I'm sorry but I fail to see the problem with this. DeepMind has made very impressive demos and papers, but they have yet to add one dollar of revenue to Google's bottom line. Further they have drained billions from Google.
Google has to, somehow, get completely out of the research paper game and into the product game.
Papers have to have little/no impact on perf going forward. Other than a small windfall to goodwill they are a misalignment between the company's goals and those of the employees.
Products Google, products. Unless Larry and Sergey want to turn Google into a non-profit research tank. Which would be fine, but likely with substantially lower headcount. Even they aren't that wealthy.
DeepMind has researched and developed features that exist in many Google products today, e.g. Wavenet: https://www.deepmind.com/research/highlighted-research/waven...
Things like the Wavenet "contributions" are just Demis paying lip service to the fact that once in a while Google was nudging them to produce something, anything really that was actually useful.
This was paid by Google for unspecified research services. But the way it’s accounted for it’s likely that it was based on some legitimate contribution. It is unlikely it would be structured this way if it was just corporate support.
DeepMind has public financial filings and you can go read the exact language they use to describe the revenue they generate.
DeepMind is profitable and paying tax on that profit. It’s public information that you can see in its UK regulatory filings.
You could say the same about OpenAI and Microsoft, they drained money for years until about 6 months ago when suddenly the partnership started to pay back big style.
The existing, top-performing product teams at Google should be taking that research and building products around it. If Google has any top-performing product teams left, that is...
This sounds like you need far more evidence. If you say academia as the institution where you share papers, sure but then that’s just a sharing mechanism. Almost like saying all advancements came out of Internet because arxiv is where research is shared.
If you want to say professors and Universities have been heralding AI advancement, that has not been true for at least 10 years possibly more. Moment industry started getting into Academia, Academia couldn’t compete and died out. Even Transformers the founding paper of the modern GPT architectures came out of Google Research. In Vision, ResNet, MaskRCNN to Segment Anything came out of Meta / MIcrosoft. The last great academic invention might have been dropout and even that involved Apple. After that I fail to see Academia coming up with a single invention in ML that the rest of the community instantly adopted because of how good it was.
they've shut down at least one deepmind office recently: https://betakit.com/alphabet-company-deepmind-shutters-edmon...
> Announcing Google DeepMind... launched DeepMind back in 2010 ...
Is it tough because one of these is a newly merged and rebranded team and the other is a machine learning model?
Yes the team which literally created transformer and almost all the important open research including Bert, T5, imagen, RLHF, ViT don't have the ability to execute on AI /s. Tell me one innovation OpenAI bought into the field. They are good at execution but i havent seen anything novel coming out of them.
Microsoft has been doing, like, negative innovation and is still relevant.
What would they have done instead? They still need to fund the operation somehow, and raising money is a good way to build partnerships.
If you were sure it will be a $1T business you have more reason to sell equity and accelerate the growth of your company, because you know the remaining 50% is going to be so valuable.
If you need 10B dollar to develop your product you have to find it from somewhere. Training an LLM is not something you can do in a garage, bootstrapped.
You're reiterating their point. Yeah, Google has competent AI people but that means nothing for their own success if they can't execute. OpenAI has proven that.
Despite what people often write and believe here, the access controls on PII data at Google are incredibly strict. You can't just arbitrarily train on people's personal data. I know, because when I was there, working on search backend data mining, in order to get access to anonymized search and web logs, I had to sign paperwork that essentially said I'd be taken to the cleaners if I abused the access.
> What gives, Google? Get on it
It's a very difficult decision to intentionally destabilize the space you are the leader in, for all the reasons you can imagine. In a sense, Google needed someone else with nothing to lose to shake up the space. How they execute in the new reality is yet to be seen. The biggest challenge they may have right now isn't technological, but that "ChatGPT" has become a sort of brand, like Kleenex and well, Google.
However, whatever's going on inside I still strongly believe in that company! Sometimes though it just feels like they don't themselves.
I'd prioritize their problems like this:
1. LLM's don't have a lucrative business model that Google needs.
2. The quality of their language model is really lacking as of now.
You fix 1 and 2, ChatGPT's branding is nothing. Google is the biggest advertisement machine in the world and they can market the hell out of their product. Just see how Chrome gained ground on Firefox for example.
Google is still used several folds more than ChatGPT and if you resolve 1 and 2, Google will make their money and their users have no incentive to go to ChatGPT.
Many markets had early leaders who got stomped by later entrants.
Microsoft and Google both have the capability and trust to make this available. When corporates start paying for LLMs, per user, or for applications, both Google and Microsoft and the two companies are in the best position.
All other industries will be users paying for LLMS model access.
And yet Google is the largest online advertiser in the world. And yet, GMail used to (I don't know if it still does) push ads into people's inboxes.
I have as much belief in their PII controls as in their "Don't be evil" motto.
https://support.google.com/mail/answer/6603?hl=en
When you open Gmail, you'll see ads that were selected to show you the most useful and relevant ads. The process of selecting and showing personalized ads in Gmail is fully automated. These ads are shown to you based on your online activity while you're signed into Google. We will not scan or read your Gmail messages to show you ads.
