[xbmcuser]

This is the main thing that I keep harping about that human knowledge is too vast today for a person or even a group of people and llm will change that many discoveries that require serendipity in the past will be more likely than ever

[yourapostasy]

I want to hear James Burke of Connections [1] and his writer team to have a free wheeling discussion for a few hours on what they see will happen with LLM’s making these connections with more conscious intent a lot easier. The awesome compression of knowledge aspect of LLM’s is a far undersold aspect of the technology.

[1] https://en.wikipedia.org/wiki/Connections_(British_TV_series...

[Nashooo]

I don't think you can look at the current AI landscape and say any aspect of it is undersold.

[serf]

I totally disagree. AI is overhyped, but the demos that have impressed me most are things/elements that very few people are touching and that no one seems to be talking about.

the hype is where the money is, as is always -- marketing & porn. Both touched heavily by AI already.

[pojzon]

There is one issue with this. When noone can prove or disprove what AI came up with.

Currently we can live with it because someone can review that work. Soon we wont be able.

[aspenmartin]

We can use things like Lean for math proofs, verification is tractable.

[pluc]

Even a broken clock is right twice a day.

[anthonyrstevens]

I'm not sure what your point is - can you expand?

[runarberg]

I am also an AI skeptic, but I would rather have used the 1000 monkeys with a 1000 typewriters will eventually write the whole works of Shakespeare analogy.

When you consider the amount of computation which went into this discovery it is less impressive. Like if you spend a lot of fuel you can travel really fast, much faster then a bicyclist. Similarly Go-engines can beat the best humans at go, but they spend several orders of magnitude more energy to do so.

Mathematicians prove or disprove conjectures all the time and use orders of less energy to do so. Using LLMs is kind of just throwing money at the problem and hoping it works. In this case it did. But this is not the most efficient way to do this, and it won‘t scale.

[xylenox]

> Go-engines can beat the best humans at go, but they spend several orders of magnitude more energy to do so.

From what I can find AlphaGo Zero runs at ~400 watts, while a human brain uses ~20. So really only about 1 order of magnitude difference.

Of course, training costs are a different question entirely.

[pluc]

People are marveling at what AI can discover purely out of time and chance. AI will undoubtedly find awesome things because there's very few things we've thrown this much money at. For every awesome thing AI finds, there's a million mistakes, fake leads, hallucinations, etc. Amaze away but let's not forget this is an exception much more than the norm.

[throwaway27448]

Presumably with automated proofs, hallucinations (aka, creativity or confabulation, if we hadn't botched the naming) is a good thing. It's in the empirical world where trusting an LLM is a stupid thing because there's no automated form of fact checking.

[bnchrch]

I'm sorry. But whats the point of this critique?

That a raw LLM hallucinates?

That we never see all the mistakes and dead ends a complex system using AI hits?

Does it even matter if its accuracy rate across all its experiments is < 100% if it can run trillions of experiments in the same time a human could run 1?

We don't see many of the failed attempts of Human Researchers. Why? Because it doesn't matter.

What amazing here is that it shows our society can make discoveries faster in the post LLM world. Thats incredible.

Your "critique" of how it happened. Not so much.

[switchbak]

As a human, I have many stupid and wrong ideas all day long - most of those don't bubble up to my conscious awareness. If LLMs hallucinate and come up with crazy things, maybe that's ok given that we can filter out the sensible and novel ones.

[runarberg]

And in doing so you spend what, a 100 watt hours per bad idea? Compared to how many megawatt hours of AIs failed attempts at proving math capabilities to investors only to prolong the AI bubble another month?

I bet your stupid ideas also taught you a valuable lesson and you learned at least something from the experience, maybe your next idea won’t be so dumb, and those 100 watt hours weren’t actually wasted (though it may feel like they were). Compered to a failed LLM experiment, where all those billions of billions of computations are completely wasted. the model knows exactly as much after a failed experiment as it did going into it. Those Megawatt hours were simply wasted, turned into heat energy, paid for by raising the power bills of the of the datacenter’s neighbors.

[falcor84]

> As a human, I have many stupid and wrong ideas all day long - most of those don't bubble up to my conscious awareness

I'm sorry to nitpick, but isn't an unconscious idea an oxymoron?

[salad-tycoon]

Kind of seems to me, the heart of the critique is that 1. unthinkable amounts of financial and social and political credit have been thrown at this which necessarily has deducted it from other fields we could have invested in, instead. 2. Thus, with such wealth you would expect at least a couple of discoveries.

Not my post, but I think point 1 is stronger than 2.

[eru]

> 1. unthinkable amounts of financial and social and political credit have been thrown at this which necessarily has deducted it from other fields we could have invested in, instead.

That's not necessarily true. If our only counterfactual to investing resources in project A were to invest them in some other project B, then, yes, the conclusion above follows. But often people just consume the resources.

(In the end, the goal of all economic activity is consumption. We invest resources so that we can consume more later. If there's no good enough project around, might as well consume more now.)