exists.
The original intent I do believe was not separating the hammers per se, but also helping the hands alternate, so they would naturally not jam as much.
However, I use a Dvorak layout and my hands feel like they alternate better on that due to the vowels being all on one hand. The letters are also in more sensical locations, at least for English writing.
It can get annoying when G and C are next to each other, and M and W, but most of the time I type faster on Dvorak than I ever did on Qwerty. It helps that I learned during a time where I used qwerty at work and Dvorak at home, so the mental switch only takes a few seconds now.
And it does a bad job at it, which is further evidence that it was not the design consideration. People may not have been able to run a quick perl script over a few gigabytes of English text, but they would have gotten much closer if that was the desire. I don't believe that was their goal but they were just too stupid to get it even close to right.
The reason we use qwerty on a smartphone is extremely straightforward: people tend to know where to look for the keys already, so it's easy to adopt to even though it's not "efficient". We know it better than we know the positions of letters in the alphabet. You can easily see the difference if you're ever presented with an onscreen keyboard that's in alphabetical order instead of qwerty (TVs do this a lot, for some reason, and it's a different physical input method but alpha order really does make you have to stop and hunt). It slows you down quite a bit.
That's definitely a good reason why, but perhaps if iOS or Android were to research what the best layout is for typical touch screen typing and release that as a new default, people would find it quite quick to learn a second layout and soon get just the benefits?
After all, with TVs I've had the same experience as you with the annoying alphabetical keyboard, but we type into they maybe a couple of times a year, or maybe once in 5 years, whereas if we changed our phone keyboard layout we'd likely get used to it quite quickly.
Even if not going so far as to push it as a new default for all users (I'm willing to accept the possibility that I'm speaking for myself as the kind of geeky person who wouldn't mind the initial inconvenience of a new kb layout if it meant saving time in the long run, and that maybe a large majority of people would just hate it too much to be willing to give it a chance), they could at least figure out what the best layout is (maybe this has been studied and decided already, by somebody?) and offer that as an option for us geeks.
Even most technically-minded people still use QWERTY on full-size computer keyboards despite it being a terrible layout for a number of reasons. I really doubt a new, nonstandard keyboard would get much if any traction on phones.
It only worked because it had to, given phones just had keypads back then. As soon as qwerty-with-your-thumbs was available, everyone abandoned T9 and never looked back.
Path dependency is the reason for this, and is the reason why a lot of things are the way they are. An early goal with smart phone keyboards was to take a tool that everyone already knew how to use, and port it over with as little friction as possible. If smart phones happened to be invented before external keyboards the layouts probably would have been quite different.
I know marketing is marketing, but it's bad form IMO to "demo" something in a manner totally detached from its actual manner of use. A swype keyboard takes practice to use, but the demos of that sort of input typically show it being used in a realistic way, even if the demo driver is an "expert".
This is the sort of demo that 1) gives people a misleading idea of what the product can actually do; and 2) ultimately contributes to the inevitable cynical backlash.
If the product is really great, people can see it in a realistic demo of its capabilities.
Showing a product in its best light is one thing. Demonstrating a mode of operation that doesn't exist is entirely another. It would be like if a demo of your swipe keyboard included telepathic mind control for correcting errors.
I’m not sure I’d agree that what they showed will never be possible and in fact my whole point is that I think Google can most likely deliver on that in this specific case. Chalk it up to my experience in the space, but from what I can see it looks like something Google can actually execute on (unlike many areas where they fail on product regularly).
I would agree completely that it’s not ready for consumers the way it was displayed, which is my point.
I do want to add that I believe that the right way to do these types of new product rollout is not with these giant public announcements.
In fact, I think generally speaking the “right” way to do something like this demonstrates only things that are possible robustly. However that’s not the market that Google lives in. They’re capitalists trying to make as much money as possible. I’m simply evaluating that what they’re showing I think is absolutely technically possible and I think Google can deliver it even if its not ready today.
Do I think it’s supremely ethical the way that they did it? No I don’t.
