It's amazing how many researchers underestimate the importance of UX and design.
Personally, seeing non-technical people using prompts for the first time -- and getting results that make sense -- is so incredible. Their eyes light up, they are surprised, and they want to keep playing with it. Amazing!
A sizeable part of the population just got access to an incredibly complex AI model and can play with it. What a collective experience.
Clarke's third law: "Any sufficiently advanced technology is indistinguishable from magic."
Most people don't understand that ChatGPT has no idea what they're talking about. It lacks it's own thought and heuristic patterns, and only gives you the most-likely response to your prompt. People don't know that though, so they think the Mechanical Turk is actually playing chess.
I mostly agree with the headline here. ChatGPT is hardly any more innovative than a Markov chain.
> ChatGPT has no idea what they're talking about. It lacks it's own thought and heuristic patterns, and only gives you the most-likely response to your prompt
Funny thing is, same could be said about lots of people. Try listening to any political debate.
Which leads me to believe that the thing that's missing from AI is the same thing that we miss in those political debates: ability to explain and justify your own thought process.
As long as AI is a blackbox, we won't consider it to be a real intelligence.
If it weren’t that remarkable, how come Siri, alexa, and whatever google calls it’s unused voice assistant are so useless? If it were not a substantial advance, why are chatbots so useless? The reality is, if Alexa had half the ability to “understand” as chatgpt it wouldn’t be nearly as frustrating. It’s not like Amazon, apple, and google haven’t dumped metric tons of investment into these things.
But the essay brings up another connotation: to simulate
is to pretend to be something. A simulator wears many masks.
If you ask GPT to complete a romance novel, it will simulate
a romance author and try to write the text the way they
would. Character.AI lets you simulate people directly,
asking GPT to pretend to be George Washington or Darth Vader.
[…]
This answer is exactly as fake as the last answer where it
said it liked me, or the Darth Vader answer where it says it
wants to destroy me with the power of the Dark Side. It’s
just simulating a fake character who happens to correspond
well to its real identity.
[…]
The whole point of the shoggoth analogy is that GPT is
supposed to be very different from humans. But however
different the details, there are deep structural
similarities. We’re both prediction engines fine-tuned
with RHLF.
And when I start thinking along these lines, I notice that
psychologists since at least Freud, and spiritual traditions
since at least the Buddha, have accused us of simulating a
character. Some people call it the ego. Other people call
it the self.
Innovation is not invention. Innovation includes combining pedestrian concepts in novel ways. ChatGPT has the best ux and richest corpus of any markov chain I've ever used. It fits my bill of innovation combining sevaral things into a dang appealing product.
I don't think he's underestimating the UX/design piece, I think it's just outside of the scope of what he's talking about. It seems pretty clear to me that his statement is talking about the underlying AI technology (which makes sense given his background).
And in any case, I wouldn't call anything about the UX or design here innovative - it's obviously a huge improvement in terms of usability, but a chat UI is a pretty well-established thing.
I think the idea of a "prompt" is actually pretty cool. I never saw that framing prior to GPT-3 and I think it reframes the entire idea behind what a model does and how you interact with it.
Researchers point of view is based on their area of research and that's fair and expected.
Yann LeCun compares ChatGPT in the context of the related research.
Imagine a ChatGPT equivalent that memorizes many questions and does a brute force strategy for an answer. It may "look" magic, but there's nothing magic about it. We all accepted that this is the case with Blue Gene - https://en.wikipedia.org/wiki/Deep_Blue_(chess_computer)
What's different here?
Productization and usability are difference concerns here and Yann LeCun is not a usability researcher. Granted, that doesn't mean usability/accessibility doesn't impact research outcomes.
OpenAI's really interesting approach to GPT was to scale the size of the underlying neural network. They noticed that the performance of an LLM kept improving as the size of the network grew so they said, "Screw it, how about if we make it have 100+ billion parameters?"
Turns out they were right.
From a research perspective, I'd say this was a big risk and it turned out they were right -- bigger network = better performance.
Sure, it's not as beautiful as inventing a fundamentally new algorithm or approach to deep learning but it worked. Credit where it's due -- scaling training infrastructure + building a model that big was hard...
It's like saying SLS or Falcon Heavy are "just" bigger rockets. Sure, but that's still super hard, risky, and fundamentally new.
But ChatGPT does not use brute force search to look for an answer. It interpolates among the answers in its training set. I.e. in Yann LeCun's analogy of a cake, interpolation or unsupervised learning is the cake; direct feedback on each single data point or supervised learning is the icing on the cake; and general feedback of the "how I am doing" sort or reinforcement learning is the cherry on top. Now LeCun is just saying that the cake is a lie, and leaving it at that. I don't think this is a helpful understanding.
