Live data from Hacker News

ChatGPT is a blurry JPEG of the web

newyorker.com

201–210 of 317 posts

Re: ChatGPT is a blurry JPEG of the web

#201

Earlier quoted context omitted.

It absolutely can generate new data, it does so all the time. If you are claiming otherwise I think we need a more formal definition of what you mean by new data. Are you suggesting because it can't predict the future it can't generate novel data?

It's not just the future, though the examples I gave were future oriented. But it's all very interpolation/summarization-focused. A "song lyrics in the style of Taylor Swift" isn't an actual song by Taylor Swift. A summary of the history of Texas isn't actually vetted by any historian to ensure accuracy. The answer to a math problem may not be correct. To me, those things don't qualify as "new data." They aren't suit…

One way to think about prompting is as a conditional probability distribution. There is a particular song by Taylor Swift or the set of all songs by Taylor Swift but ChatGPT is particularly talented at sampling the "set of all songs in the style of Taylor Swift".

One of the worst problems in the "Expert Systems" age of A.I. was reasoning over uncertainty, for instance this system

https://en.wikipedia.org/wiki/Mycin

had a half-baked approach that worked well enough for a particular range of medical diagnosis. In general it is an awful problem because it involves sampling over a joint probability distribution. If you have 1000 variables you have to sample a 1000-dimensional space, to do it the brute force way you'd have sample the data in an outrageous number of hypercubes.

Insofar as machine learning is successful it is that we have algorithms that take a comparatively sparse sample and make a good guess of what the joint p.d. is. The success of deep learning is particularly miraculous in that respect.

Re: ChatGPT is a blurry JPEG of the web

#202

Earlier quoted context omitted.

its no fad. I use to spend an hour going over an email to management. make it less technical, make it sound nicer / polite, etc. now I can take a sentence and say 'make this more succinct. ok, but take that and make it more polite, great thanks'. I even used it for project proposals. 'write me a 1 page document for this idea'. and then I just use the scaffolding from there. its a huge time saver. also you are notn se…

But would you like to be on the receiving end of this? Would you buy a book written by AI? Content creation might explode, but nobody will read it.

can't be any worse than the crap they are putting out there now. I don't care if its written by a human or not, if its good, fine

Re: ChatGPT is a blurry JPEG of the web

#203

Earlier quoted context omitted.

The thing is that generalization is good enough to make people squee and not notice that the output is wrong but not good enough to get the right answer. If it were going to produce ‘explainable’ correct answers for most of what it does that would be a matter of looking up the original sources to make sure they really say what it thinks they do. I mean, I can say, “there’s this paper that backs up my point” but I hav…

There is definitely a misconception about how to use a tool like ChatGPT. If you give it an analytic prompt like "turn this baseball box score into an entertaining outline" it will reliably act as a translator because all of the facts about the game are contained in the prompt. If you give it a synthetic prompt like "give me quotes from the broadcasters" it will reliably acts as a synthesizer because none of the fact…

That's a touch beyond state of the art but we might get there.

If there was one big problem w/ today's LLMs it is that the attention window is too short to hold a "complete" document. I can put the headline of an HN submission through BERT and expect BERT to capture it but there is (as of yet) no way to cut up a document up into 512 (BERT) or 4096 (ChatGPT) token slices and then mash those embeddings together to make an embedding that can do all the things the model is trained to do on a smaller data set. I'm sure we will see larger models, but it seems a scalable embedding that grows with the input text would be necessary to move to the next level.

Re: ChatGPT is a blurry JPEG of the web

#204
post #14

Ugh I’m beginning to think I’m going to spend the next 6-12 months commenting “no, large language models aren’t supposed to somehow know everything in the world. No, that’s not what they’re designed for. Yes, hooking one up to our long-standing record-of-everything-in-the-world (google’s knowledge graph) is going to be powerful.” It’s getting to point where I need to consider stop going on HN. This is like when my fa…

Powerful for what? To use Chiang's analogy, do you think that an LLM trained on Web content will actually derive the rules of arithmetic, physics, etc. I think it is more likely that in decade or more a majority of Internet content will be generated by machine and search engines will do a great job of indexing increasingly meaningless information.

You’re missing an important point - it’s not /trained/ on live internet content, it /reads/ that content at runtime. I mean it is trained on the internet but please try to separate the concerns. Remember that the goal of this model is language, not learning facts about the world - they could’ve trained it completely on fictional novels if there was a big enough corpus.

