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

The Sell ∀ ∃ as ∃ ∀ Scam

win-vector.com

Post reply on HN
Print view View original

71–80 of 148 posts

Re: The Sell ∀ ∃ as ∃ ∀ Scam

#71

Post by ummonk »

cjbprime wrote: post #22
dvt wrote: post #9

I can walk and chew bubble gum at the same time: on one hand, yes, there's certainly a lot of Kool-Aid being drank by the AI folks. Even on HN, I constantly argue with people that genuinely think LLMs are some kind of magical black box that contain "knowledge" or "intelligence" or "meaning" when in reality, it's just a very fancy Markov chain. And on the other hand, I think that language interfaces are probably the n…

The Microsoft Research "Sparks of AGI" paper spends 154 pages describing behaviors of GPT-4 that are inconsistent with the understanding of it being a "fancy Markov chain": https://arxiv.org/abs/2303.12712 I expect that the reason people are constantly arguing with you is that your analysis does not explain some easily testable experiences, such as why GPT-4 has the ability to explain what some non-trivial and unique…

The problem is the false dichotomy in thinking that a fancy Markov chain can't contain intelligence / knowledge.

Re: The Sell ∀ ∃ as ∃ ∀ Scam

#72

Post by 317070 »

jameshart wrote: post #61
justeleblanc wrote: post #59

Earlier quoted context omitted.

Pretty sure you should read "for every task there exists a setting".

But what's the inverse?

there exists a setting which will work for all your test points.

The caveat of the author is (I think) that if you have a task, you collect a set of points (questions) on which you will test this task. Then you tune your setting (prompt) to start working for your test point (questions).

After that procedure, you do not know if that prompt solves the original task. You might have overfitted to your test points.

And by repeatedly doing this overfitting for various tasks, you are not gathering evidence that a good setting truly exists for all tasks

Re: The Sell ∀ ∃ as ∃ ∀ Scam

#73

Post by CyberRabbi »

sebzim4500 wrote: post #51
dvt wrote: post #9

I can walk and chew bubble gum at the same time: on one hand, yes, there's certainly a lot of Kool-Aid being drank by the AI folks. Even on HN, I constantly argue with people that genuinely think LLMs are some kind of magical black box that contain "knowledge" or "intelligence" or "meaning" when in reality, it's just a very fancy Markov chain. And on the other hand, I think that language interfaces are probably the n…

>it's just a very fancy Markov chain Could you provide an argument for why an LLM is a fancy markov chain that does not apply equally well to a human?

Good point. While it seems obvious to me that LLMs can never be anything more than fancy Markov chains, in my experience it seems the majority of human "logic" does not operate much differently. Very rare to encounter someone who is able to think or speak critically. Most regurgitate canned responses based on keywords.

Re: The Sell ∀ ∃ as ∃ ∀ Scam

#74

Post by ummonk »

dvt wrote: post #70
sebzim4500 wrote: post #51

Earlier quoted context omitted.

>it's just a very fancy Markov chain Could you provide an argument for why an LLM is a fancy markov chain that does not apply equally well to a human?

Well, for one, humans are obviously at least more than a fancy Markov chain because we have genetically hard-wired instincts, so we are, in some sense, "hard-coded" if you forgive my programming metaphor. Hard-coded to breed, multiply, care for our young, seek shelter, among many other things.

Markov chains, like any algorithm, are hard-coded. And just as evolution hard-codes our genes, supervised learning (and in the future reinforcement learning) hard-codes LLMs and other AI models.

Re: The Sell ∀ ∃ as ∃ ∀ Scam

#75

Post by cpitman »

alextheparrot wrote: post #53
tialaramex wrote: post #34

Earlier quoted context omitted.

Crucially there is not a program for every problem. Many (presumably "Almost all" in a mathematical sense) problems are Undecidable and so a program can't do that.

Why are we not OK with the program producing the undecidability result?

Because it is not generally possible to determine that the result is undecidable, when maybe searching a little longer would find it.

Re: The Sell ∀ ∃ as ∃ ∀ Scam

#76

Post by renewiltord »

Ultimately, this is a tool like Google. If using a search engine doesn't work for you for some task, then it's not the tool for the job.

You even get to trial it for free.

You shouldn't use LLMs if you can't make them work. I can make them work for me for many things. That makes them useful for me.

Re: The Sell ∀ ∃ as ∃ ∀ Scam

#77

Post by ummonk »

version_five wrote: post #8
sebzim4500 wrote: post #5

>To conclude: one must have different standards for developing systems than for testing, deploying, or using systems. Or: testing on your training data is a common way to cheat, but so is training on your test data. Isn't this already a solved problem? Every reasonable paper on ML separates their test data from their validation data already.

That in no way prevents overfitting though hyperparameter optimization / graduate student descent. All the common benchmarks, by definition of being a common benchmark, are susceptible to overfitting

That's why you split it 3-way into train, validation, and test datasets, to ensure you didn't use too many hyperparameters and overfit to the validation data.

Re: The Sell ∀ ∃ as ∃ ∀ Scam

#78

Post by civilized »

analog31 wrote: post #49

>>> Build a system that solves problems, but with an important user-facing control. ... >>> Convince the user that it is their job to find a instantiation or setting of this control to make the system work for their tasks. By golly, you just described playing the cello.

But the cello sounds good.

Re: The Sell ∀ ∃ as ∃ ∀ Scam

#79

Post by ftxbro »

In my opinion it's a weak take that only got so upvoted and commented on hacker news because it has the existential and universal logical quantifier symbols in the title and also because it uses the chiasmus rhetoric device both in the title and at the end of the article.

The argument is that several technologies don't work 'out of the box' and you have to tweak their settings for each problem that you face, and that this means it's a scam. For example you have to change prompts in LLMs or change hyperparameters in other machine learning solutions. This argument is some combination of not insightful and not true.

Post reply on HN