...
To opt-out of the use of personal information for personalized Gmail ads, go to the Ads Settings page
--- end quote ---
They literally train their datasets on people's personal data.
OpenAI subverted this by riding on the “open” part of their name at first—before doing a 180-degree turn and selling out to Microsoft.
Receiving traffic to sites is nice, especially for already highly-ranked results, but these are not the people buying the ads.
To be clear, Google does use AI. They use it so heavily that they've designed four generations of training accelerators. All the fancy knowledge graph features used to keep you from clicking anything on the SERP are powered by large language models. The only thing they didn't do is turn Google Search into a chatbot, at least not until Microsoft and OpenAI one-upped them and Google felt competitive pressure to build what they thought was garbage.
And yes, Google's customers share that belief. Remember that when Google Bard gets a fact about exoplanets wrong, it's a scandal. When Bing tries to gaslight its users into thinking that time stopped at the same time GPT-4's training did, it's funny. Bing can afford to make mistakes that Google can't, because nobody uses Bing if they want good search results. They use Bing if they can't be arsed to change the defaults[1].
[0] Or at least they did, then they fired the woman who wrote it
[1] And yes that is why Microsoft really pushes Bing and Edge hard in Windows.
This +100 Somehow there is a perception that chat bots are the only example of AI research or product that matters and all AI organisations ability will be judged by their ability to create chatbots.
Sadly, I think I'd argue that nobody has good search results anymore. Google's results have been SEO'd to the hilt and most of the results are blog spam garbage nowadays.
Ethics is a false excuse because rushing that out show they never cared either. It was just PR and their bluff was called.
Also I skimmed over that Stochastic Paper and I’m unimpressed. I’m unfamiliar with the subject but many points seems unproven/political rather than scientific, with a fixation on training data instead of studying the emerging properties and many opinions notably regarding social activism, but maybe it was already discussed here on HN. Edit: found here: https://news.ycombinator.com/item?id=34382901
You're exactly the kind of person Stochastic Parrots was trying to warn us about - you bought into the AI hype.
AI are extremely sensitive to the initial statistical conditions of their dataset. A good example of this is image regurgitation in diffusion models: if you include the same image n times in the data set, it gets n times the number of training epochs, and is far more likely to be memorized. Stable Diffusion's propensity to draw bad copies of the Getty Images logo is another example; there's so many watermarks and signatures in the training data that learning how to draw them measurably reduces loss. In my own AI training adventures[0], the image generator I trained loves to draw maps all the time, no matter what the prompt is, because Wikimedia Commons hosts an absolutely unconscionable number of them.
Stochastic Parrots is arguing that we can't effectively filter five terabytes[1] of training set text for every statistical bias. Since HN is allergic to social justice language, I'll put it in terms that are more politically correct here: gradient descent is vulnerable to Sybil attacks. Because you can only scrape content written by people who are online, the terminally online will decide what the model thinks, filtered through the underpaid moderators who are censoring your political opinions on TwitBook.
Of course, OpenAI will try anyway[2]. The best they've come up with is to use RLHF to deliberately encode a center-left bias into a language model that otherwise would be about as far-right as your average /pol/ user. This has helped ChatGPT avoid the fate of, say, Microsoft's Tay; but it is just sweeping the problem under the rug.
The other main prong of Stochastic Parrots is energy usage. The reason why OpenAI hasn't been outcompeted by actual open AI models is because it takes shittons of electricity and hardware to train these things. Stable Diffusion and BLOOM are the biggest open competitors to OpenAI, but they're being funded purely through burning venture capital. FOSS is sustainable because software development is cheap enough that people can do it as volunteer work. AI training is almost the opposite: extremely large capital costs that can only be recouped by the worst abuses of proprietary software.
[0] I am specifically trying to build a diffusion model trained purely on public domain images, called PD-Diffusion.
[1] No problem. We are Google. Five terabytes is so little that I've forgotten how to count that low.
[2] When filtering the dataset for DALL-E 2, OpenAI found that removing porn from the training set made the image generator's biases far worse. i.e. if you asked for a stock photo of a CEO, pre-filter DALL-E would give about 60% male, 40% female examples; post-filter DALL-E would only ever draw male CEOs.
No, they turned google search into what it is now.
For me, trying google bard was an instant reminder of the change in behavior in google search from 15 years ago to today.
We used to have a search that you could give obscure flags to Linux commands and find their documentation or source code. Today we have a google search that often only tell you about how some kardashian or recent political drama is a sounds-alike with the technical term that you were searching for.
GPT4 has some of the same "excessively smart" failure modes, but it (and GPT3.5 for that matter) is so much more useful than bard (which hits the user with "I can't do that dave" 100x more often than chatgpt's already excessive behavior) that they're a useful addition to the toolbox. Too bad the toolbox hardly includes plain search anymore.
There were plenty of touch screen phones before the iphone.
You are doing a whole lot of tea leave reading with basically zero visibility, which I can’t really reconcile with how absolute you’re being with your language.