The only way that LLM-enhanced search returns misinformation is if the internet is full of misinformation. So yeah we’re still in trouble, but the inclusion of the LLM isn’t going to affect that factor either way IMO

EDIT: this is completely separate from using LLMs to, say, write political statements for Facebook bots and drown out all human conversations. That’s obviously terrifying, but not related to their use in search engines IMO.

Re: ChatGPT is a blurry JPEG of the web

#205

Earlier quoted context omitted.

In fact there's a potent new theory(1) that human consciousness (and probably all mammalian "consciousness") is just a memory system involving some form of lossy compression. Your sense of awareness happens ~20-50 ms after the memory is created. A lot of life is buffering and filtering, and reading that lossy record is very much who we are. Einstein's brain must have been amazing at throwing away information about th…

Why is this comment downvoted?

I dunno. It is pretty official that this what the Hippocampus does. What I do know is that people are pretty incredulous when you feed back conventional wisdom about how the brain works, how SSRIs work, etc.

Re: ChatGPT is a blurry JPEG of the web

#206
post #14

Ugh I’m beginning to think I’m going to spend the next 6-12 months commenting “no, large language models aren’t supposed to somehow know everything in the world. No, that’s not what they’re designed for. Yes, hooking one up to our long-standing record-of-everything-in-the-world (google’s knowledge graph) is going to be powerful.” It’s getting to point where I need to consider stop going on HN. This is like when my fa…

I understand why the mainstream thinks that but it's incredibly annoying that even in tech circles there is very little meaningful discussion, it's mainly just people posting amusing screenshots purportedly showing how smart GPT3 is or in other cases how it's politically biased. Anyone who's played around with it knows that it's fun but it's not a search engine replacement and it doesn't know nor understand things. I…

> It's nothing like described in the article and I don't understand why people who should know better don't call out the bullshit media reporting more.

I'm kind of assuming you didn't read the article, but if you did then I'm kind of assuming that you've never done machine learning, but if you have: how did you manage to do that without ever noticing that you were doing approximation?

Objectively, neural networks are approximators. Like, truly objectively, as in, the literal objective function, objectively minimizes approximation error. We call them objective functions and minimizing the approximation error is typically the objective of these objective functions. This isn't bullshit. It isn't. If you think it is, you are deeply and profoundly mistaken.

The article advances this view of language models. This is a reasonable view of language models for the same reason that machine learning papers exploring neural networks describe them as universal function approximators.

Re: ChatGPT is a blurry JPEG of the web

#207

Earlier quoted context omitted.

its no fad. I use to spend an hour going over an email to management. make it less technical, make it sound nicer / polite, etc. now I can take a sentence and say 'make this more succinct. ok, but take that and make it more polite, great thanks'. I even used it for project proposals. 'write me a 1 page document for this idea'. and then I just use the scaffolding from there. its a huge time saver. also you are notn se…

But would you like to be on the receiving end of this? Would you buy a book written by AI? Content creation might explode, but nobody will read it.

I’m less excited about “books written by AI” than “authors using AI for certain mundane tasks to increase their efficiency”. And multiply that out to games, movies, podcasts, etc etc etc

Re: ChatGPT is a blurry JPEG of the web

#208
An essay making reasonable points, but overall it strikes me like a dismissal circa 1980 of personal computers as toys.

My first day with ChatGPT I tried teaching it my hobby dialect of Lisp (unlikely to be in its training set) and then asking it to implement symbolic differentiation. Its attempt was very scatterbrained, but not completely hopeless. If you don't think that required any thinking from it, I don't want to argue -- unless you're in some position of influence that'd make such an ostrich attitude matter.

Re: ChatGPT is a blurry JPEG of the web

#210

Damn, I hate to plug products on HN, but I'd say that the New Yorker is the one subscription I've loved maintaining throughout my life. First got it right out of college and appreciate it 20 years later. Everyone is publishing think pieces about ChatGPT - yawn. But only the New Yorker said, hmm, how about if we get frickin' Ted Chiang to write a think piece? (It is predictably very well written.)

In his short story Understand, he talks about two superintelligent individuals who are having high bandwidth conversations. Maybe ChatGPT and Bard are those bespoke intelligent agents.

https://web.archive.org/web/20140527121332/http://www.infini...

We continue. We are like two BARDs, each cueing the other to extemporize another stanza, jointly composing an epic poem of knowledge. Within moments we accelerate, talking over each other's words but hearing every nuance, until we are absorbing, concluding, and responding, continuously, simultaneously, synergistically.

Post reply on HN