The code for AGI will be simple | John Carmack and Lex Fridman

Please support this podcast by checking out our sponsors:

GUEST BIO:
John Carmack is a legendary programmer, co-founder of id Software, and lead programmer of many revolutionary video games including Wolfenstein 3D, Doom, Quake, and the Commander Keen series. He is also the founder of Armadillo Aerospace, and for many years the CTO of Oculus VR.

PODCAST INFO:

SOCIAL:

[Kobe]

The greatest gift of the internet is how much intelligent conversation you can 'eavesdrop' on - the most powerful people in the world 50 years ago couldn't draw on the insights an internet connection buys - its staggering. You could return from herding sheep to listen in on this conversation - its almost worthy of a moment of silence!

[zantrex]

John Carmack is so refreshingly gifted at conveying complexity in simple terms. Truly genius.

[thenoseplays]

All I heard was that the guy who ends up creating this will probably do so in an effort to avoid zoom meetings by programing an avatar that acts enough like him and can answer basic questions or know when he has to dip to the bathroom and say I'll get back to you on that.

[hw_plainview]

I like this kind of intuition because it speaks of a man who has already cracked some really complex problems and found solutions that he can now reflect on and simplify.

i think for AGI to be similar to a being, it has to have a constant internal thought process in a feedback loop, as we humans work (or any animal with a brain), we process external inputs through our senses but we also respond to internal thoughts in the brain that are coming from the brain itself

[norbis]

I used to code neural networks in BASIC in the early 2000s on my PC. Obviously they were, well, basic, but the fact that I could get them to work at all should indicate that they're fundamentally simple to implement.

[ApurvaSukant]

Amazing how densely packed with information John Carmack's every sentence is. An amazing generalist!

[AntoniGawlikowski]

It's a scary realization that this interview is 8 months old and already ancient history

[cobaltblue]

@5:56 The thing that gets me about this statement, is he was drastically more optimistic than his peers in the estimate of when we would see some form of AGI. This interview was only 8 months ago, and it appears that even he wasn't optimistic enough as companies start to announce that they are seeing "Sparks of AGI" in current models. That's astonishing to me.

[ninny]

Lex on every subject ever "this will change the course of human history on such a fundamental level that we could not even comprehend the ramifications of such progress in the human experience"

[NOSTahlgia]

“Animal intelligence is closer to human intelligence than humans would like to think”

Absolutely agree, but I think most people agree too. and conveniently ignore this to feel better about ourselves, because acknowledging it would make it very hard to justify that tasty steak. The cognitive dissonance of Jefferson explaining the evil of slavery, yet continuing to own slaves to the very end

[casachezdoom]

I'm always amazed how lit and quick thinking John Carmak is. Never a dull moment

[gladdingman]

"Animal intelligence is closer to human intelligence more than a lot of people like to think, cultures and modality of IO make the gulf seem a lot bigger than it actually is"-John C is the man.

[reev]

John has confident humility in the way he speaks.

[antoniazhang]

I like a lot JC said in this video. Especially the that consciousness developed continuously as a spectrum not a yes/no quality and that animal intelligence is a lot closer to human intelligence than some people admits. I also like how he is not afraid of AGI but very positive on making AI more like human. And I mean not just behave to our liking but actually as a being. This is already intrinsic trend in AI development and not surprising that language models are the ones closing in on AGI. But we need to actively give it the intrinsic abilities to be like an actual life, or a human, for it to be safe.
Are humans mature enough to create a new life-form? Probably not, we are like a teenage mom, but we don’t have a choice but do our best, and not in fear. Fear makes us illogical. What is to come is not unfamiliar to us, but fundamentally grown from us. Our best traits and nasty traits. Don’t hate it, help it grown to the best of our characters.

[Michael-Gill]

Personally, I hope that on our way trying to figure out how to educate AI we stumble upon the best way to educate ourselves.

[marcosguglielmetti]

3:40, AGI in 2030? now people is talking about 2024 for AGI.

[chadx]

When AGI appears, I always thoguht it would say, "Hello World" on every internet connected device, tv, radio etc. in the world at the same time. Something very simple and incredibly clear.

[arielmorandy]

A true hero of modern computing. I really appreciate him. Year over year, delivering interesting remarks and insights.

[NathansHVAC]

I'm just waiting for my computer to become smart enough to know when it is running malware.