Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

What is ChatGPT doing and why does it work?

writings.stephenwolfram.com

Post reply on HN
Print view View original

151–160 of 518 posts

Re: What is ChatGPT doing and why does it work?

#151

Post by Upvoter33 »

scottlawson wrote: post #112

I spent about an hour reading this. I know most people probably won't be able to do the same because of time constraints in their day, but seriously this is the most awesome post about large language models I've read. Take a second to think about how much time it must have taken to write this, to prepare all of these examples. This post gave me another way to look at something I already partially understood pieces of…

I had the same reaction. Wow, what a thing to put together. I wonder how many people could produce such a doc, and how long it would take them.

Re: What is ChatGPT doing and why does it work?

#152

Post by Reptur »

criddell wrote: post #149

Off topic: I’m looking for a good web clipper to grab just the article and related images from a page like this and save it (as html or a pdf) for me to read later. I’d like it on iPadOS because that’s where I like to read and write. I tried reader mode, but it lost a lot of the images. Any suggestions? Edit: I was able to get a good PDF using the OneNote web clipper on my desktop.

https://www.google.com/search?q=web+to+pdf

Re: What is ChatGPT doing and why does it work?

#153

Post by janalsncm »

bottlepalm wrote: post #96

The easiest way for ChatGPT to generate good output is to plain understand it. Given the vast amount of input data fed into it, it has no choice, but to start reducing the input into fundamental rules which is basically what understanding is. Understanding is a form of compression. More efficient for a neural network to understand a concept than memorize permutations. Same with statistics and markov chains, people fo…

Understanding is a form of compression, but not all compression is understanding, at least in the way I would describe it. Given an arbitrary function, you could simply compress some number of inputs and outputs, giving a lossy mapping between the two. Maybe you would interpolate between the two closest points in the range for unknown values.

Or, you could store the function itself.

To see the difference, compare for yourself what would happen for very large or very small inputs in both types of “understanding”.

Re: What is ChatGPT doing and why does it work?

#154

Post by williamcotton »

efitz wrote: post #40
mr_toad wrote: post #36

Earlier quoted context omitted.

> Sincere question, because to me it feels like my brain is improvising word by word when typing out this sentence. I often delete and retype it until it feels right, but in the process of typing a single sentence, I'm just chaining words one after another the way they feel right. That’s your conscious experience, but it doesn’t necessarily match what your subconscious mind has actually been doing. I’d hazard a guess…

I don’t generally think in individual words when writing. When I want to write a sentence, the first thing I think about is the idea that I want to convey. Then clusters of words start to come into my mind, as ways of expressing that idea, and I select and organize those fragments into a sentence, sometimes fine tuning words to avoid undesirable connotations or emphasize desired ones. Finally I glue the tuned fragmen…

I definitely just start saying things or hammering out words and then figuring out midway that I need to reword the paragraph at some point or at least choose to sacrifice legibility for my own laziness. It reminds me of improvised music or comedy. You just gotta roll with it and build a structure out of what you’ve already said. Like 60s amphetamine beat poetry epics.

Re: What is ChatGPT doing and why does it work?

#155

Post by yafbum »

codeulike wrote: post #34
spion wrote: post #15

The answer to this is: "we don't really know as its a very complex function automatically discovered by means of slow gradient descent, and we're still finding out" Here are some of the fun things we've found out so far: - GPT style language models try to build a model of the world: https://arxiv.org/abs/2210.13382 - GPT style language models end up internally implementing a mini "neural network training algorithm" (…

That Kenneth Li Othello paper is great. The accompanying blog post https://thegradient.pub/othello/ was discussed on HN here https://news.ycombinator.com/item?id=34474043 A lot of people didn't seem to get it when it was discussed on HN. A GPT had _only_ ever seen Othello transripts like: "E3, D3, C4 ..." and NOTHING else. It knows nothing of the board. It doesnt event know that there are two players. It learned Othe…

This is a potent argument for some interpretations of the "Chinese room experiment"

Re: What is ChatGPT doing and why does it work?

#156

Post by kilgnad »

Rury wrote: post #148
kilgnad wrote: post #135

Earlier quoted context omitted.

I didn't say it's not a probabilistic word generator. I'll quote more of what I wrote: >There are clearly different angles of interpreting what these models are actually doing but people are stubbornly refusing to believe it's anything more then just statistical word jumbles. So CLEARLY because I said there are different angles of interpretation I'm implying that from one of these angles we can interpret it as a stat…

How is this significant? I mean you can view a moving car as stationary, and reality is the thing that moves around it, or you can view reality as stationary, but the car moves through it. Or you can make analogies between electricity and water. (Frankly, I like viewing neural nets as a numerical method for doing nonlinear multivariate regression, and ChatGPT is nothing but a resulting equation(s) of such). In the en…

Well you're just a jumble of electrons and protons interacting with each other. That's a 100% true interpretation is it not?

It's also a view that misses the point that you are a molecular intelligence made of DNA that continually mutates and reconstructs it's physical form with generational copies to increase fitness in an ever changing environment.

But that viewpoint also misses the point that you're a human with wants, needs, desires and capability of understanding the world around you.

All viewpoints are valid. But Depending on the context one viewpoint is more valid then others. For example, in my day to day life do I go around treating everyone as if they're useless jumbles of molecules and atoms? Do I treat them like biological natural selection entities? Or do I treat them like humans?

What I'm complaining about is the fact that a lot of people are taking the most simplest viewpoint when looking at these LLMs. They ARE MORE then statistical word generators and it's OBVIOUS this is the case when you talk to it in-depth. Why are people denying the obvious? Why are people choosing not to look at LLMs for what they are?

