On: Learning to Reason with LLMs

[exe34]

the first one anyway. after that it will find more efficient ways. we did, afterall.

[wahnfrieden → exe34]

it's not obviously achievable. for instance, we don't have the compute power to simulate cellular organisms of much complexity, and have not found efficiencies to scale that

[HeatrayEnjoyer → wahnfrieden]

Human level AGI only requires 20 watts

[oblio → HeatrayEnjoyer]

With a mechanism for AGI we don't comprehend at all.

Airplanes don't fly by flapping their wings.

[exe34 → oblio]

knowing that a mechanism exists is enough to motivate us to find one that works with our current or achievable tech.

[oblio → exe34]

We were motivated to fly since the earliest human.

[exe34 → oblio]

and we've had stories of artificial humans since we've been writing down stories!