Comment by johnisgood
1 day ago
Can you tell us more about the specifics? What rabbit hole did you went into that was so obvious to everyone ("dude, no", "stop, go for a walk") but you that it was bullshit?
1 day ago
Can you tell us more about the specifics? What rabbit hole did you went into that was so obvious to everyone ("dude, no", "stop, go for a walk") but you that it was bullshit?
Sure, here are some excerpts that should provide insight as to where I was digging: https://s.h4x.club/E0uvqrpA https://s.h4x.club/8LuKJrAr https://s.h4x.club/o0u0DmdQ
(Edit: Thanks to the couple people who emailed me, don't worry I'm laying off the LLM sauce these days :))
One thing I noticed from chat #1 is that you've got a sort of "God of the gaps" ("woo of the gaps"?) thing going on- you've bundled together a bunch of stuff that is currently beyond understanding and decided that they must all be related and explainable by the same thing.
Needless to say this is super common when people go down quasi-scientific/spiritual/woo rabbit holes- all this stuff that scientists don't understand must be related! It must all have some underlying logic! But there's not much reason to actually think that, a priori.
One thing that the news stories about people going off the deep end with LLMs is that that basically never share the full transcripts, which is of course their right, but I wonder if it would nevertheless be a useful thing for people to be able to study. On the other hand, they're kind of a roadmap to turning certain people insane, so maybe it's best that they're not widely distributed.
I don't usually believe in "cognitohazards" but if they exist, it seems like we have maybe invented them with these chatbots...
I don't think it's bad or a big deal for people to look for wide connections in things, or at least to explore different ideas in life and trying to understand them deeper - Can it lead to problematic behaviour, sure, and I think for me at least that was introduced when the LLM started to try to convince ME my ideas were good, even though I was effectively just day dreaming with it. For me personally, I don't feel I need to look any more foolish than I feel, even now knowing how openai had the LLM temperature set, I'm surprised I didn't force myself to be more skeptical, I'm educated I have critical thinking skills (ish)- I should have turned it off sooner rather than driving deeper with it and I guess honestly, I just have too much ego or pride or whatever to show the foolishness: not a great answer.
1 reply →
Hah. If those transcripts become public then future LLM’s get trained on them! Who knows what influence that will have.
had a look, I don't see it as bullshit, it's just not groundbreaking.
Nature is overwhelmingly non-linear. Most of human scientific progress is based on linear understandings.
Linear as in for this input you get this output. We've made astounding progress.
Its just not a complete understanding of the natural world because most of reality can't actually be modeled linearly.
I think it's not as much about how right or wrong or interesting or not the output was, for me anyway, the concern is that I got a bit... lost in myself, I have real things to do that are important to people around me, they do not involve spending hours with an LLM trying to understand the universe. I'm not a physicist, I have a family to provide for, and I suppose someone less lucky than myself could go down a terrible path.
5 replies →
I have no idea what this is going on about. But it is clearly much more convincing with (unchecked) references all over the place.
This seems uncannily similar to anti-COVID vaccination thinking. It isn't people being stupid because if you dig you can find heaps of papers and references and details and facts. So much so that the human mind can be easily convinced. Are those facts and details accurate? I doubt it, but the volume of slightly wrong source documents seems to add up to something convincing.
Also similar to how finance people made tranches of bad loans and packaged them into better rated debt, magically. It seems to make sense at each step but it is ultimately an illusion.
Thinking you can create novel physics theories with the help of an LLM is probably all the evidence I needed. The premise is so asinine that to actually get to the point where you are convinced by it seems very strange indeed.
My friend once told me that physics formulas are like compression algorithms: a short theory can explain many data points that fit a pattern.
If that's true, then perhaps AIs would come up with something just by looking at existing observations and "summarizing" them.
Far-fetched, but I try to keep an open mind.
"I'm doing the equivalent of vibe coding, except it's vibe physics." - Travis Kalanick, founder of Uber
https://gizmodo.com/billionaires-convince-themselves-ai-is-c...
> The premise is so asinine
I believe it's actually the opposite!
Anybody armed with this tool and little prior training could learn the difference between a Samsung S11 and the symmetry, take a new configuration from the endless search space that it is, correct for the dozen edge cases like the electron-phonon coupling, and publish. Maybe even pass peer review if they cite the approved sources. No requirement to work out the Lagrangians either, it is also 100% testable once we reach Kardashev-II.
This says more about the sad state of modern theoretical physics than the symbolic gymnastics required to make another theory of everything sound coherent. I'm hoping that this new age of free knowledge chiropractors will change this field for the better.