It's quite fascinating how information retrieval and search engines have evolved..
From trying to teach people how to google via "I am feeling lucky", to using language models for ranking, to building LLMs to better adapt to user queries and move beyond keywords, to having arguably useful chatbots that can synthesize responses.
I am curious to see what the future holds. Maybe an AI that can anticipate our queries?
> It's amazing how many researchers underestimate the importance of UX and design.
Yes. Also the fact that ChatGPT's UX and design leaves much to be desired. They could add/improve the product in so many obvious ways. I wonder if they either 1.) don't care, or 2.) have plans to, but are holding off until they take on a subscription fee.
>> using prompts for the first time -- and getting results that make sense -- is so incredible.
Years ago it was possible to insert prompts into Google and get results that made sense, results that were meaningfully tied to the information you requested. The young people today who missed that time think it magic.
I think this is fundamentally different because you can play with it, and people want to play with it.
Google isn't a playground -- it's much more utilitarian. You go there when you need to find a doctor in Seattle, or to research politics in Poland, or something. You get results which are great, but you don't need to stick around.
GPT-3 and ChatGPT allow you to spend time playing with the system and having fun with it[1]. I think this is what makes it so viral and interesting to people.
chatGPT is innovative in the way the iPhone was innovative. That, of course can be market winning, but it's dangerous, especially for businesses and investors, to get caught up in hype that there is some new "AI" out there that can do things we previously couldn't
Can you point to any product where I could do what chatgpt does before chatgpt? Feels like we just leapfrogged from dumb markov chains to something incredible that provides real value.
This is very much people seeing the Apple Lisa(and/or Mac) for the first time. It had all been done better in the labs at Xerox over a decade earlier. No one got to use it though.
Show me the AI chat system from a decade ago that could hallucinate a git repo.
Edit: let me be more careful:
Show me the AI chat system from a decade ago that wasn't specifically designed to hallucinate git repos, but would hallucinate a git repo in response a detailed natural language prompt. It should have been good enough in its output to cause the user to write more prompts in an attempt to gauge whether the AI system is querying the internet in realtime to generate its response.
Decade ago, definitely not. But the answer to that question, time aside, is, well, GPT-3. No doubt ChatGPT is better, but it does seem like a decent bit of the improvement was UX related, since with the right prompts people have shown similarly miraculous output from GPT-3.
Of course, I'm pretty sure the quoted person is referring to the fact that Google and others likely have similarly capable LLMs already that they simply won't release the models for. I can see the argument that new architectures and designs can be innovative, but that simply being the one to throw a million dollars of compute time into training is much less so, if that's the kind of argument.
According to the article: "LeCun made his remarks about OpenAI in response to a question during the colloquium posed by New York Times journalist Cade Metz. Metz asked if Meta's AI team, FAIR, which LeCun built, will ever be identified in the public mind with breakthroughs the way that OpenAI is."
It seems to me "well actually, ChatGPT isn't really the breakthrough you think it is" is an appropriate response. He wasn't dismissive of ChatGPT and even praised it to some degree.
Well, the transformer was invented at Google, language models were decades old. But scaling it was not done before, and preparing the dataset at this size, and babysitting the model so its loss doesn't explode during training - all were innovations that required lots of work and money to be achieved, so we can just copy the same formula without redoing all the steps.
I think if you read the headline it seems childish. His main point is that Google/Meta couldn't really release this, since the accuracy is really really poor.
Google is the new Xerox - tons of cool tech that nobody gets to see besides white papers and articles. And the only real product that makes money is the search engine that may get disrupted very soon by newcomers using tech based on Google's own research.
He's not wrong. The key ingredient in ChatGPT is not brilliance but capital. It takes a lot of money not just to do the raw training but to get all the data together and process everything.
There's no break through insights that make ChatGPT work, just a lot of consolidated wealth.
The hacker part of me finds AI less and less interesting for this reason. We're seeing what the limits of pouring resources into the problem are. The results are cool, but I think we'll very soon see progress bound since we're using most of the data we can and it doesn't look like adding even more parameters to these models might not yield that much of result.
But this shift in AI means it's increasing something programmers consume rather than produce.
As far as interesting AI goes, I found the series of posts on stuff people were doing with Prolog that showed up during the last month much more interesting.
When language models run out of more trillions of words to train, there is one way ahead - we need to generate more. But wait, you might say, garbage in garbage out. It won't work.