The voice interaction part didn't look a far cry from what we are doing with Dynamic Interaction at SoundHound. Because of this I assumed (like many it seems) that they had caught up.
And it's dangerous to assume they can just "deliver later". It's not that simple. If it is why not bake it in right now instead of committing fraud?
This is damaging to companies that walk the walk and then people have literally said to me "but what about that Gemini"? and dismiss our work.
That was basically what magic leap did to the whole AR development market. Everyone deep in it knew they couldn’t do it but they messed up so badly that it basically killed the entire industry
I don't care what google could, in theory, deliver on some time in the future maybe. That's irrelevant. They are demonstrating something that can't be done with the product as they are selling it.
> However because it’s so easy to fix this is not an issue and it doesn’t slow me down at all.
But that's a different issue than LLM hallucinations.
With Swype, you already know what the correct output looks like. If the output doesn't match what you wanted, you immediately understand and fix it.
When you ask an LLM a question, you don't necessarily know the right answer. If the output looks confident enough, people take it as the truth. Outside of experimenting and testing, people aren't using LLMs to ask questions for which they already know the correct answer.
I think you mean swipe. Swype was a brilliant third party keyboard app for Android which was better at text prediction and manual correction than Gboard is today. If however you really do still use Swype then please tell me how because I miss it.
Ha good point, and yes I agree Swype continues to be the best text input technology that I’ll never be able to use again. I guess I just committed genericide here but I meant the general “swiping” process at this point
The insight here is that the speed of correction is a crucial component of the perceived long-term value of an interface technology.
It is the main reason that handwriting recognition did not displace keyboards. Once the handwriting is converted to text, it’s easier to fix errors with a pointer and keyboard. So after a few rounds of this most people start thinking: might as well just start with the pointer and keyboard and save some time.
So the question is, how easy is it to detect and correct errors in generative AI output? And the unfortunate answer is that unless you already know the answer you’re asking for, it can be very difficult to pick out the errors.
Yeah the feedback loop with consumers has a higher likelihood of being detrimental, so even if the iteration rate is high, it’s potentially high cost at each step.
I think the current trend is to nerf the models or otherwise put bumpers on them so people can’t hurt themselves. That’s one approach that is brittle at best and someone with more risk tolerance (OpenAI) will exploit that risk gap.
It’s a contradiction then at best and depending on the level of unearned trust from the misleading marketing, will certainly lead to some really odd externalities
Think “man follows google maps directions into pond” but for vastly more things.
I really hated marketing before but yeah this really proves the warning I make in the AI addendum to my scarcity theory (in my bio).
You make a decent point, but you might underestimate how much this Gemini demo is faked[0].
In your Swype analogy, it would be as if Swype works by having to write out on a piece of paper the general goal of what you're trying to convey, then having to write each individual letter on a Post-it, only for you to then organize these Post-its in the correct order yourself.
This process would then be translated into a slick promo video of someone swiping away on their keyboard.
This is not a matter of “eh, it doesn't 100% work as smooth as advertised.”
I watched this video, impressed, and thought: what if it’s fake. But then dismissed the thought because it would come out and the damage wouldn’t be worth it. I was wrong.
The worst part is that there won't be any damage. They'll release a blog post with PR apologies, but the publicity they got from this stunt will push up their brand in mainstream AI conversations regardless.
There’s no such thing as bad publicity only applies to people and companies that know how to spin it.
Reading the comments of all these disillusioned developers, it’s already damaged them because now smart people will be extra dubious when Google starts making claims.
They just made it harder for themselves to convince developers to even try their APIs, let alone bet on them.
Gemini demo looks like ChatGPT with a video feed, except it doesn't exist, like ChatGPT. I have ChatGPT on my phone right now, and it works (and it can process images, audio, and audio feed in).
This means Google has shown nothing of substance. In my world, it's a classic stock price manipulation move.
I've been using Gemini in Bard since the launch, with respect to coding it is outperforming GPT4 in my opinion. There is some convergence in the answers,but Bard is outputting really good code now.
Even a year ago, this advert would have been obvious puffery in advertising.
