qemqemqem's profile picture. Ideas and ideals bring us together. Talk to me about neuroscience, AI, or effective altruism.

Andrew Keenan Richardson

@qemqemqem

Ideas and ideals bring us together. Talk to me about neuroscience, AI, or effective altruism.

Soliciting anonymous constructive feedback: admonymous.co/andrew-keenan-…


What can't AI solve even if it becomes superintelligent? There are challenges in everyday problems that frustrate you: planning efficient routes, packing for vacation, scheduling meetings. Some problems are provably, permanently hard: mechanisticmind.substack.com/p/many-common-… #AI #NPHardProblems


I compared human learning to LLM scaling laws (Chinchilla). By this heuristic, humans are over-parameterized by 4 orders of magnitude! Why is that? @karpathy @elonmusk @ylecun #neuroscience #machinelearning mechanisticmind.substack.com/p/humans-are-o…


Personalizing your terminal with bash aliases is a game-changer. Stop memorizing complex terminal commands! I share my .bash_aliases file and how it has helped me grow. #bash #linux #terminal mechanisticmind.substack.com/p/using-the-te…


I reviewed Claude Code and compared it to Aider. CC has a great design, and I'll be using it sometimes (especially in large code bases), but I'm sticking with Aider as my primary tool for now. @AnthropicAI @paulgauthier mechanisticmind.substack.com/p/claude-code-…


A loss for Imbue! Vincent is extremely hard-working, knowledgeable, and great to work with

i’ve decided to take a break from working on codegen at @Imbue and am looking to explore a bit + figure out what to focus on 🙂 please send cool papers to read, problems that are not being thought about enough, people you think i should meet!!



Andrew Keenan Richardson reposted

Have you ever done a dense grid search over neural network hyperparameters? Like a *really dense* grid search? It looks like this (!!). Blueish colors correspond to hyperparameters for which training converges, redish colors to hyperparameters for which training diverges.


As a Boston -> SF transplant, this chart showing biotech markers by city (JLL 2022), captures how I feel about the cities. Boston has more talent (in biotech) but SF has more money In SF more likely to hear how biotech will change the world, in Boston more likely to hear details

qemqemqem's tweet image. As a Boston -> SF transplant, this chart showing biotech markers by city (JLL 2022), captures how I feel about the cities. Boston has more talent (in biotech) but SF has more money

In SF more likely to hear how biotech will change the world, in Boston more likely to hear details

Andrew Keenan Richardson reposted

Some exciting updates from @AsteraInstitute! Check out our new website for latest on our work in AGI, science, & climate 👇Quick thread on our newest area of growth - evolution of open scientific publishing 📖🧵 astera.org


I've asked a lot of people about their experience of the inner monologue and some people do report having no linguistic inner thought process. Some of those people seem smart and thoughtful, somewhat to my surprise. There's more variation in inner experience than it seems.

reading about completely normal stuff again reddit.com/r/self/comment…

QiaochuYuan's tweet image. reading about completely normal stuff again

reddit.com/r/self/comment…


I recently switched to @NotionHQ and set up a personal knowledge wiki. Great experience and I recommend. mechanisticmind.com/my-experience-…


I used to struggle to get out of bed every morning, but I made a few changes and now I reliably get up when I want to. Sharing some reflections about what I changed. mechanisticmind.com/how-i-learned-…


Andrew Keenan Richardson reposted

90 Days of Diffusion 90 🤯 AI Advances 👇


Andrew Keenan Richardson reposted

Today, AI systems can create stunning art & beat humans at chess & Go. But they can't do things a 3-year-old can do. Why? We're launching @genintelligent & open sourcing our research environment, Avalon, so as a community we can answer these questions: generallyintelligent.com/launch/


Loading...

Something went wrong.


Something went wrong.