thmzlt

programmer

[thmzlt @thmzlt]

@andreshalders Derrubar site não funciona. Tu tem um negócio chamado CloudFlare por onde passa 1/3 da Internet. Se bloquear para de funcionar tudo.

[thmzlt @thmzlt]

@ayubio @surtasa_dev A gente roda uma infra aqui em bare metal (leasing) por ~50k usd/mês que não sairia por menos de 200k na AWS. Nem precisa ir tão longe no self-hosting.

[thmzlt @thmzlt]

@ayubio Como assim representante do Debian? Não tenho familiaridade com a governança do projeto.

[thmzlt @thmzlt]

@annanay @QFEX "CS problems we have to tackle" can you share a vague example?

[thmzlt @thmzlt]

@roderix1966 Craque!

[thmzlt @thmzlt]

Dwarkesh is killing it with these lectures

[Dwarkesh Patel @dwarkesh_sp]

New blackboard lecture w @ericjang11

He walks through how to build AlphaGo from scratch, but with modern AI tools.

Sometimes you understand the future better by stepping backward. AlphaGo is still the cleanest worked example of the primitives of intelligence: search, learning from experience, and self-play. You have to go back to 2017 to get insight into how the more general AIs of the future might learn.

Once he explained how AlphaGo works, it gave us the context to have a discussion about how RL works in LLMs and how it could work better – naive policy gradient RL has to figure out which of the 100k+ tokens in your trajectory actually got you the right answer, while AlphaGo’s MCTS suggests a strictly better action every single move, giving you a training target that sidesteps the credit assignment problem. The way humans learn is surely closer to the second.

Eric also kickstarted an Autoresearch loop on his project. And it was very interesting to discuss which parts of AI research LLMs can already automate pretty well (implementing and running experiments, optimizing hyperparameters) and which they still struggle with (choosing the right question to investigate next, escaping research dead ends). Informative to all the recent discussion about when we should expect an intelligence explosion, and what it would look like from the inside.

Timestamps:

0:00:00 – Basics of Go
0:08:06 – Monte Carlo Tree Search
0:31:53 – What the neural network does
1:00:22 – Self-play
1:25:27 – Alternative RL approaches
1:45:36 – Why doesn’t MCTS work for LLMs
2:00:58 – Off-policy training
2:11:51 – RL is even more information inefficient than you thought
2:22:05 – Automated AI researchers

[thmzlt @thmzlt]

@PixelRainbowNFT @antirez What's your thermal treatment? I have the same machine.

[thmzlt @thmzlt]

"why should a project maintainer spend time reviewing and discussing that PR as opposed to firing up their own LLM to solve the same problem?"

[Simon Willison @simonw]

The Zig project's rationale for their blanket ban on AI-assisted contributions makes a lot of sense to me - for them, time spent reviewing PRs isn't about the code, it's about growing new contributors for the future of the project https://t.co/9DoJqBGphp

[thmzlt @thmzlt]

@chrislakin Person successfully self-teaches skill and then believes they can extrapolate to any other skill

[thmzlt @thmzlt]

@mitsuhiko Love it!

[thmzlt @thmzlt]

@henriquebastos @KentBeck Beautiful idea!

[thmzlt @thmzlt]

@lhcaroli Acredita numa diminuição do 2/20 tradicional?

[thmzlt @thmzlt]

@lhcaroli Do Barça, difícil falar de todos

[thmzlt @thmzlt]

@RealSpaceMonkey Yes! You still get most of the benefits while not feeling like you're playing Age of Empires

[thmzlt @thmzlt]

@paulsaladinomd What's up with the macbook display? Also there's variance between models.

[thmzlt @thmzlt]

@ayubio Como ipv6 ajuda com DDoS?

[thmzlt @thmzlt]

@ayubio agora apareceu o restante da thread (tinha visto como um post individual)

[thmzlt @thmzlt]

@ayubio @ayubio qual é a sua opinião? Até agora ipv6 tem sido um desastre (trabalho em infra)

[thmzlt @thmzlt]

@lhcaroli Por que não iria? Quem paga a conta é o investidor do CDB que não quebra