Normally it wouldn't, but we add an extra ingredient here. We get a validation signal. This is problem specific, but for code it would mean to integrate the LM with a compiler and runtime so it can iterate until it solves the task, step by step. For other tasks it could mean hooking the AI to simulators, games and robots to solve tasks. It is also possible to use LLMs as simulators of text.
Basically doing Reinforcement Learning with a Language Model and not just for human preferences, but for problem solving on a grand scale. Collect data from problem solving, train on it, and iterate. It costs just electricity, but LLMs can make their own data. Anthropic's Constitutional AI which is RLAIF - reinforcement learning from AI feedback is proof it can be done.
There are a lot of things that require an amount of processing power that the average person doesn't yet possess. This doesn't mean that they aren't interesting.
50 years ago you would have said "these computer things aren't interesting, only big institutions and corporations have them".
I find it the opposite. There were a lot of projects I didn't have the time to work on and got bored to fast learning. So, now I can tell chatgpt to build it for me. and I only have to debug the 1 or 2 things it can't do. I built a blogging platform, a voice based chatbot using gpt3, something that takes blog article and trims out all the garbage html so only the basic html is left. created a python api, and called it from shared hosting using php. something that trims the times out of youtube captions so I can post it into chatgpt to give me an outline. I just tell it what I want the function to do and it does it.
It's helped me way up my own personal productivity and creativity.
I think what’s really being emphasized here is that ChatGPT is an first generation product without much of a novel technology moat compared to what others with capital can deliver.
Microsoft will soak this work up as a pillar of their own AI assistance tech, but the other big tech firms (and probably some startups) are positioned to try to leapfrog what we see here, now that the market has been demonstrated.
In the language of tech history, ChatGPT is the Lycos or Altavista to some not-yet-launched Google.
Sure, the ideas could be old (in ml research timeframes) but it works and they shipped it. It’s most people’s first exposure to a lot of innovative ideas that make up LLM at scale.
I agree with his assertion that the only reason other people haven’t seen this out of Meta or Google is that they “have a lot to lose” from putting out bots “that make stuff up” - that’s the root of Google
seemingly losing the narrative war I think we’re seeing here. Not sure how to get them to work through that fear.
>Google seemingly losing the narrative war I think we’re seeing here.
Google doesn't care about the narrative war. They care about getting sued by the EU.
They can put out a similar model whenever they want. The just don't want to, because the thing about AI research (vs., say, social media) is there's no first mover advantage. The best model wins and there's no network effect to bolster incumbents.
> The just don't want to, because the thing about AI research (vs., say, social media) is there's no first mover advantage.
That’s true and I am not worried for Google in the slightest. However, once a tool is good enough, both the branding and use of it becomes somewhat sticky. It’s why people still call any tablet an iPad, and it’s one of the reasons even if a large search engine launched tomorrow people would keep using google for a long time. The problem isn’t being first, it’s if they crack changing consumer habits while google is biding their time.
> they “have a lot to lose” from putting out bots “that make stuff up”
2023 should be the year of AI validation. For starters, we could extract facts from all the training examples (trillions) and organise them in a knowledge base. Where there is inconsistency or variation, the model should learn the distribution, so later it can confidently say a fact is not in its training data or is controversial. Note that I didn't say the model should find the truth, just the distribution of "facts".
We can add source reputation signals to this KB to improve its alignment with truth. And if it is controversies we want to know, we can look at news headlines, they usually reference facts being debated. So we can know a fact is contested and by who.
Another interesting approach - probing the model for "truth" by identifying a direction in the latent space that aligns with it. The logic being, even when the model is deceptive, it has an interest to know it is not telling the truth to keep the narrative consistent. So we need to just identify this signal. This is also part of work for AI alignment.
I hope this year we will see large investments in validation, because the unverified outputs of a generative model are worthless and they know it. At the very least hook the model up with a calculator and a way to query the web.
"In terms of underlying techniques, ChatGPT is not particularly innovative," said Yann LeCun, Meta's chief AI scientist, in a small gathering of press and executives on Zoom last week.
"It's nothing revolutionary, although that's the way it's perceived in the public," said LeCun. "It's just that, you know, it's well put together, it's nicely done."
It seems he s trying to say something that is technically correct, but exactly at the wrong time. It s going to be a lot of backlash if you go against the hype wave that we are riding. Better to explain in a more nuanced way how chatGPT managed to become so popular despite being an incremental improvement of fundamental findings of the past few years.