Current-day Google churns out sterile, uninspiring products, and kills them.
If your argument is “this company is going to act out of character and do something innovative!” then…yeah, sure. That’s a good way to be right, sometimes. Just don’t let everyone see the majority of the time where you’ve been wrong.
It's not a matter of skill as much as objective. And DeepMind would still be starting at zero if they decide to pivot to products.
Yes yes that’s right the algorithm is the most important part.
While PaLM demolished benchmarks scores openai with a chat tune of a sizeable but not unwieldy model took the world by storm.
This, but non-sarcastically. Google has spectacularly, so far, failed to execute on products (even of the “selling shovels” kind, much less end-user products) for generative AI, despite both having lots of consumer products to which it is naturally adaptable and a lot of the fundamental research work in generative AI.
The best explanation is that they actually are, institutionally and structurally, bad at execution in this domain, because they have all the pieces and incentives that rule out most of the other potential explanations for that.
> OpenAI bought into the field. They are good at execution but i havent seen anything novel coming out of them.
Right, OpenAI is good at execution (at least, when it comes to selling-shovels tools, I don’t see a lot of evidence beyond that yet), whereas Google is, to all current evidence, not good at execution in this space.
Then somebody else reads the papers, decides to execute on it, and hires all the researchers who are frustrated at discovering all this cool stuff but never seeing it launch.
It seems like that's what they were doing with DeepMind for the last decade. But it's also possible DeepMind as an institution lacked the pressure/product sense/leadership to produce consumable products/services. Maybe their instincts were more centered around R&D and being isolated left them somewhat directionless?
So now that AI suddenly really matters as a business, not just some indefinite future potential, Google wants to bring them inside.
They could have created a 3rd entity, their own version of OpenAI, combining DeepMind with some Google management/teams and other acquisitions and spinning it off semi-independently. But this play basically has to be from Google itself for their own reputation's sake - maybe not for practicality's sake but politically/image-wise.
Being hungry and scrappy seems to be a necessary precondition for bringing innovative products to market. If you don't naturally come from hungry & scrappy conditions (eg. Gates, Zuckerburg, Bezos, PG), being in an environment where you're surrounded by hungry & scrappy people seems to be necessary.
For that matter, a number of extremely well-resourced startups (eg Color, Juicero, WebVan, Secret, Pets.com, Theranos, WeWork) have failed in spectacular ways. Being well-resourced seems to be an anti-success criteria even for independent companies.
I survived ALL the layoffs somehow. Boots on the ground agrees with "doesn't really work all that well" but the people collecting rents keep collecting. Given the size all of these received significant DOJ reviews though the only detail I remember is basketball sized court rooms filled with printed paper for the depositions. I'm sure they burned down the Amazon to print all that legalese, speaking of scaling problems.
i'm thinking:
- Lotus
- Macromedia
- CA
edit: i take it all back! my memory is not as good as i thought it was re: software companies. i will leave up my sorry list as penance for my crappy recent tech history skills.
Indeed, you are right on: Legent, Platinum, CA, and Broadcom in order from little fish to big. CA was the second largest software company in the world behind Microsoft then.
The weird part you couldn't see from this telling is that I worked in the Legent office in Pittsburgh, moved to Boston post-CA acquisition and worked in the CA office in Andover. Resigned and went to Platinum in Burlington. Moved to Seattle. Second CA acquisition in 5 years. I should have quit while I was ahead. Moved back to Pittsburgh. Worked in the exact same office I'd worked in 5 years earlier with the same crew. Weird feeling is a mild understatement. I still know people who work for Broadcom now. I should reach out.
i used to read BYTE mag over in the UK in the early 90s before i moved to USA; CA was such a heavy hitter in the early 90s!! i guess it never really was the same in the post-Wang era(s).
Digital (DEC) had no substantial connection with Western Digital; see https://en.wikipedia.org/wiki/Western_Digital#History
[1] https://www.computerhistory.org/collections/catalog/10275038...
Cisco has done a great job balancing this, actually - they keep contact with engineers who leave to do startups, and then acquire their companies if they become successful enough to prove the product.
Getting a X-million-per-year budget from a parent company gives you a very different sort of situation. IME this results in less urge to get something out the door and more urge to get "the best thing" built. Shipping early risks your budget in a way that "look at all this cool theoretical progress" doesn't, because the public and press can critique you more directly.
It seems like this is more a Google problem than a DeepMind problem though, no? Google created one of the most successful R&D labs for ML/AI research the world has ever known, then failed to have their other business units capitalize on that success. OpenAI observed this gap and swooped in to profit off all of their research outputs (with backing from Microsoft).
IMO what they’re doing here is doubling down on their mistakes: instead of disciplining their other business units for failing to take advantage of this research, they’re forcing their most productive research team to assume responsibility and correct for those failures. I expect this will go about as well as any other instance of subjecting a bunch of research scientists to internal political struggles and market discipline, i.e. very poorly.
Is this really the solution? Is there an example of a company that escaped its fate with this tactic?
I think this is what Christensen and Schumpeter suggest, but I don’t think it works.