But right now, all the bits needed to do this already exist (just need to be assembled and -to be fair- given a LOT of polish), so it would be somewhat reasonable to think that someone had actually Put In The Work already.
That demo was much further on the "marketing" end of the spectrum when compared to some of their other videos from yesterday which even included debug views: https://youtu.be/v5tRc_5-8G4?t=43
This is endemic to public product demos. The thing never works as it does in the video. I'm not excusing it, I'm saying: don't trust public product demos. They are commercials, they exist to sell to you, not to document objectively and accurately, and they will always lie and mislead within the limits of the law.
682 comments
[ 2.8 ms ] story [ 376 ms ] thread...OpenAI solved this by generating LLM text for you to wade through?
LLMs have been very useful, but regular search is still a big part of everyday life for me.
Today, LLMs excel in providing concise responses, addressing simple, curious questions like, "Do all bees live in colonies?"
For technical questions, ChatGPT has almost completely replaced Google & Stack Overflow for me.
Though because you don’t see the answers it doesn’t show you, it’s hard to really validate the quality, so I’m still wary, but when I look for specific stuff it tends to find it.
Hopefully realtime interaction will be part of an app soon. Doesn’t seem like there would be too many technical hurdles there.
A local model could send relevant still images from the camera feed to Gemini, along with the text transcript of the user’s speech. Then Gemini’s output could be read aloud with text-to-speech. Seems doable within the present cost and performance constraints.
I'm not saying there have been no improvements in AI. There is and this includes Google. But the reason why ChatGPT has really taken over the world is that the demo is in your own hands and it does quite well there.
Thinking back to the firm's early days, it strikes me that some HN users and perhaps even some Googlers have no memory of a time before Google Maps and simply can't imagine how disruptive and innovative things like that were at the time. Being able to browse satellite imagery for the whole world was something previously confined to the upper echelons of the military-industrial complex.
That's one reason I wish the firm (along with several other tech giants) were broken up; it's full of talented innovative people, but the advertising economics at the core of their business model warp everything else.
That's different from "Gemini was shown selected still images and not video".
They do disclose most of the details elsewhere, but the video itself is produced and edited in such a way that it's extremely misleading. They really want you to think that it's responding in complex ways to simple voice prompts and a video feed, and it's just not.
The video fooled many people, including myself. This was not your typical super optimized and scripted demo.
This was blatant false advertising. Showing capabilities that do not exist. It’s shameful behavior from Google, to be perfectly honest.
In my estimation, given the context around AI-generated content and general fakery, this video was deceptive. The only impressive thing about the video (to me) was how snappy and fluid it seemed to be, presumably processing video in real time. None of that was real. It's borderline fraudulent.
edit: s/stuck/stock
This is common in all industries. Take gaming, for example. Game publishers love this kind of publicity, as it creates hype, which leads to sales. There have been numerous examples of this over the years: Watch Dogs, No Man's Sky, Cyberpunk 2077, etc. There's a period of controversy once consumers realize they've been duped, the company releases some fake apology and promises or doubles down, but they still walk out of it richer, and ready to do it again next time.
It's absolutely insidious, and should be heavily fined and regulated.
But I recognize we're all different programmers in different circumstances. But at a minimum, I'd like to be honest with my work. My bosses seem to agree with me and I've never been pressured into hosting a fake demo or lie about the features.
In most cases, demos are needed because there's that dogfood problem. Its just not possible for me to know how my (prospective) customers will use my code. So I need to show off what has been coded, my progress, and my intentions for the feature set. In response, the (prospective) customer may walk away, they may have some comments that increases the odds of adoption, or they think its cool and amazing and take it on the spot. We can go back and forth with regards to feature changes or what is possible, but that's how things should work.
------------
I've done a few "I could do it like this" demos, where everyone in the room knew that I didn't finish the code yet and its just me projecting into the future of how code would work and/or how it'd be used. But everyone knew the code wasn't done yet (despite that, I've always delivered on what I've promised).
There is a degree of professional ethics I'd expect from my peers. Hosting honest demos is one of them, especially with technical audience members.