People love it because it understands language, every time and without fail. So what if it spits out lies, this thing makes people happy, which is the value at play here
Not the parent, but to me it generally means 1 million (likely non-dev, regular people) users in 5 days.
For me personally: I have had success asking for code snippets and general brainstorming in areas which are not my forte, all the while using a very nice and clean UI.
Readit News
Personally, seeing non-technical people using prompts for the first time -- and getting results that make sense -- is so incredible. Their eyes light up, they are surprised, and they want to keep playing with it. Amazing!
A sizeable part of the population just got access to an incredibly complex AI model and can play with it. What a collective experience.
Most people don't understand that ChatGPT has no idea what they're talking about. It lacks it's own thought and heuristic patterns, and only gives you the most-likely response to your prompt. People don't know that though, so they think the Mechanical Turk is actually playing chess.
I mostly agree with the headline here. ChatGPT is hardly any more innovative than a Markov chain.
Funny thing is, same could be said about lots of people. Try listening to any political debate.
Which leads me to believe that the thing that's missing from AI is the same thing that we miss in those political debates: ability to explain and justify your own thought process.
As long as AI is a blackbox, we won't consider it to be a real intelligence.
Y’all simply have lost the ability to be amazed.
That "only" carries a lot of weight on its back.
I wonder why people keep saying this. Is it some type of psychological defense to future AI displacement?
I’m not sure these are well-defined enough terms to make this claim.
Innovation is not invention. Innovation includes combining pedestrian concepts in novel ways. ChatGPT has the best ux and richest corpus of any markov chain I've ever used. It fits my bill of innovation combining sevaral things into a dang appealing product.
A smartphone is hardly any more innovative than the ENIAC. And yet ...
Do not underestimate the power of making tech useful and accessible. The lightbulb means nothing without the power grid after all.
And in any case, I wouldn't call anything about the UX or design here innovative - it's obviously a huge improvement in terms of usability, but a chat UI is a pretty well-established thing.
Yann LeCun compares ChatGPT in the context of the related research. Imagine a ChatGPT equivalent that memorizes many questions and does a brute force strategy for an answer. It may "look" magic, but there's nothing magic about it. We all accepted that this is the case with Blue Gene - https://en.wikipedia.org/wiki/Deep_Blue_(chess_computer)
What's different here?
Productization and usability are difference concerns here and Yann LeCun is not a usability researcher. Granted, that doesn't mean usability/accessibility doesn't impact research outcomes.
OpenAI's really interesting approach to GPT was to scale the size of the underlying neural network. They noticed that the performance of an LLM kept improving as the size of the network grew so they said, "Screw it, how about if we make it have 100+ billion parameters?"
Turns out they were right.
From a research perspective, I'd say this was a big risk and it turned out they were right -- bigger network = better performance.
Sure, it's not as beautiful as inventing a fundamentally new algorithm or approach to deep learning but it worked. Credit where it's due -- scaling training infrastructure + building a model that big was hard...
It's like saying SLS or Falcon Heavy are "just" bigger rockets. Sure, but that's still super hard, risky, and fundamentally new.
From trying to teach people how to google via "I am feeling lucky", to using language models for ranking, to building LLMs to better adapt to user queries and move beyond keywords, to having arguably useful chatbots that can synthesize responses.
I am curious to see what the future holds. Maybe an AI that can anticipate our queries?
Yes. Also the fact that ChatGPT's UX and design leaves much to be desired. They could add/improve the product in so many obvious ways. I wonder if they either 1.) don't care, or 2.) have plans to, but are holding off until they take on a subscription fee.
The # of times I've tried to clone a GitHub repo and run a model on my own only to get 50 errors I can't debug.... :)
Deleted Comment
Someone on twitter likened this to when Xerox invented the mouse, but Apple/Microsoft shipped it with their PCs
Years ago it was possible to insert prompts into Google and get results that made sense, results that were meaningfully tied to the information you requested. The young people today who missed that time think it magic.
Google isn't a playground -- it's much more utilitarian. You go there when you need to find a doctor in Seattle, or to research politics in Poland, or something. You get results which are great, but you don't need to stick around.
GPT-3 and ChatGPT allow you to spend time playing with the system and having fun with it[1]. I think this is what makes it so viral and interesting to people.
[1] https://10millionsteps.com/ux-ai-gpt-3-click
Sometimes "making the first X that doesn't suck" is a lot more important than "making the first X".
Edit: let me be more careful:
Show me the AI chat system from a decade ago that wasn't specifically designed to hallucinate git repos, but would hallucinate a git repo in response a detailed natural language prompt. It should have been good enough in its output to cause the user to write more prompts in an attempt to gauge whether the AI system is querying the internet in realtime to generate its response.
