AGI is an engineering problem, not a model training problem

Post by habinero » Wed, Aug 27, 2025, 6:35 PM UTC

danenania wrote: post #430

player1234 wrote: post #429

Earlier quoted context omitted.

[flagged]

Sorry you got triggered. I know it can be an emotional topic for some people. I'll try to explain in a simple way. We clearly are replicating at least some significant aspects of human intelligence via LLMs, despite biological complexity. So we obviously don't need a 100% complete understanding of the corresponding biology to build things which achieve similar goals. In other words, we can (conceivably) figure out ho…

> We clearly are replicating at least some significant aspects of human intelligence via LLMs

That's a very load-bearing "clearly". I don't think it's clear at all lol.

Again, you are vastly underestimating the scale here.

Heavier-than-air flight is (relatively) straightforward and it's easy for humans to build models. Also, you know when you're flying.

Building a star is theoretically also relatively straightforward -- just collect a lot of gas and dust in one area and wait for gravity to do its thing.

Actually doing that is left as an exercise to the reader.

Post by danenania » Wed, Aug 27, 2025, 7:18 PM UTC

habinero wrote: post #441

danenania wrote: post #430

Earlier quoted context omitted.

Sorry you got triggered. I know it can be an emotional topic for some people. I'll try to explain in a simple way. We clearly are replicating at least some significant aspects of human intelligence via LLMs, despite biological complexity. So we obviously don't need a 100% complete understanding of the corresponding biology to build things which achieve similar goals. In other words, we can (conceivably) figure out ho…

> We clearly are replicating at least some significant aspects of human intelligence via LLMs That's a very load-bearing "clearly". I don't think it's clear at all lol. Again, you are vastly underestimating the scale here. Heavier-than-air flight is (relatively) straightforward and it's easy for humans to build models. Also, you know when you're flying. Building a star is theoretically also relatively straightforward…

> That's a very load-bearing "clearly". I don't think it's clear at all lol.

Are you really disputing that LLMs can replicate some aspects of human intelligence? I mean, they're often passing the Turing test, writing non-trivial programs, and got a gold medal in the IMO.

Maybe you aren't well informed on their actual capabilities?

> Heavier-than-air flight is (relatively) straightforward and it's easy for humans to build models.

Actually, before the first airplanes were created, plenty of people were making arguments very similar to yours to dismiss human flight as impossible.

"Heavier-than-air flying machines are impossible." - Lord Kelvin, 1895

Re: AGI is an engineering problem, not a model training problem

Re: AGI is an engineering problem, not a model training problem