Maybe the closest is Microsoft, but they didn’t do this. They changed their revenue model by emulating AWS.
I thought OpenAI’s unique advantage over many big tech companies is that they’ve somehow figured out how to fast track research into product, or have researchers much more willing to worry about “production”.
They're also paying for their product managers' cancellation culture. (Sorry.) I'm seeing a lot of AI pitch decks; none suggest trusting Google. That saps not only network effects, but what ill term earned research: work done by others on your product. Google pays for all its research and promotion. OpenAI does not.
I saw quotes from independent scientists referring to it as the greatest breakthrough of their lifetime, and I saw similarly strong language used in regard to the potential for good of alpha fold as a product.
So they gave it away, but it is still a product they followed through on and continue to.
Was it wrong of them that they gave it away, and right, that Microsoft’s primary intent with their open AI technology, seems to be to provoke an arms race with google?
And if it isn't? Literally every single argument I've seen towards this being AGI is "We don't know at all how intelligence works, so let's say that this is it!!!!!"
> nowhere near the game changer ChatGPT(4) is, even if ChatGPT was only available for the subset of scientists that benefit from Alpha Fold
This is utter nonsense. For anyone who actually knows a field, ChatGPT generates unhelpful, plausible-looking nonsense. Conferences are putting up ChatGPT answers about their fields to laugh at because of how misleadingly wrong they are.
This is absolutely okay, because it can be a useful tool without being the singularity. I'd sure that in a couple of years time, most of what ChatGPT achieves will be in line with most of the tech industry advances in the past decade - pushing the bottom out of the labor market and actively making the lives of the poorest worse in order to line their own pockets.
I really wish people would stop projecting hopes and wishes on top of breathless marketing.
>pushing the bottom out of the labor market and actively making the lives of the poorest worse in order to line their own pockets.
This makes zero sense. GPT4 has little effect on a janitor or truck driver. It doesn't pick fruit, or wash cars.
As far as I know, the exception to this is the bar exam, which GPT-4 can also pass, but that exam plays into GPT-4's strengths much more than other professional exams.
FWIW, this is more true for CA than most states.
Either it's a really obscure usage of the word or I got the president wrong.
Signed,
Guess Who"
That is not an instance of passing a standardized academic test through "autocompletion" or "regurgitation." It's rudimentary synthetic thought.
If it had named a different president, I could have argued with it, which is what I find especially interesting.
Making GPT sit it is like getting someone with no knowledge but a computer full of past questions and answers and a search button to sit the exam. It has metaphorical written it’s answers on it’s arm.
It knows a lot of stuff, but it can't do much thinking, so the minute your problem and its solution are far enough off the well-trodden path, its logic falls apart. Likewise, it's not especially good at math. It's great at understanding your question and replying with a good plain-english answer, but it's not actually thinking
It's able to define new concepts and new words. It's masters have gone to great lengths to prevent it from writing out particular types of judgements (eg https://sharegpt.com/c/uPztFv1). Hell, it's got a great imagination if you look at all the hallucinations it produces.
All of that sum up to many thinking-adjacent things, if not actual thinking! It all really hinges on your definition of thinking.
See GPT4's reply to pclmulqdq at https://news.ycombinator.com/item?id=35648144 .
That's not a response from someone who wrote the answers on the inside of their elbow before coming to class. That's genuine inductive reasoning at a level you wouldn't get from quite a few real, live human students. GPT4 is using its general knowledge to speculate on the answer to a question that has possibly never been asked before, certainly not in those particular words.
(Shrug) Exactly the same as with a human child.
Unlike a human child who tends to know when you are lying to them.
LOL. If that were true, it might have saved Fox News $800 million. Nobody would bother lying, either to children or to adults, if it didn't work as well as it does.
I asked GPT-4 to give me a POSIX compliant C port of dirbuster. It spit one out with instructions for compiling it.
I asked it to make it more aggressive at scanning and it updated it to be multi-threaded.
I asked it for a word list, and it gave me the git command to clone one from GitHub and the command to compile the program and run the output with the word list.
I then told it that the HTTP service I was scanning always returned 200 status=ok instead of a 404 and asked it for a patch file. It generated that and gave me the instructions for applying it to the program.
There was a bug I had to fix: word lists aren’t prefixed with /. Other than that one character fix, GPT-4 wrote a C program that used an open source word list to scan the HTTP service running on the television in my living room for routes, and found the /pong route.
This week it’s written 100% of the API code that takes a CRUD based REST API and maps it to and from SQL queries for me on a cloudflare worker. I give it the method signature and the problem statement, it gives me the code, and I copy and paste.
If you’re laughing this thing off as generating unhelpful nonsense you’re going to get blind sided in the next few years as GPT gets wired into the workflows at every layer of your stack.
> pushing the bottom out of the labor market and actively making the lives of the poorest worse in order to line their own pockets.
I’m in a BNI group and a majority of these blue collar workers have very little to worry about with GPT right now. Until Boston Dynamics gets its stuff together and the robots can do drywalling and plumbing, I’m not sure I agree with your take. This isn’t coming for the “poorest” among us. This is coming for the middle class. From brand consultants and accountants to software engineers and advertisers.
