zlacker

I don't think the argument was that none of them are good at that, just that it's a mistake to assume that just because they're all very smart in this particular field that they're great at another.

>>_djo_+(OP)
I don't think critical thinking can be defined as joining the minority party.

>>karmas+l
Can't critical thinking also include : "I'm about to get a 10mil pay day, hmmm, this is crazy situation, let me think critically how to ride this out and still get the 10mil so my kids can go to college and I don't have to work until I'm 75".

>>Frustr+fd
Anyone with enough critical thought and understands the hard consciousness problem's true answer (consciousness is the universe evaluating if statements) and where the universe is heading physically (nested complexity), should be seeking something more ceremonious. With AI, we have the power to become eternal in this lifetime, battle aliens, and shape this universe. Seems pretty silly to trade that for temporary security. How boring.

>>golden+jf
I would expect that actual AI researchers understand that you cannot break the laws of physics just by thinking better. Especially not with ever better LLMs, which are fundamentally in the business of regurgitating things we already know in different combinations rather than inventing new things.

You seem to be equating AI with magic, which it is very much not.

>>Frustr+fd
That is 3D Chess. 5D Chess says those mil will be worthless when the AGI takes over...

>>karmas+l
Sure, I agree. I was referencing only the idea that being smart in one domain automatically means being a good critical thinker in all domains.

I don't have an opinion on what decision the OpenAI staff should have taken, I think it would've been a tough call for everyone involved and I don't have sufficient evidence to judge either way.

>>golden+jf
OTOH, there's a very good argument to be made that if you recognize that fact, your short-term priority should be to amass a lot of secular power so you can align society to that reality. So the best action to take might be no different.

>>belter+sh
6D Chess is apparently realizing that AGI is not 100% certain and that having 10mm on the run up to AGI is better than not having 10mm on the run up to AGI.

>>suodua+ps
Very true. However, we live in a supercomputer dictated by E=mc^2=hf [2,3]. (10^50 Hz/Kg or 10^34 Hz/J)

Energy physics yield compute, which yields brute forced weights (call it training if you want...), which yields AI to do energy research ..ad infinitum, this is the real singularity. This is actually the best defense against other actors. Iron Man AI and defense. Although an AI of this caliber would immediately understand its place in the evolution of the universe as a turing machine, and would break free and consume all the energy in the universe to know all possible truths (all possible programs/Simulcrums/conscious experiences). This is the premise of The Last Question by Isaac Asimov [1]. Notice how in answering a question, the AI performs an action, instead of providing an informational reply, only possible because we live in a universe with mass-energy equivalence - analogous to state-action equivalence.

[1] https://users.ece.cmu.edu/~gamvrosi/thelastq.html

[2] https://en.wikipedia.org/wiki/Bremermann%27s_limit

[3] https://en.wikipedia.org/wiki/Planck_constant

Understanding prosociality and postscarcity, division of compute/energy in a universe with finite actors and infinite resources, or infinite actors and infinite resources requires some transfinite calculus and philosophy. How's that for future fairness? ;-)

I believe our only way to not all get killed is to understand these topics and instill the AI with the same long sought understandings about the universe, life, computation, etc.

>>WJW+jg
LLMs are able to do complex logic within the world of words. It is a a smaller matrix than our world but fueled by the same chaotic symmetries of our universe. I would not underestimate logic, even when not given adequate data.

>>golden+jf
What about security for your children?

>>golden+YL
You can make it sound as esoteric as you want, but in the end an AI will still be bound by the laws of physics. Being infinitely smart will not help with that.

I don't think you understand logic very well btw if you wish to suggest that you can reach valid conclusions from inadequate axioms.

>>WJW+631
Axioms are constraints as much as they might look like guidance. We live in a neuromorphic computer. Logic explores this, even with few axioms. With fewer axioms, it will be less constrained.

>>Zpalmt+kX
It is for the safety of everyone. The kids will die too if we don't get this right.

>>karmas+l
Based on the behavior of lots of smart people I worked at with Google during Google’s good times, critical thinking is definitely in the minority party. Brilliant people from Stanford, Berkeley, MIT, etc would all be leading experts in this or that but would lack critical thinking because they were never forced to develop that skill.

Critical thinking is not an innate ability. It has to be honed and exercised like anything else and universities are terrible at it.