Live data from Hacker News

Many in the AI field think the bigger-is-better approach is running out of road

economist.com

211–220 of 354 posts

Re: Many in the AI field think the bigger-is-better approach is running out of road

#211

Earlier quoted context omitted.

Only you, a human developer can be truly creative. An LLM can only ever reproduce what it has seen before.

Where, exactly, did the LLM see this epilogue to The Great Gatsby before? https://twitter.com/tsimonite/status/1653065940463157248

I responded 'way to not getsby the point'.

How is that in ANY WAY an epilogue to The Great Gatsby? This is exactly the problem. That original story builds with a series of revelations into the conclusion 'so we beat on, boats against the current, borne back ceaselessly into our past': establishing a PURPOSE, perhaps a bleak and unwelcome one. Fitzgerald's revealing an insight into the delusions of humanity. He's picturing even the greatest of us as surfers on the river of nihilism and reality. Our aspirations sparkle prettily… and are gone, like froth in the rapids.

And this is just a little bit beautiful. We can imagine beyond our reach. That's a human thing. The fact that our minds can cling so longingly to something that is simply not real, is kind of wonderful. Everyone's their own little world of unreality, and we aspire so earnestly (much like these AI folks do).

And then the AI, having drunk up all of Fitzgerald and everybody else, 'continues' past the point. With what? Hardly matters. It has no point to make. I'd be impressed if it refused, said 'nope, that was where it ended. Can't add anything worth adding, try reading it again'. But no, because the LLM has no intention and doesn't successfully get one from what it's 'read'.

It's constructed an epilogue out of nebulous religious feelgoodism in the rough style of Fitzgerald's sentence construction, and undermined the whole conclusion of the story… and not maliciously, for that would require intent. Nope… it sort of ambled on, going 'what would feel nice here? ok, now what seems like it would go with this sort of thing? ok, something else, let's have more, what kind of concepts go here? what do people normally say when they talk in this way?'

In so doing, it's less than Gatsby and way less than Fitzgerald. There is nothing here in this 'epilogue'.

GPT4 criticised the 'epilogue', in a rather fascinating way! https://pastebin.com/B0zxbvNv

It successfully works out some of the problems with the first AI's writing, and yet it too fails to get the idea expressed by Fitzgerald, and rather than pivot to religious feelgoodism, it pressures the first AI to instead emphasize how its narrator and Gatsby shared a special bond, the very special specialness of being "the only one who understood what it meant to be young and restless in this restless world."

A lot of people can write that idea, and in fact a lot of people did and that's why GPT4 found it a probable argument to make.

Fitzgerald gave us a moment of viscerally grokking that it doesn't mean s*t… and yet, we will still paddle against the current of time and decay and collapse, because what else can we do? Tomorrow we'll get it. Tomorrow we'll really understand and it'll all make sense.

And so…

Re: Many in the AI field think the bigger-is-better approach is running out of road

#212

Earlier quoted context omitted.

TBH, it looks like metric manipulation to me. They have used GPT-3.5 to generate their data(and not use textbooks at all like the title suggests). And their dataset is very much like their benchmark data. While there was some filtering, but still it is very possible that lot of the benchmark questions were in training data. We likely wouldn't ever know how good the model is as it not only closed but they haven't prov…

They seemed to be pretty mindful of this contamination, and call out that they agressively pruned some training dataset and still observed strong performance. That said, I agree, I really want to try it out myself and see how it feels, and if the scores really translate to day-to-day capabilities. From section 5: In Figure 2.1, we see that training on CodeExercises leads to a substantial boost in the performance of t…

I read that but there is no good technique to rule out close duplicates. I know because I had tried to build one for my product. At best it relies on BLEU, embedding distance and other proxies which are far from ideal.

Re: Many in the AI field think the bigger-is-better approach is running out of road

#213

Isn't the fundamental problem that LLM's don't actually understand anything (as greater concepts), but rather operate as complex probability machines? My 2 month active experience with ChatGPT-4 gave me the following takeaways: - when it's right, it's amazing; and when you, the operator, can recognize the niche use case where it performs really well, it can be a game-changer (although you could have programmed a tool…

Have you been to a bar lately and overheard people talking about politics ? They 100% are probability machine, of lower quality than ChatGPT

Re: Many in the AI field think the bigger-is-better approach is running out of road

#214

Isn't the fundamental problem that LLM's don't actually understand anything (as greater concepts), but rather operate as complex probability machines? My 2 month active experience with ChatGPT-4 gave me the following takeaways: - when it's right, it's amazing; and when you, the operator, can recognize the niche use case where it performs really well, it can be a game-changer (although you could have programmed a tool…

> avoid thinking and effort and instead rely on an external brain

For me personally this was probably the biggest game changer because I'm now able to offload a lot of thinking to GPT-based tools and use my brain cycles for the less automatable activities.

Now instead of searching for something on Google and going through multiple pages before finding an answer, I can ask a precise question and get the answer right away. Especially when I know that the answer _is_ there somewhere, and all I need is to find it. If I'm not happy with the answer, I can continue the conversation until I get what I'm looking for.

I made Bing Chat, ChatGPT, and Warp AI (a feature of the Warp terminal that allows you to access GPT from within the terminal) a part of my daily life and I feel like I'm achieving much more with the time I have.