Of course, I'm pretty sure the quoted person is referring to the fact that Google and others likely have similarly capable LLMs already that they simply won't release the models for. I can see the argument that new architectures and designs can be innovative, but that simply being the one to throw a million dollars of compute time into training is much less so, if that's the kind of argument.
It seems to me "well actually, ChatGPT isn't really the breakthrough you think it is" is an appropriate response. He wasn't dismissive of ChatGPT and even praised it to some degree.
https://www.quora.com/How-did-the-original-Macintosh-compare...
There's no break through insights that make ChatGPT work, just a lot of consolidated wealth.
The hacker part of me finds AI less and less interesting for this reason. We're seeing what the limits of pouring resources into the problem are. The results are cool, but I think we'll very soon see progress bound since we're using most of the data we can and it doesn't look like adding even more parameters to these models might not yield that much of result.
But this shift in AI means it's increasing something programmers consume rather than produce.
As far as interesting AI goes, I found the series of posts on stuff people were doing with Prolog that showed up during the last month much more interesting.
Normally it wouldn't, but we add an extra ingredient here. We get a validation signal. This is problem specific, but for code it would mean to integrate the LM with a compiler and runtime so it can iterate until it solves the task, step by step. For other tasks it could mean hooking the AI to simulators, games and robots to solve tasks. It is also possible to use LLMs as simulators of text.
Basically doing Reinforcement Learning with a Language Model and not just for human preferences, but for problem solving on a grand scale. Collect data from problem solving, train on it, and iterate. It costs just electricity, but LLMs can make their own data. Anthropic's Constitutional AI which is RLAIF - reinforcement learning from AI feedback is proof it can be done.
50 years ago you would have said "these computer things aren't interesting, only big institutions and corporations have them".
It's helped me way up my own personal productivity and creativity.
Microsoft will soak this work up as a pillar of their own AI assistance tech, but the other big tech firms (and probably some startups) are positioned to try to leapfrog what we see here, now that the market has been demonstrated.
In the language of tech history, ChatGPT is the Lycos or Altavista to some not-yet-launched Google.
I agree with his assertion that the only reason other people haven’t seen this out of Meta or Google is that they “have a lot to lose” from putting out bots “that make stuff up” - that’s the root of Google seemingly losing the narrative war I think we’re seeing here. Not sure how to get them to work through that fear.
Google doesn't care about the narrative war. They care about getting sued by the EU.
They can put out a similar model whenever they want. The just don't want to, because the thing about AI research (vs., say, social media) is there's no first mover advantage. The best model wins and there's no network effect to bolster incumbents.
That’s true and I am not worried for Google in the slightest. However, once a tool is good enough, both the branding and use of it becomes somewhat sticky. It’s why people still call any tablet an iPad, and it’s one of the reasons even if a large search engine launched tomorrow people would keep using google for a long time. The problem isn’t being first, it’s if they crack changing consumer habits while google is biding their time.
2023 should be the year of AI validation. For starters, we could extract facts from all the training examples (trillions) and organise them in a knowledge base. Where there is inconsistency or variation, the model should learn the distribution, so later it can confidently say a fact is not in its training data or is controversial. Note that I didn't say the model should find the truth, just the distribution of "facts".
We can add source reputation signals to this KB to improve its alignment with truth. And if it is controversies we want to know, we can look at news headlines, they usually reference facts being debated. So we can know a fact is contested and by who.
Another interesting approach - probing the model for "truth" by identifying a direction in the latent space that aligns with it. The logic being, even when the model is deceptive, it has an interest to know it is not telling the truth to keep the narrative consistent. So we need to just identify this signal. This is also part of work for AI alignment.
I hope this year we will see large investments in validation, because the unverified outputs of a generative model are worthless and they know it. At the very least hook the model up with a calculator and a way to query the web.
"In terms of underlying techniques, ChatGPT is not particularly innovative," said Yann LeCun, Meta's chief AI scientist, in a small gathering of press and executives on Zoom last week.
"It's nothing revolutionary, although that's the way it's perceived in the public," said LeCun. "It's just that, you know, it's well put together, it's nicely done."
It's not free from error but what it's able to put together is pretty great...
Deleted Comment
People love it because it understands language, every time and without fail. So what if it spits out lies, this thing makes people happy, which is the value at play here
Deleted Comment
For me personally: I have had success asking for code snippets and general brainstorming in areas which are not my forte, all the while using a very nice and clean UI.