Live data from Hacker News

AI training method exceeds GPT-3 performance with fewer parameters

infoq.com

71–80 of 82 posts

Re: AI training method exceeds GPT-3 performance with fewer parameters

#71
post #54

Earlier quoted context omitted.

Unfortunately, your correction is still misleading because it fails to capture what really sets GPT-3 apart. Ironically, GPT-3's limitation also highlights its strength. As it is not capable of learning (few shot or otherwise) in the strict sense of permanently changing its parameters based on examples, all its demonstrated capabilities are completely at inference time. It somehow configures itself at inference time…

> It somehow configures itself at inference time so that state machines which produce plausible continuations of whatever pattern it was fed, are most probably generated. What is this sentence supposed to convey? I'm an NLP practicioner/researcher and this isn't even true - as GPT isn't a "state machine" as the latent space is continuous and not finite. Moreover, there is nothing that makes GPT-3 "not capable of lear…

> What is this sentence supposed to convey?

There are examples where it is able to recognize and continue patterns in strings which if manually generated, would have required a FSM. In fact, some of the more impressive examples would require a stack of some sort so I thought I was rather underselling its capabilities in that arena.

> as GPT isn't a "state machine" as the latent space is continuous and not finite.

Technically speaking, that is impossible since these models leverage floating point numbers and are limited in memory to whatever hidden and self-attention layers.

Practically speaking, in order to generate strings based on patterns as mentioned prior, there must be abstract states which correspond to states and state changes such that thinking in terms of at least state machines is useful.

Studying LMs in terms of automata is not strange, there have been papers which do this for specific trained RNNs (such as https://arxiv.org/abs/1711.09576). I contend GPT-3 is capable, to a certain extent, of generating these dynamically at inference time.

As far as I know this way of extracting what LMs are doing hasn't been done for Transformers but you can also frame Transformers in term of RNNs so there's no reason why such methods wouldn't readily apply to them too.

> Moreover, there is nothing that makes GPT-3 "not capable of learning."

I specifically addressed that: to count as learning, without diluting the utility of the term, it has to be capable of remembering. Without permanent changes to its weights, the use of the term learning stretches the word beyond utility.

Re: AI training method exceeds GPT-3 performance with fewer parameters

#72
post #71

Earlier quoted context omitted.

> It somehow configures itself at inference time so that state machines which produce plausible continuations of whatever pattern it was fed, are most probably generated. What is this sentence supposed to convey? I'm an NLP practicioner/researcher and this isn't even true - as GPT isn't a "state machine" as the latent space is continuous and not finite. Moreover, there is nothing that makes GPT-3 "not capable of lear…

> What is this sentence supposed to convey? There are examples where it is able to recognize and continue patterns in strings which if manually generated, would have required a FSM. In fact, some of the more impressive examples would require a stack of some sort so I thought I was rather underselling its capabilities in that arena. > as GPT isn't a "state machine" as the latent space is continuous and not finite. Tec…

> Without permanent changes to its weights, the use of the term learning stretches the word beyond utility.

Yes, my claim is that there is nothing that makes its weights incapable of being fine-tuned and thus changed.

Re: AI training method exceeds GPT-3 performance with fewer parameters

#73

Earlier quoted context omitted.

A shorter reply would be: It would be great to compare PET not only to GPT-3, but also to other models, especially ones geared towards few-shot learning. Do you know of any other models that should be used for such a comparison, or are there already any relevant results on SuperGLUE that should be mentioned?

This appears to be SOTA on SuperGLUE with few-shot learning. PET (well, a version called iPET from the same author) is at #9 on the SuperGLUE leaderboard [1], and none of the models above it mention being evaluated by few-shot learning. 1: https://super.gluebenchmark.com/leaderboard/

The results reported there are what most people would call ‘semi-supervised learning’, not ‘few-shot’. The true few-shot results are in a few places in the paper, https://arxiv.org/abs/2009.07118, labeled with ‘- dist’.

Re: AI training method exceeds GPT-3 performance with fewer parameters

#74
post #54

Earlier quoted context omitted.

Unfortunately, your correction is still misleading because it fails to capture what really sets GPT-3 apart. Ironically, GPT-3's limitation also highlights its strength. As it is not capable of learning (few shot or otherwise) in the strict sense of permanently changing its parameters based on examples, all its demonstrated capabilities are completely at inference time. It somehow configures itself at inference time…

Number of parameters remains an important practical concern, and an important goal for research (see [1] for a recent example). It is true that this model requires a large amount of unlabeled data, in addition to a small amount of labeled data, but gathering unlabeled data is often easy. So I think it's potentially misleading to say "this approach remains at a severe disadvantage when amount of effort and time requir…

Let me preface this by saying I generally agree with what you're saying. My focus below is on the delta.