Implying they’ve solved single token latency, however, is very distasteful.
[1] https://www.bloomberg.com/opinion/articles/2023-12-07/google...
[2] https://www.bloomberg.com/opinion/articles/2023-12-07/google...
The eventual removal of that tier and anything even close speaks to Google's general issues with cancelling services, but that doesn't mean it was less real while it existed.
I don't remember the "increasing forever" ever being particularly fast. I found some results from 2007 and 2012 both saying it was 4 bytes per second, <130MB per year.
So it's true that the number hasn't increased in ten years, but that last increase was +5GB all by itself. They've done a reasonable job of keeping up.
Arguably they should have kept adding a gigabyte each year, based on the intermittent boosts they were giving, but by that metric they're only about 5GB behind.
Very misleading and sad Google would so obviously fake a demo like this. Mentioning in the description that it's edited is not really in the realm of doing enough to make clear the fakery.
mea cupla i should have looked at the bottom of the description box on youtube where it probably says "this demonstration is based on an actual interaction with an LLM"
All they've done is completely destroy my trust in anything they present.
AI is in the "hold it together with hope and duct tape" phase, and marketing videos claiming otherwise are easy to spot and debunk.
While I agree, I wouldn't call proofs or concepts and demos pointless. They often illustrate a goal or target functionality you're working towards. In some cases it's really just a matter of allotting some time and resources to go from a concept to a product, no real engineering is needed, it all exists, but there's capital needed to get there.
Meanwhile some proof of concepts skip steps and show higher level function that needs some serious breakthrough work to get to, maybe multiple steps of that. Even this is useful because it illustrates a vision that may be possible so people can understand and internalize things you're trying to do or the real potential impact of something. That wasn't done here, it was embedded in a side note. That information needs to be before the demo to some degree without throwing a wet blanket on everything and needs to be in the same medium as the demo itself so it's very clear what you're seeing.
I have no problem with any of that. I have a lot of problems when people don't make it explicitly clear beforehand that it's a demo and explain earnestly what's needed. Is it really something that exists today in working systems someone just needs to invest money and wire it up without new research needed? Or is it missing some breakthroughs, how many/what are they, how long have these things been pursued, how many people are working on them... what does recent progress look like and so on (in a nice summarized fashion).
Any demo/poc should come up front with an earnest general feasibility assessment. When a breakthrough or two are needed then that should skyrocket. If it's just a lot of expensive engineering then that's also a challenge but tractable.
I've given a lot of scientific tech demonstrations over the years and the businesses behind me obviously want me to be as vague as possible to pull money in. I of course have some of those same incentives (I need to eat and pay my mortgage like everyone else). None-the-less the draw of science to me has always been pulling the veil from deception and mystery and I'm a firm believer in being as upfront as possible. If you don't lead with disclaimers, imaginations run wild into what can be done today. Adding disclaimers helps imaginations run wild about what can be done tomorrow, which I think is great.
I am curious, if the voice, email, chat, and shortly video can all be entirely generated in real or near real time, how can we be sure that remote employee is actually not a full or partially generated entity?
Shared secrets are great when verifying but when the bodies are fully remote - what is the solution?
I am traveling at the moment. How can my family validate that it is ME claiming lost luggage and requesting a Venmo request?
PGP
(I say this in jest, as a PGP user)
If I demoed swype texting as it functions in my day to day life to someone used to a querty keyboard they would never adopt it
The rate at which it makes wrong assumptions about the word, or I have to fix it is probably 10% to 20% of the time
However because it’s so easy to fix this is not an issue and it doesn’t slow me down at all. So within the context of the different types of text Systems out there, I t’s the best thing going for me personally, but it takes some time to learn how to use it.
This is every product.
If you demonstrated to people how something will actually work after 100 hours of habituation and compensation for edge cases, nobody would ever adopt anything.
I’m not sure how to solve this because both are bad.
(Edit: I’m keeping all my typos as meta-comment on this given that I’m posting via swype on my phone :))
Unfortunately iOS text editing is also completely worthless. It forces strange selections and inserts edited text in awkward ways.