Software engineers with GPT are about to replace software engineers without GPT. Accountants with GPT are about to replace accountants without GPT.
> Literally every single argument I've seen towards this being AGI is
Here is one: it can simultaneously pass the bar exam, port dirbuster to POSIX compliant C, give me a list of competing brands for conducting a market analysis, get into deep philosophical debates, and help me file my taxes.
It can do all of this simultaneously. I can't find a human capable of the simultaneous breadth and depth of intelligence that ChatGPT exhibits. You can find someone in the upper 90th percentile of any profession and show that they can out compete GPT4. But you can't take that same person and ask them to out compete someone in the bottom 50th percentile of 4 other fields with much success.
Artificial = machine, check. Intelligence = exhibits Nth percentile intelligence in a single field, check General = exhibits Nth percentile intelligence in more than one field, check
This is AGI, now we are nit-picking. It's here.
Hahaha, if you want nit-picking, all the language tasks chatGPT is good at are strictly human tasks. Not general tasks. Human tasks are all related to keeping humans alive and making more of us, they don't span the whole spectrum of possible tasks where intelligence could exist.
Of course inside language tasks it is as general as can be, yet still needs to be placed inside a more complex system with tools to improve accuracy, LLM alone is like brain alone - not that great at everything.
This extreme difficulty in discerning what it hallucinates and what is "true" is what it's most obvious problem is. I guess it can be fixed somehow but right now it has to be heavily fact-checked manually.
It does this for computing questions as well, but there is some selection bias so people tend to post the success-stories and not the fails. However it's less dangerous if it's in computing as you'll notice it immediately so maybe require less manual labour to keep it in check.
I have a feeling they had access to a lot of code on GH, who knows how much code they actually accessed. The conspiracy theorist in me wonders if MS just didn’t provide access to public and private code to train on, they wouldn’t have even told Open AI, just said, “here’s some nice data”, it’s all secret and we can’t see the models inputs so I’ll leave it at that.
So yeah I feel your enthusiasm but if you think about it a little more, or maybe not so hard to imagine what you saw being actually rather simple ? Every time I write code I feel kind of depressed because I know almost certainly someone has already written the same thing and that it’s in GitHub or somewhere else. ChatGPT just takes the knowing where to find it and gives it to you directly.
Nothing is magic, right ?
Regardless of the truthiness of your assertion, that general description of work is a majority of my profession. Novel contribution is a very small percentage of the work I do. I'm looking forward to that shifting significantly in the coming years.
It doesn't matter if the "POSIX compliant C" version of dirbuster was in it's training set. Or if my cloudflare wrangler JS API <-> Neon.tech SQL database implementation was. (Copyright arguments withstanding for sure.)
What matters is that it did 100% of the work of generating the code. All I did was guide it, code review it, and ship it. It felt like I had a team of junior engineers working in the trenches on the problem for me and I was getting to have fun playing at a level up the stack. It can generate 100LoC faster than I can type it, and I can code review it as it generates it. I can fix the code up in the editor and paste it back in with the next prompt for another round of work.
This is with the current model.
Unlike GP's assertion that people are falling for marketing, I actually don't know what OpenAI's marketing is. 100% of my belief in GPT-4 comes from it doing my real day-to-day tasks along side me. I don't need a crystal ball to predict what AI is going to do, I'm already using it as a partner in writing my code. It's already well enough adept at my profession to pair-program through problems with me.
This week I had a particularly hairy problem I needed to solve, and I was already behind schedule on the project. ChatGPT and I knocked it out in ~30 minutes. I suspect, if I were alone, it would have taken me several hours. The velocity boost came from: ChatGPT correctly identifying it as a binpacking problem (which I had missed), listing out several algorithms for approaching binpacking, and giving me an initial (incorrect) first implementation. I was able to go back and forth on that response and get the rough constraints figured out for a solution, ask it to generate that solution, and then clean it up in my editor.
The flow I'm using today is already a 100% increase in effective "workforce" for a software organization with a $20 per month per employee subscription to ChatGPT. Everyone gets a real-time, always available, pair programmer.
Okay, now try being a scientist in a scientific field that isn't basic coding.
It's not people laughing at pretences, it's people who know even basic facts about their field literally looking at the output today and finding it deeply, fundamentally incorrect.
I wonder what your personal success rate would be if we did a Turing test with the “people” who “know basic facts about their field.” If they sat at a computer and asked you all these questions, would you get them right? Or would you end up in slide decks being held up as a reason why misnome doesn’t qualify as AGI?
I find comfort in knowing that it can’t “do science.” There is a massive amount of stuff it can do. I’m hopeful there will be stuff left for humans.
Maybe we’ll all be scientists in 10 years and I won’t have to waste my life on all this “basic coding” stuff.
Absolutely not! I created a powershell script for converting one ASM label format to another for retro game development and i used ChatGPT to write it. Now, it fumbled some of the basic program logic, however, it absolutely nailed all of the specific regex and obtuse powershell commands that i needed and that i merely described to it in plain English.