> It is true that this model requires a large amount of unlabeled data, in addition to a small amount of labeled data, but gathering unlabeled data is often easy.

Gathering unlabeled data is easier than labeled but can still be challenging. You'll often require careful thought in assembling a distribution of examples. Being able to skip that step yields a significant savings even if not as much as that gained from going from labeled to largely unlabeled data.

> I'd say this approach actually has a serious advantage compared to GPT3, which is locked inside the walls of OpenAI, can only be used with their permission, and is too big for most people to use (let alone train) anyway.

I fully agree and said as much too.

> The cost and effort to use this approach on large real world problems is probably less than using GPT3.

I'd say it depends. Most of the effort with GPT3 will involve edge cases. Having a system in front to handle these might eat into labor savings but you could still end up net positive. It's difficult to say without real world data, you might be correct.

> That is, it may be a more efficient way to train new models from scratch for tasks for which few labeled examples are available.

You're right in general, I think. But it's still worth pointing out GPT3's advantage. It combines a lot of general capabilities, which together with its generative ability and flexibility to input means the level of expertise required to get something useful will be much lower compared to this semi-supervised learning approach. And there are some capabilities it's displayed, one example of many being discussing, querying, pattern matching on computer code that seem hard to replicate with this method.

Re: AI training method exceeds GPT-3 performance with fewer parameters

#75
post #71

Earlier quoted context omitted.

> What is this sentence supposed to convey? There are examples where it is able to recognize and continue patterns in strings which if manually generated, would have required a FSM. In fact, some of the more impressive examples would require a stack of some sort so I thought I was rather underselling its capabilities in that arena. > as GPT isn't a "state machine" as the latent space is continuous and not finite. Tec…

> Without permanent changes to its weights, the use of the term learning stretches the word beyond utility. Yes, my claim is that there is nothing that makes its weights incapable of being fine-tuned and thus changed.

Oh okay, then in that case we don't disagree and are talking about different things. My issue is the few shot learning done by a model which is performing gradient updates should be distinguished from whatever GPT-3 is doing when it's continuing context patterns.

Re: AI training method exceeds GPT-3 performance with fewer parameters

#76
post #50

Earlier quoted context omitted.

I had this discussion with Timo Schick, the first author, on Reddit. My final comment is copied below, with the relevant context. Note that GPT-3's approach to *GLUE involved no training on the task, just a good choice of prompt, whereas PET and iPET also use fine-tuning. Also, because distillation takes large amounts of training data, they use ensembles in the true few-shot regime, so their parameter efficiency is s…

A shorter reply would be: It would be great to compare PET not only to GPT-3, but also to other models, especially ones geared towards few-shot learning. Do you know of any other models that should be used for such a comparison, or are there already any relevant results on SuperGLUE that should be mentioned?

There are many BERT-based models that would have made for a good numeric comparison, had they tested on few-shot learning, but I'm not aware of any that have.

Re: AI training method exceeds GPT-3 performance with fewer parameters

#77
post #76

Earlier quoted context omitted.

A shorter reply would be: It would be great to compare PET not only to GPT-3, but also to other models, especially ones geared towards few-shot learning. Do you know of any other models that should be used for such a comparison, or are there already any relevant results on SuperGLUE that should be mentioned?

There are many BERT-based models that would have made for a good numeric comparison, had they tested on few-shot learning, but I'm not aware of any that have.

Well, in table 1 they compare to RoBERTa trained in a vanilla supervised fashion?

Re: AI training method exceeds GPT-3 performance with fewer parameters

#78

Earlier quoted context omitted.

http://www.frankmcsherry.org/assets/COST.pdf

It's a bit different, right, because that paper presents better algorithms on the same datasets as the work it cites.

Yes it's different. Another part of Benchmarklandia is where researchers make custom algorithms for the same dataset. Benchmarks targeting MNIST is a great example. :)

Re: AI training method exceeds GPT-3 performance with fewer parameters

#79
post #42

Earlier quoted context omitted.

Yes they tell you how to do it. But training a model of this size requires you to use thousands of GPUs, or wait forever. That will sum up to millions of $$$ in rental and electricity costs.

That seems like the reason why won't they share the trained model. Somebody needs to pay for that, how would they fund it if they just shared it? As far as my thinking goes, they are open more than enough. Thank you for your input!

They claim to be a well funded nonprofit.

Re: AI training method exceeds GPT-3 performance with fewer parameters

#80
post #42

Earlier quoted context omitted.

That seems like the reason why won't they share the trained model. Somebody needs to pay for that, how would they fund it if they just shared it? As far as my thinking goes, they are open more than enough. Thank you for your input!

They claim to be a well funded nonprofit.

Yeah, and that nonprofit probably wouldn't really be doing its mission it got funding for if they were handing out computing, instead of research
Post reply on HN