Because of fear. Because of bias.

Re: What is ChatGPT doing and why does it work?

#157

Post by Baeocystin »

britzkopf wrote: post #132
codeulike wrote: post #32

Wow this is 19,000 words. I like his summary at the end: At some level it’s a great example of the fundamental scientific fact that large numbers of simple computational elements can do remarkable and unexpected things. And this: ... But it’s amazing how human-like the results are. And as I’ve discussed, this suggests something that’s at least scientifically very important: that human language (and the patterns of th…

The thing I'm sort of confused about, but maybe someone can explain why I shouldn't be, is, why does there seem to be no implication for language translation? Or is there but coverage is overwhelmed by the fascination with chatGPT? In short, is machine language translation now a fully solved problem? A couple years ago when I tested Google translate in a non-esoteric conversation with my Russian speaking girlfriend a…

ChatGPT has been blowing every single translation task I've thrown it out of the water, even compared to other modern systems. I have no idea why more people aren't talking about that aspect of it either, other than the Anglosphere in general is kind of oblivious to things that aren't English.

Re: What is ChatGPT doing and why does it work?

#158

Post by kilgnad »

netsharc wrote: post #143
kilgnad wrote: post #129

Earlier quoted context omitted.

Finally I'm tired of people saying it's just a probabilistic word generator and downplaying everything as if they know. If you said something along these lines before... then these papers show that you're not fully grasping the situation here. There are clearly different angles of interpreting what these models are actually doing but people are stubbornly refusing to believe it's anything more then just statistical w…

> deep understanding of self and awareness Come on... you're making it sound like the thing is sentient. It's impressive but it's still a Chinese Room. Although, for searching factual information it still failed me.. I wanted to find a particular song - maybe from Massive Attack or a similar style - with a phrase in the lyrics, I asked Chatty, and it kept delivering answers where the phrase did not appear in the lyri…

It's clearly early technology, so it's not perfect. But what it is able to get right is clear proof it's more then what you think:

https://www.engraved.blog/building-a-virtual-machine-inside/

Read to the end. The beginning and middle doesn't show off anything too impressive. It's the very end where chatGPT displays a sort of self awareness.

Also here's a scientific paper showing that LLMs are more then a chinese room: https://arxiv.org/abs/2210.13382

Re: What is ChatGPT doing and why does it work?

#159

Post by gtoubassi »

codeulike wrote: post #34
spion wrote: post #15

The answer to this is: "we don't really know as its a very complex function automatically discovered by means of slow gradient descent, and we're still finding out" Here are some of the fun things we've found out so far: - GPT style language models try to build a model of the world: https://arxiv.org/abs/2210.13382 - GPT style language models end up internally implementing a mini "neural network training algorithm" (…

That Kenneth Li Othello paper is great. The accompanying blog post https://thegradient.pub/othello/ was discussed on HN here https://news.ycombinator.com/item?id=34474043 A lot of people didn't seem to get it when it was discussed on HN. A GPT had _only_ ever seen Othello transripts like: "E3, D3, C4 ..." and NOTHING else. It knows nothing of the board. It doesnt event know that there are two players. It learned Othe…

I agree this is an incredibly interesting paper. I am not a practitioner but I interpreted the gradient article differently. They didn’t directly find 64 nodes (activations) that represented the board state as I think you imply. They trained “64 independent two-layer MLP classifiers to classify each of the 64 tiles”. I interpret this to mean all activations are fed into a 2 layer MLP with the goal of predicting a single tile (white, black, empty). Then do that 64 times once for each tile (64 separately trained networks).

As much as I want to be enthusiastic about this, it’s not entirely clear to me that it is surprising that such a feat can be achieved. For example it may be possible to train a 2 layer MLP to predict the state of a tile directly from the inputs. It may be that the most influential activations are closer to the inputs then the outputs, implying that Othello-GPT itself doesn’t have a world model, instead showing that you can predict board colors from the transcript. Again, not a practitioner but once you are indirecting internal state through a 2 layer MLP it gets less obvious to me that the world model is really there. I think it would be more impressive if they were only taking “later” activations (further from the input), and using a linear classifier to ensure the world model isn’t in the tile predictor instead of Othello-GPT. I would appreciate it if somebody could illuminate or set my admittedly naive intuitions straight!

That said, I am reminded of another OpenAI paper [1] from way back in 2017 that blew my mind. Unsupervised “predict the next character” training on 82 million amazon reviews, then use the activations to train a linear classifier to predict sentiment. And it turns out they find a single neuron activation is responsible for the bulk of the sentiment!

[1] https://openai.com/blog/unsupervised-sentiment-neuron/

Re: What is ChatGPT doing and why does it work?

#160

Post by criddell »

eclipxe wrote: post #150
criddell wrote: post #149

Off topic: I’m looking for a good web clipper to grab just the article and related images from a page like this and save it (as html or a pdf) for me to read later. I’d like it on iPadOS because that’s where I like to read and write. I tried reader mode, but it lost a lot of the images. Any suggestions? Edit: I was able to get a good PDF using the OneNote web clipper on my desktop.

Take a screen shot. On iPad/iPhone you can save the screenshot of a webpage as a full PDF of the page. Make sure you are in Safari.

I don’t want the right column or the footer. I only want the main content. The Evernote web clipper used to be pretty good at this but it didn’t work on this article. I tried Pocket and it lost the images for me as well. I just tried saving it to Bear and so far, that looks pretty good.
Post reply on HN