It essentially aced the "hard parts" of the script and i was able to take what it generated and make it fit my needs perfectly with some minor tweaking. The end result was far cleaner and far beyond what i would have been able to write myself, all in a fraction of the time. This ain't no breathless marketing dude: this thing is the real deal.
ChatGPT is an extremely powerful tool and an absolute game changer for development. Just because it is imperfect and needs a bit of hand holding (which it may not soon), do not underestimate it, and do not discount the idea that it may become an absolute industry disrupter in the painfully near future. I'm excited ...and scared
It does, quite often. Not only that, as you describe. But it does.
For example, I asked it what my most cited paper is, and it made up a plausible-sounding but non-existent paper, along with fabricated Google Scholar citation counts. Totally unhelpful.
It also can produce very useful things.
If you're a robotresearcher, maybe try getting it to whip up some ...verilog circuits or something? I don't know much about your field or what you do specifically, but tasks like regular expressions or specific code syntax it is absolutely brilliant at, whatever the equivalent to that is in hardware. ...I've only ever replaced capacitors and wired some guitar pickups.
> it made up a plausible-sounding but non-existent paper, along with fabricated Google Scholar citation counts
I ran into a similar issue: I asked it for codebases of similar romhacks to a project i'm doing, and it provided made up Github repos with completely unrelated authors for romhacks that do actually exist: non-existent hyperlinks and everything.
Now, studying the difference in GPT generations, it seems like more horsepower and more data solves alot of GPT problems and produces emergent capabilities with the same or similar architecture and code. The current data points to this trend continuing. I find it both super exciting and super ...concerning.
You asked a machine learning model to tell you something about yourself that you already knew?
This is not what any of the US economic stats have looked like in the last decade.
Especially since 2019, the poorest Americans are the only people whose incomes have gone up!
I use ChatGPT daily to generate code in multiple languages. Not only does it generate complex code, but it can explain it and improve it when prompted to do so. It's mind blowing.
It isn't and nobody with any experience in the field believes this. This is the Alexa / IBM Watson syndrome all over again, people are obsessed with natural language because it's relatable and it grabs the attention of laypeople.
Protein folding is a major scientific breakthrough with big implications in biology. People pay attention to ChatGPT because it recites the constitution in pirate English.
I use chatGPT every day to solve real problems as if it’s my assistant, and most people with actual intelligence I know do as well. People with “experience in the field”, in my opinion can often get a case of sour grapes that they internalize and project with their seeming expertise and go blind to persist some sense of calm to avoid reality.
For example, it can describe concepts like risk neutral pricing and replication of derivatives but it cannot apply that logic to show how to replicate something non-trivial (i.e., not repeating well published things).
/s
Except its not, because they gave it away without any kind of commercialization. Its possible to give something away for free in some context and still have it be a product (Stable Diffusion is doing quite a bit of that, though its very unclear if they’ll be able to do it sustainably), but AlphaFold doesn’t seem to be an example. It seems to be an example of something cool they did that they had no desire to make into a product. Which is great! But isn’t the same as executing on product in a space.
Moreover, it has been quite some time since Google successfully developed and sustained a high-quality product without ultimately discontinuing it. The organizational structure at Google seems to inadvertently hinder the creation of exceptional products, exemplifying Conway's Law in practice.
Read more about this topic here: https://www.wsj.com/articles/google-ai-chatbot-bard-chatgpt-...
That's a pretty short time ago. So it seems that so far it hasn't really been a failure to execute, but more about problems with product vision or with reading the market right leading to not even attempting to have actual products in this space. That's definitely a problem, but not one that's particularly predictive of how well they'll be able to execute now that they're actually working on products.
The hardware costs alone of running something like GPT 3.5 for real time results is 6-7 figures a year. By the time you scale for user numbers and add redundancy... The infra needs to be doing useful work 24/7 to pay for itself.
It's more than possible Google knows exactly what it can do, but was waiting for it to be financially viable before acting on that. Meanwhile Microsoft has decided to throw money at it like no tomorrow - if they corner the market and it becomes financially viable before they lose that it could pay off. That is a major gamble...
Can you unpack your thinking there? Even at 5% interest for ownership costs to be six figures a year you're talking about millions of dollars in hardware. Inference is just not that expensive, not even with gigantic models.
To the extent that there is operating cost (e.g. energy)-- that isn't generated when the system is offline.
I don't know how big GPT 3.5 is, but I can _train_ LLaMA 65B on hardware at home and it is nowhere near that expensive.
That's 8 $200k GPUs + all the other hardware + power consumption for one instance. You could run it on cheaper hardware, but then you'll get to nowhere near realtime output which is required for the majority of the use cases not already handled well by much smaller models.
Even if Google/Microsoft are getting the hardware at a 50% reduction (bearing in mind these are already not consumer prices) it gets to $1mn in hardware alone - again for a single instance that can handle one user interacting with it at a time.
It makes a lot of the bespoke usecases people are getting excited about (i.e. anything with data privacy concerns) far from financially viable.