I’m a QWERTY texter but text entry on iOS is a complete disaster that has only gotten worse over time.
My only issue is that no keyboard implementation really supports more than two languages which makes me switch back to plain qwerty with autocomplete all the time.
It's even worse on the watch somehow. I take care to hit every key exactly, the correct word is there, I hit space, boom replaced with a completely different word. On the watch it seems to replace almost every word with bullshit, not just technical terms.
Sort of related, it also doesn't let you cuss. It will insist on replacing fuck with pretty much anything else. I had to add fuck to the custom replacement dictionary so it would let me be. What language I choose to use is mine and mine alone, I don't want Nanny to clean it up.
[0] https://f-droid.org/en/packages/inc.flide.vi8/
No, that is how you're told to type. You have to be told to type that way precisely because QWERTY is not designed to keep your fingers on the home row. If you type in a layout that is designed to do that, you don't need to be told to keep your fingers on the home row, because you naturally will.
Nobody really knows what the designers were thinking, which I do not mean as sarcasm, I mean it straight. History lost that information. But whatever they were thinking that is clearly not it because it is plainly obvious just by looking at it how bad it is at that. Nobody trying to design a layout for "keeping your fingers on the home row" would leave hjkl(semicolon) under the resting position of the dominant hand for ~90% of the people.
This, perhaps in one of technical history's great ironies, makes it a fairly good keyboard for swype-like technologies! A keyboard layout like Dvorak that has "aoeui" all right next to each other and "dhtns" on the other would be constantly having trouble figuring out which one you meant between "hat" and "ten" to name just one example. "uio" on qwerty could probably stand a bit more separation, but "a" and "e" are generally far enough apart that at least for me they don't end up confused, and pushing the most common consonants towards the outer part of the keyboard rather than clustering them next to each other in the center (on the home row) helps them be distinguishable too. "fghjkl" is almost a probability dead zone, and the "asd" on the left are generally reasonably distinct even if you kinda miss one of them badly.
I don't know what an optimal swype keyboard would be, and there's probably still a good 10% gain to be made if someone tried to make one, but it wouldn't be enough to justify learning a new layout.
My understanding of QWERTY layout is that it was designed so that characters frequently used in succession should not be able to be typed in rapid succession, so that typewriter hammers had less chance of colliding. Or is this an urban myth?
People don't hunt and peck after years of keyboard use because of the keyboard; they do it because of the keyboard layout.
If you want to prove I'm wrong, go learn Dvorak or Colemak and show me that once you're comfortable you still hunt and peck. You won't be, because it wouldn't even make sense. Or, less effort, find a hunt & peck Dvorak or Colemak user who is definitely at the "comfortable" phase.
I understand how dvorak is designed. I am still not convinced people will be using all their fingers especially their pinkys in a consistent manner that without learning that this is what you should work towards.
I did it. It happened. Theorizing about why it didn't happen is not terribly productive.
The design was to spread out the hammers of the most frequently used letters to reduce the frequency of hammer jamming back when people actually used typewriters and not computers.
The problem it attempted to improve upon, and which is was pretty effective at, is just a problem that no longer exists.
However, I use a Dvorak layout and my hands feel like they alternate better on that due to the vowels being all on one hand. The letters are also in more sensical locations, at least for English writing.
It can get annoying when G and C are next to each other, and M and W, but most of the time I type faster on Dvorak than I ever did on Qwerty. It helps that I learned during a time where I used qwerty at work and Dvorak at home, so the mental switch only takes a few seconds now.
And it does a bad job at it, which is further evidence that it was not the design consideration. People may not have been able to run a quick perl script over a few gigabytes of English text, but they would have gotten much closer if that was the desire. I don't believe that was their goal but they were just too stupid to get it even close to right.
That's a folk myth that's mostly debunked.
https://www.smithsonianmag.com/arts-culture/fact-of-fiction-...
After all, with TVs I've had the same experience as you with the annoying alphabetical keyboard, but we type into they maybe a couple of times a year, or maybe once in 5 years, whereas if we changed our phone keyboard layout we'd likely get used to it quite quickly.
