[40four]

I’m sure I’m not the only one, but it seriously bothers me, the high ranking discussion and comments under this post about whether or not a model trained on data from this time period (or any other constrained period) could synthesize it and postulate “new” scientific ideas that we now accept as true in the future. The answer is a resounding “no”. Sorry for being so blunt, but that is the answer that is a consensus among experts, and you will come to the same answer after a relatively small mount of focus & critical thinking on the issue of how LLMs & other categories of “AI” work.

[saberience]

> The answer is a resounding “no”.

This is your assertion made without any supportive data or sources. It's nice to know your subjective opinion on the issue but your voice doesn't hold much weight making such a bold assertion devoid of any evidence/data.

[friendzis]

I understand where you are coming from, but not every field is hard science. In many fields we deal with some amount of randomness and attribute causality to correlations even if we do not have as much as a speculative hypothesis for a mechanism of action behind the supposed causality.

LLMs trained on data up to a strictly constrained point are our best vehicle to have a view (however biased) on something, detached from its origins and escape a local minima. The speculation is that such LLMs could help us look at correlational links accepted as truths and help us devise an alternative experimental path or craft arguments for such experiments.

Imagine you have an LLM trained on papers up to some threshold, feed your manuscript with correlational evidence and have an LLM point out uncontrolled confounders or something like that.

[hare2eternity]

Outside of science it would be an interesting pedagogic tool for many people. There is a tendency to imagine that people in the past saw the world much the same as we do. The expression "the past is a foreign country" resonates because we can empathise at some level that things were different, but we can't visit that country. "Talking" to a denizen of London in 1910 regarding world affairs, gender equality, economic opportunities, etc would be very interesting. Even if it can never be entirely accurate I think it would be enlightening.

[mexicocitinluez]

> but that is the answer that is a consensus among experts

Do you have any resources that back up such a big claim?

> relatively small mount of focus & critical thinking on the issue of how LLMs & other categories of “AI” work.

I don't understand this line of thought. Why wouldn't the ability to recognize patterns in existing literature or scientific publications result in potential new understandings? What critical thinking am I not doing?

> postulate “new” scientific ideas

What are you examples of "new" ideas that aren't based on existing ones?

When you say "other categories of AI", you're not including AlphaFold, are you?

[roywiggins]

I think it's pretty likely the answer is no, but the idea here is that you could actually test that assertion. I'm also pessimistic about it but that doesn't mean it wouldn't be a little interesting to try.

[PxldLtd]

I'm sorry but this is factually incorrect and I'm not sure what experts you are referring to here about there being concensus on this topic. I would love know. Geoffrey Hinton, Demis Hassabis, and Yann LeCun all heavily disagree with what you claim.

I think you might be confusing creation ex nihilo with combinatorial synthesis which LLMs excel at. The proposed scenario is a fantastic testcase for exactly this. This doesn't cover verification of course but that's not the question here. The question is wether an already known valid postulate can be synthesized.

[nomel]

I think the question is more about the concept, rather than the specific LLM architectures of today.