If you want a dedicated instance of full capability ChatGPT for example (32K content) OpenAI are charging $468k for a 3 month commitment / $1,584k for a year.
So under the assumption that 8 80GB gpus are required, we're talking about a somewhat more than $100k one time cost (for 8x 80gb A100 plus the host) plus power, not 6-7 figures annually. Huge difference!
Evaluating it in a latency limited regime but without enough workload to enable meaningful batching is truly a worst case. I admit that there are applications where you're stuck with that, but there are plenty that aren't.
Anyone in that regime should try to figure out how to get out of it. E.g. concurrently generating multiple completions can sometimes help you hide latency, at least to the extent that you're regenerating outputs because you were unhappy with the first sample.
> that can handle one user interacting with it at a time.
That bit I don't follow. The argument given there is without batching. You can do N samples concurrently at far less than N times the cost.
> OpenAI are charging
Ah the joys of having a monopoly!
Android is the biggest OS in the world
Chrome is the biggest browser in the world
Gmail is the biggest email service in the world
YouTube is the biggest video platform in the world
Google is the biggest search engine in the world
Google is the biggest digital advertiser in the world
and I'm probably missing more things they're #1 in.
Not bad for a company that has "spectacularly failed to execute on products"
Here’s the whole thing (leaving out a parenthetical that isn’t important here):
“Google has spectacularly, so far, failed to execute on products […] for generative AI”
You listed a bunch of products in other domains, some of which are the reasons why it has institutional incentives not to push generative AI forward, even if it also stands to lose more if someone else wins in it.
Execution is 9/10 of the battle.
I wouldn't bet against Google DeepMind originating the next big thing, at the very least, their odds are higher than OpenAIs.
If Google spends billions of it's ad money doing original research that spawns a new industry with thousands of companies, that would seem to be a great result to me.
https://arxiv.org/pdf/2009.01325.pdf
CLIP also seems novel?
So OpenAI it is.
Google could have built a search engine where paid results were indistinguishable from organic results, but the negative externalities of that were too great.
Google could have remained in China, but the negative externalities of developing and managing a censorship engine were too great.
Google could have productized AI before the risks were controlled, but they sacrificed revenue and first-mover advantage to be more responsible, and to protect their reputation.
This behavior is so rare, it's hard to think of another megacorp that would do that.
Google's far from perfect, they've made ethical lapses, which their competitors love to yell and scream about, but their competitors wouldn't hold up well under the same scrutiny.
Have you not used Google search in the past 5 years?
Before Google, search engines didn't do this. Paid results were indistinguishable from organic results.
Here's an example --> https://imgur.com/a/bSJTBeD
If you have a counter-example, please share!
Prominently? Bold text?
Here's how they repeatedly made ads indistinguishable from search results: https://atechnocratblog.wordpress.com/2016/07/26/color-fade-...
Or this: https://twitter.com/garybernhardt/status/1648496387640938496
To quote from the above, here's what they said in the beginning: "we expect that advertising funded search engines will be inherently biased towards the advertisers and away from the needs of the consumers"
I do agree however that the labeling has gotten less prominent over time. I don't however agree that it has become subtle enough to considered indistinguishable from search results.
This is what it looks like on mobile. A tiny "sponsored" text is the only thing that distinguishes ads from search results: https://imgur.com/a/WOk4NdR
It's deliberately designed this way compared to what it once was https://atechnocratblog.files.wordpress.com/2016/07/history-...
To deny this is quite bizarre
Now, when I search any even slightly remotely commercial search term on mobile, about the entire first page and a half of results are ads. Yes, they're identified with a "Sponsored" message, but as you can see from the "evolution" link the other commenter replied, this was obviously done to make the visual treatment between ads and organic results less clear.
The reason I'm thrilled about Google finally getting competition in their bread-and-butter is not because I want them to fail, but I want them to stop sucking so bad. For about the past 10 or so years Google has gotten so comfy with their monopoly position that the vast majority of their main search updates have been extremely hostile to both end users and their advertisers as Google continually demands more and more of "the Google tax" by pushing organic results down the page.
In the meantime I've switched to Bing, not because I think Microsoft is so much better, because I desperately want multiple search alternatives.
Edit: Great article from a couple years ago about how Google tried to make ads even more indistinguishable from organic results: https://www.theverge.com/tldr/2020/1/23/21078343/google-ad-d...
> ...
> In the meantime I've switched to Bing
Hilarious!
Here's Bing --> https://imgur.com/a/N3HCTtw
Here's Google --> https://imgur.com/a/bSJTBeD
Identical search terms.
Which ads look more like organic results?
Reread the comment you are replying to. It explicitly said that the research is good.
It might help to reflect on what the upsides of this have been for OpenAI, re execution.
On the face of it, execution is often all that matters. FB v myspace, AMD v Intel (eventually), Uber v Lyft, MS v Apple (pre 2001), Apple v MS (post 2001) etc.
