Celento George

[Andrej Karpathy @karpathy]

I had the same thought so I've been playing with it in nanochat. E.g. here's 8 agents (4 claude, 4 codex), with 1 GPU each running nanochat experiments (trying to delete logit softcap without regression). The TLDR is that it doesn't work and it's a mess... but it's still very pretty to look at :)

I tried a few setups: 8 independent solo researchers, 1 chief scientist giving work to 8 junior researchers, etc. Each research program is a git branch, each scientist forks it into a feature branch, git worktrees for isolation, simple files for comms, skip Docker/VMs for simplicity atm (I find that instructions are enough to prevent interference). Research org runs in tmux window grids of interactive sessions (like Teams) so that it's pretty to look at, see their individual work, and "take over" if needed, i.e. no -p.

But ok the reason it doesn't work so far is that the agents' ideas are just pretty bad out of the box, even at highest intelligence. They don't think carefully though experiment design, they run a bit non-sensical variations, they don't create strong baselines and ablate things properly, they don't carefully control for runtime or flops. (just as an example, an agent yesterday "discovered" that increasing the hidden size of the network improves the validation loss, which is a totally spurious result given that a bigger network will have a lower validation loss in the infinite data regime, but then it also trains for a lot longer, it's not clear why I had to come in to point that out). They are very good at implementing any given well-scoped and described idea but they don't creatively generate them.

But the goal is that you are now programming an organization (e.g. a "research org") and its individual agents, so the "source code" is the collection of prompts, skills, tools, etc. and processes that make it up. E.g. a daily standup in the morning is now part of the "org code". And optimizing nanochat pretraining is just one of the many tasks (almost like an eval). Then - given an arbitrary task, how quickly does your research org generate progress on it?

[Celento George @celento]

#MyXAnniversary

[Serotonin @serotonin_hq]

Hello again, world. 🙂

[Celento George @celento]

@serotonin_hq 👀👀👀

[Celento George @celento]

@serotonin_hq 👀

[Celento George @celento]

@ali_moiz @sandhill_mkts @StonkyOli @m2jr @ericries Love the rebrand! The new landing page looks sleek. Congrats @ali_moiz @StonkyOli

[Tim Urban @waitbutwhy]

22 thoughts from 2022 I'd like to take into 2023:
https://t.co/bo83E67GbW

[Years Progress @YearsProgress]

▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓ 99%

[Years Progress @YearsProgress]

▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓ 98%

[Years Progress @YearsProgress]

▓▓▓▓▓▓▓▓▓▓▓▓▓░░ 86%

[Shane Parrish @shaneparrish]

“Your personal experiences make up maybe 0.00000001% of what’s happened in the world but maybe 80% of how you think the world works.”
— Morgan Housel

Studying the past is a shortcut to expanding your experience.

[VV @visualizevalue]

NFTs, explained.

[Years Progress @YearsProgress]

░░░░░░░░░░░░░░░ 0%

[Andrej Karpathy @karpathy]

floats aren't real! 😂I can't be the first one to notice

[Shane Parrish @shaneparrish]

Boring progress compounds into exceptional results.

[Years Progress @YearsProgress]

▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓ 98%

[Celento George @celento]

Merry Christmas Y’all 🎅🌲

[Years Progress @YearsProgress]

▓▓▓▓▓▓▓▓▓▓▓▓▓▓░ 95%

[Ali Moiz @ali_moiz]

hot off the press 🔥 "How AngelList and Stonks solve for Adverse Selection in Investing [Part 2]"
https://t.co/ObNVu7du4p

For old AngelList users, curious to hear if this jibes with your experience over the years investing there?