Re: Many in the AI field think the bigger-is-better approach is running out of road

#215
post #206

Earlier quoted context omitted.

It only matters in matters of free will and ethics. One actual scenario where it's relavant would be the discussion around the criminal justice system. If the universe is deterministic, how can punitive justice be justified?

> If the universe is deterministic, how can punitive justice be justified? Determinism doesn't necessarily mean that organisms always act in the same way. They act in the same way given the exact configuration of them and the world. Obviously, justice changes the configuration of an organism (fines, prison, ...). To me it boils down to the question whether justice decreases the likelihood to commit crimes again. Give…

Our justice systems have evolved over a long time, and thus include many remnants of earlier times when prevailing values were much different than they are today. I'd be wary about giving them the benefit of the doubt.

Re: Many in the AI field think the bigger-is-better approach is running out of road

#216
post #49

Earlier quoted context omitted.

> that don't hallucinate “Hallucination” is part of thought. Solving a new problem requires hallucinating new, non existing, possible outcomes and solutions, to find one that will work. It seems that eliminating the ability to interpolate and extrapolate (hallucinations) would make intelligence impossible. It would eliminate creativity, tying together new concepts, creation, etc. Is the goal AI, or a nice database fr…

>Is the goal AI, or a nice database front end, to reference facts? The latter given the kind of products that are currently being built with it. You don't want your code completion or news aggregator to hallucinate for the same reason you don't want your wrench to hallucinate, it's a tool. And as for hallucinations, that's a PR friendly misnomer for "it made **** up". Using the same phrase doesn't mean it has functio…

You see, there's your problem. Now that you mention it, I absolutely do want my wrench to hallucinate.

A more salient question would be, 'how do I know it ISN'T hallucinating'…

Re: Many in the AI field think the bigger-is-better approach is running out of road

#217

Earlier quoted context omitted.

Let's try to rewrite this in a somewhat more dispassionate style: A pragmatic perspective requires one to accept the present reality as it is, rather than hypothesize an exaggerated potential of what could be. Not all concerns surrounding existential risks in technology are necessarily grounded in empirical evidence. When it comes to artificial intelligence, for instance, current models operate at a speed vastly supe…

Hahaha, thanks ChatGPT! This is better said than my snarky, frustrated at the FUD version, and I can learn from the approach.

No, it's really not, because your riff on 'shoggoths that are both so brilliant as to be dangerous, yet so stupid that they maximize paperclips' touches on an important point that the summarized version completely omits.

AI is exactly that kind of stupid. What it lacks isn't 'brilliance' but intentionality. It can do all sorts of rhetorical party tricks, including those that are good at influencing humans, it can even very likely work out WHICH lines of argument are good at influencing humans from context, and yet it has no intentionality. It's wholly incapable of thinking 'wait, I'm making people turn the world to paperclips. This is stupid'.

So it IS likely to turn its skills to paperclip maximization, or any other hopelessly quixotic and destructive pursuit. It just needs a stupid person to ask it to do that… and we're not short of stupid people.

So what you said was better, snark and all :)

Re: Many in the AI field think the bigger-is-better approach is running out of road

#218

Isn't the fundamental problem that LLM's don't actually understand anything (as greater concepts), but rather operate as complex probability machines? My 2 month active experience with ChatGPT-4 gave me the following takeaways: - when it's right, it's amazing; and when you, the operator, can recognize the niche use case where it performs really well, it can be a game-changer (although you could have programmed a tool…

This probability thing might be a red herring. I reckon it is just that softmax/cross entropy loss is a handy tool to get words out of the NN (and now transformers)

Re: Many in the AI field think the bigger-is-better approach is running out of road

#219
post #111

Isn't the fundamental problem that LLM's don't actually understand anything (as greater concepts), but rather operate as complex probability machines? My 2 month active experience with ChatGPT-4 gave me the following takeaways: - when it's right, it's amazing; and when you, the operator, can recognize the niche use case where it performs really well, it can be a game-changer (although you could have programmed a tool…

I’m not convinced the language part of my brain isn’t just a complex probability machine, just with different trade-offs.

Nope, LLM is and that's all he meant.

Re: Many in the AI field think the bigger-is-better approach is running out of road

#220
post #167

Earlier quoted context omitted.

What nonsense. I've spent over a decade 100% focused on AI, and the broad consensus among everyone I've worked with is not to be that concerned at all. The only consensus is that a small group of self proclaimed experts who make a lot of noise is that they get lots of press coverage if they scream and shout making predictions based on zero scientific evidence. We can understand the physics of greenhouse gases and tak…

> Show me any evidence for AI risk today beyond people's theories and beliefs? Deduction. Empirical evidence isn't the only source of insight. You don't have to conduct experiments in order to reasonably conclude that an entity that 1. outperforms humans at mental tasks 2. shares no evolutionary commonality with humans 3. does not necessarily have any goals that align with those of humans is a potential threat to hum…

The word “entity” is doing some quiet but heavy lifting here. I think it would be a good idea to specify what you really mean by this term, and how we can logically deduce the development of such a thing from existing technology (deep learning).
Post reply on HN