Yet Google does not have a slam-dunk product despite so many great research results. This looks a gross failure of the CEO, especially given that he's been chanting AI First in the past few years.
https://www.linkedin.com/in/ashish-vaswani-99892181/
https://www.linkedin.com/in/noam-shazeer-3b27288/
https://www.linkedin.com/in/nikiparmar/
https://www.linkedin.com/in/jakob-uszkoreit-b238b51/
https://www.linkedin.com/in/aidangomez/
https://www.linkedin.com/in/lukaszkaiser/
https://www.linkedin.com/in/illia-polosukhin-77b6538/
Only one remains at Google:
ah yes, stealth startup, my favorite successful product :P
But ya their strong point is execution and doing the hundreds of little things that make the model do well and it turns out that that's more important than "novel" ideas
Google is great at research, one of the best companies in the world. They are also not very good at product. It will not be possible for Google to research their way out of the business problems they’re facing. They may win, but if so it will be because they get good at product, not because the transformers research team comes up with something even more amazing.
How did they drop the ball so hard? OpenAI has been around for less than a decade and as a smaller team with less resources was able to make a better product.
Think about how many decades head start IBM had to perfect search, but search wasn't their core competency.
Delivering advertisements is Google's core competency.
The first casualty was Nest shutting down its APIs, cutting off an ecosystem of third party integrations.
The next casualty was replacing the Nest app with the Google Home app. I stopped following Nest after that because I sold all the Nest stuff I owned and replaced it all with HomeKit.
It's astounding how Google keeps doing this, and its shareholders seem to go along with it. I agree, given their track record, its hard to be optimistic about anything Google slaps their name in front of.
You are comparing an organization to a DNN model? It would be tough for anybody.
I think they are. e.g., https://www.deepmind.com/publications/an-empirical-analysis-...
What a shortsighted statement for a race that has barely gotten out of the gates. But, if any one company should be panicking then it's OpenAI at the thought of losing their minimal lead and getting crushed by the company, that invented most of the technology they use, put a significant amount of resources behind their AI initiatives.
Google Search had an outage yesterday. Google just underwent its first round of layoffs ever which definitely affects internal morale and makes all employees aware of their company's mortality. Google's CEO was in the news last week for hiding communications while under a legal hold. Google stock tanked with the rushed demo of Bard. And, even if all those things weren't true, Google has continually failed to establish revenue streams independent from ads and continually abandons products that don't meet their expectations. Consumer confidence in new Google product announcements is lower than any other major tech company - the default assumption is that the product will be pulled months/years later.
Microsoft is giving their full support to OpenAI through their 49% partnership. $13B investment compared to Google buying DeepMind for $500M and investing $300M in Anthropic. Microsoft has good working agreements with the US government, a long history of unreasonable support for their flagship products, clawed their way back to being one of the most valuable companies in the world by finding diverse revenue streams, and, frankly, comes across as the wise adult in the room given they already had their day in the sun with legal battles.
I agree completely that if there continue to be marked revolutions in AI that invalidate current SOTA then those innovations are likely to arise from Google's research labs, but from an execution standpoint I have nothing but concerns for Google. It's crazy that I feel they need a second chance in the AI revolution when LLMs originated from inside their org just a few years ago. And it's not like they don't feel similarly - there've been countless articles about "Code Red" at Google as they try to rapidly adjust their strategy around AI.
I think OpenAI has a wider leader than people are acknowledging. It's like everyone was forced to show their AI-hand the last couple of months, in an attempt to appease shareholders, and it seemed like a fair fight until GPT4 hit the ground running. Now we're looking at agents and multi-modal support ontop of $200M/yr revenue when everyone else has no business plan and has yet to announce any looming upgrades. At a certain point, first-mover advantage compounds, the foremost AI app store becomes established, and people building commercial products will become entrenched.
One day, with AGI and autonomous agents, the goal will be to merge neural network meshes together in order to gather highly specialized datasets.
Google trying to "win" the super-human AGI race is even more flawed than a nation trying to "win" the nuclear arms race.
At least with a nuclear arms race we all die quickly. Super-human AGI will probably just bring about unthinkable levels of suffering before finally killing us all.
- This does not seem unexpected. Google is panicked about losing the AI race and pushing resources into DeepMind is a logical step to mitigating those fears.
- Google has currently given ~300M to Anthropic and has a partnership with them. I assume Google continues to see potential in both avenues and won't neglect one AI team for the other. I'm guessing that DeepMind will be their primary focus because of the numerous, real-world applications already at play.
- It's tough for me to compare Google DeepMind to OpenAI GPT4. They seem to be very different approaches. Yet, they both have support for language and imagery. So, perhaps they aren't that different afterall?
- Still waiting to hear more from Google on how they plan to leverage their novel PaLM architecture. The API for it was released a month ago, but, to my awareness, has yet to take the world by storm. (Q: Bard isn't powered by PaLM, right?)
Overall, I am not convinced this will be massively beneficial. I don't trust Google's ability to execute at scale in this area. I trust DeepMind's team and I trust Google's research teams, but Google's ability to execute and take products to market has been quite weak thus far. My gut says this action will hamstring DeepMind in bureaucracy.