Even if not going so far as to push it as a new default for all users (I'm willing to accept the possibility that I'm speaking for myself as the kind of geeky person who wouldn't mind the initial inconvenience of a new kb layout if it meant saving time in the long run, and that maybe a large majority of people would just hate it too much to be willing to give it a chance), they could at least figure out what the best layout is (maybe this has been studied and decided already, by somebody?) and offer that as an option for us geeks.
This is the sort of demo that 1) gives people a misleading idea of what the product can actually do; and 2) ultimately contributes to the inevitable cynical backlash.
If the product is really great, people can see it in a realistic demo of its capabilities.
I would agree completely that it’s not ready for consumers the way it was displayed, which is my point.
I do want to add that I believe that the right way to do these types of new product rollout is not with these giant public announcements.
In fact, I think generally speaking the “right” way to do something like this demonstrates only things that are possible robustly. However that’s not the market that Google lives in. They’re capitalists trying to make as much money as possible. I’m simply evaluating that what they’re showing I think is absolutely technically possible and I think Google can deliver it even if its not ready today.
Do I think it’s supremely ethical the way that they did it? No I don’t.
And it's dangerous to assume they can just "deliver later". It's not that simple. If it is why not bake it in right now instead of committing fraud?
This is damaging to companies that walk the walk and then people have literally said to me "but what about that Gemini"? and dismiss our work.
That was basically what magic leap did to the whole AR development market. Everyone deep in it knew they couldn’t do it but they messed up so badly that it basically killed the entire industry
But that's a different issue than LLM hallucinations.
With Swype, you already know what the correct output looks like. If the output doesn't match what you wanted, you immediately understand and fix it.
When you ask an LLM a question, you don't necessarily know the right answer. If the output looks confident enough, people take it as the truth. Outside of experimenting and testing, people aren't using LLMs to ask questions for which they already know the correct answer.
It is the main reason that handwriting recognition did not displace keyboards. Once the handwriting is converted to text, it’s easier to fix errors with a pointer and keyboard. So after a few rounds of this most people start thinking: might as well just start with the pointer and keyboard and save some time.
So the question is, how easy is it to detect and correct errors in generative AI output? And the unfortunate answer is that unless you already know the answer you’re asking for, it can be very difficult to pick out the errors.
Yeah the feedback loop with consumers has a higher likelihood of being detrimental, so even if the iteration rate is high, it’s potentially high cost at each step.
I think the current trend is to nerf the models or otherwise put bumpers on them so people can’t hurt themselves. That’s one approach that is brittle at best and someone with more risk tolerance (OpenAI) will exploit that risk gap.
It’s a contradiction then at best and depending on the level of unearned trust from the misleading marketing, will certainly lead to some really odd externalities
Think “man follows google maps directions into pond” but for vastly more things.
I really hated marketing before but yeah this really proves the warning I make in the AI addendum to my scarcity theory (in my bio).
Maybe you can spice up a demo, but misleading to the point of implying things are generated when they're not (like the audio example) is pretty bad.
Except actual good ones, like ChatGPT or Gmail (by their time).
In your Swype analogy, it would be as if Swype works by having to write out on a piece of paper the general goal of what you're trying to convey, then having to write each individual letter on a Post-it, only for you to then organize these Post-its in the correct order yourself.
This process would then be translated into a slick promo video of someone swiping away on their keyboard.
This is not a matter of “eh, it doesn't 100% work as smooth as advertised.”
0: https://techcrunch.com/2023/12/07/googles-best-gemini-demo-w...
"There's no such thing as bad publicity."
Reading the comments of all these disillusioned developers, it’s already damaged them because now smart people will be extra dubious when Google starts making claims.
They just made it harder for themselves to convince developers to even try their APIs, let alone bet on them.
This was stupid.
Ultra is not yet available.
But right now, all the bits needed to do this already exist (just need to be assembled and -to be fair- given a LOT of polish), so it would be somewhat reasonable to think that someone had actually Put In The Work already.
Shame on them :(