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

LLMs use a surprisingly simple mechanism to retrieve some stored knowledge

news.mit.edu

Post reply on HN
Print view View original

151–156 of 156 posts

Re: LLMs use a surprisingly simple mechanism to retrieve some stored knowledge

#151

Post by mjburgess »

tel wrote: post #149
mjburgess wrote: post #148

Earlier quoted context omitted.

If they only showed the measures, yes. The meaning of the measures is determined by the experiment, not by the data. "Data" is itself meaningless, and statistics on data is only informative of reality because of how the experimenter creates the measurement-target relationship.

Okay I think I buy that. I don’t know if I agree, but trying to argue for a position against it has been sufficiently illuminating that I just need to chew on it more. There’s no doubt in my mind that experimental learning is more efficient. Especially if you can design the experiments against your personal models at the time. At the same time, it’s not clear to me that one could not gain similar value purely by, say…

I'd be interested in the GPT-4 case, if you have a paper (etc.) ?

You are correct to observe that science, as we know it, is ending. We're way along the sigmoid of what can be known, and soon enough, will be drifting back into medieval heuristics ("this weed seems to treat this disease").

This isnt a matter of efficiency, it's a necessity. Reality is under-determined by measurement; to find out what it is like, we have to have many independent measures whose causal relationship to reality is one we can control (through direct, embodied, action).

If we only have observational measures, we're trapped in a madhouse.

Let's not mistake science for pseudoscience, even if the future is largely now, pseudoscientific trash.

Re: LLMs use a surprisingly simple mechanism to retrieve some stored knowledge

#152

Post by importantbrian »

Kerb_ wrote: post #62
gwervc wrote: post #60

Earlier quoted context omitted.

"We" are not forbidding you to open a computer, start experimenting and publishing some new method. If you're so convinced that "we" are stuck in a local maxima, you can do some of the work you are advocating instead of asking other to do it for you.

You can think chemotherapy is a local maxima for cancer treatment and hope medical research seeks out other options without having the resources to do it yourself. Not all of us have access to the tools and resources to start experimenting as casually as we wish we could.

Not to mention not everyone can be devoted to doing cancer research. Some Drs. and Nurses are necessary to you know actually treat the people who have cancer.

Re: LLMs use a surprisingly simple mechanism to retrieve some stored knowledge

#153

Post by tel »

mjburgess wrote: post #151
tel wrote: post #149

Earlier quoted context omitted.

Okay I think I buy that. I don’t know if I agree, but trying to argue for a position against it has been sufficiently illuminating that I just need to chew on it more. There’s no doubt in my mind that experimental learning is more efficient. Especially if you can design the experiments against your personal models at the time. At the same time, it’s not clear to me that one could not gain similar value purely by, say…

I'd be interested in the GPT-4 case, if you have a paper (etc.) ? You are correct to observe that science, as we know it, is ending. We're way along the sigmoid of what can be known, and soon enough, will be drifting back into medieval heuristics ("this weed seems to treat this disease"). This isnt a matter of efficiency, it's a necessity. Reality is under-determined by measurement; to find out what it is like, we ha…

I thought the examples I was thinking of were in the original GPT-4 Technical Report, but all I found on re-reading were examples of it explaining "what's funny about" a given image. Which is still a decent example of this, I think. GPT-4 demonstrates a semantic model about what entails humor.

Re: LLMs use a surprisingly simple mechanism to retrieve some stored knowledge

#154

Post by mjburgess »

tel wrote: post #153
mjburgess wrote: post #151

Earlier quoted context omitted.

I'd be interested in the GPT-4 case, if you have a paper (etc.) ? You are correct to observe that science, as we know it, is ending. We're way along the sigmoid of what can be known, and soon enough, will be drifting back into medieval heuristics ("this weed seems to treat this disease"). This isnt a matter of efficiency, it's a necessity. Reality is under-determined by measurement; to find out what it is like, we ha…

I thought the examples I was thinking of were in the original GPT-4 Technical Report, but all I found on re-reading were examples of it explaining "what's funny about" a given image. Which is still a decent example of this, I think. GPT-4 demonstrates a semantic model about what entails humor.

it entails only that the associative model is coincidentally indinstiguishable from a semantic one in the cases where it's used

it is always trivial to take one of these models and expose it's failure to operate semantically, but these cases are never in the marketing material.

Consider an associative model of addition, all numbers from -1bn to 1bn, broken down into their digits, so that 1bn =

Using such a model you can get the right answers for more additions than just -1bn to 1bn, but you can also easily find cases where the addition would fail.

It's never adding.

Re: LLMs use a surprisingly simple mechanism to retrieve some stored knowledge

#155

Post by tel »

mjburgess wrote: post #154
tel wrote: post #153

Earlier quoted context omitted.

I thought the examples I was thinking of were in the original GPT-4 Technical Report, but all I found on re-reading were examples of it explaining "what's funny about" a given image. Which is still a decent example of this, I think. GPT-4 demonstrates a semantic model about what entails humor.

it entails only that the associative model is coincidentally indinstiguishable from a semantic one in the cases where it's used it is always trivial to take one of these models and expose it's failure to operate semantically, but these cases are never in the marketing material. Consider an associative model of addition, all numbers from -1bn to 1bn, broken down into their digits, so that 1bn = Using such a model you…

I think part of what I suspect is going on here too is more computation and finiteness. It seems correct that LLM architectures cannot perform too much computation (unless you unroll it in the context).

On the other hand you can look at statistical model identification in, say, nonlinear control. This can absolutely lead to unboundedly long predictions.

Re: LLMs use a surprisingly simple mechanism to retrieve some stored knowledge

#156

Post by fspeech »

cypress66 wrote: post #134
fspeech wrote: post #53

Earlier quoted context omitted.

130T unique pages? That seems highly unlikely as that averages to over 10000 pages for each human being alive. If gp merely wants texts of interest to self as opposed to an accurate snapshot it seems LLMs should be quite capable, one day.

Is it? Every user profile in every website is a page. Every single tweet is a page.

Tweets count. HN posts also count (actually as high quality texts :). IOT devices reporting status based on the same templates should not qualify as unique pages (count the number of templates if you want). Now if I do a search for some news, many almost verbatim copies would show up. They should only count as one, as we are looking for unique texts!
Post reply on HN