🤯 Studying how minds change
👩🏼💻 Building science tools
🦋 ♾️ 👓
🌐 maxine.science
This is a well-thought-out and balanced perspective. It incorporates a lot of what I’ve seen personally.
This is a well-thought-out and balanced perspective. It incorporates a lot of what I’ve seen personally.
this is truly incredible art.
youtu.be/ef568d0CrRY?...
this is truly incredible art.
"After locally tuning some of the hyperparameters, I swept out a number of models fixing the FLOPs budget. (For every FLOPs target you can train a small model a long time, or a big model for a short time.) It turns out that nanochat obeys very nice scaling laws"
Amazed people still think these are “altruists”.
Amazed people still think these are “altruists”.
“Actually, there is no such thing *at all* as a decision in a naturalistic setting, because a ‘decision’ implies a prespecified exogenous task structure supplying a rigid discretization of action and outcome spaces that does not occur in real life”.
People's intuitions seem to differ on what "decision" means.
For example: do animals (say, cats) make decisions? Most people would say yes, but I think some would disagree.
Do plants make decisions?
“Actually, there is no such thing *at all* as a decision in a naturalistic setting, because a ‘decision’ implies a prespecified exogenous task structure supplying a rigid discretization of action and outcome spaces that does not occur in real life”.
quickslice.slices.network
Our latest work shows that pretraining ViTs on procedural symbolic data (eg sequences of balanced parentheses) makes subsequent standard training (eg on ImageNet) more data efficient! How is this possible?! ⬇️🧵