Repetition rewires Repetition rewires Repetition rewires As a neuroscientist, I can tell you: the brain doesn't distinguish between the habits that serve you and the ones that destroy you. The question isn't whether you'll change, but what you'll become.

since UMAPs are arbitrary anyway, might as well trick them out with nano banana

Stop watching short-form videos. Meta-analysis of ~100k people shows how it is associated with decreased cognition and increased stress + anxiety. I know this is correlation, not causation but ask yourself - do you feel great after a session of consuming reels/shorts?

RL optimized LLM learning new skills

I would do anything for a high quality image to editable vector graphic model to add on top of this. Would save scientists an insane number of hours

👀

Hyper-successful people are disciplined on a level you cannot fathom.

The fastest way to improve your life is to put yourself in an environment where you feel like you don’t belong and force yourself to level up.

According to the brief, Meta was aware that millions of adult strangers were contacting minors on its sites; that its products exacerbated mental health issues in teens; and that content related to eating disorders, suicide, and child sexual abuse was frequently detected, yet

Mark Zuckerberg literally explained why 99% of people quit early:

Your lack of commitment is insulting to the people who believe in you. Speak clearly. Stand tall. Keep your word. Do the thing you said you were going to do. Don’t betray their trust by showing up when it’s easy and disappearing when it’s hard.

if you're applying to Ph.D. programs in the coming weeks make sure your statement includes: (1) what problems you'd like to work on (2) which advisor you want to work with and why highlights of prior work are good in moderation but not at the expense of (1) and (2).

ICLR has placed OpenReview in a difficult position, so I want to offer a few words about the OpenReview team working behind the scenes. OpenReview has long been operated at UMass Amherst as a non-profit organization founded by Andrew McCallum. Each year, Andrew must raise more

Quiet Feature Learning in Transformers This is one of the most fascinating papers I have read this week. Let me explain: It argues that loss curves can mislead about what a model is learning. The default approach to monitoring neural network training relies on loss as the

Very strong new evidence against the notion that poverty causes crime. *Experimental* data on unconditional monthly payments to poor individuals in Finland, finds that the treatment (payment) has no causal effect on criminal perpetration or victimization.

Wait... Speed cameras work? "Cumulatively, over the seven months following [speed camera's] introduction, collisions declined by 30% and injuries by 16%."

all roads lead to the same subspace "The Universal Weight Subspace Hypothesis" 500 ViTs, 500 Mistral LoRAs, 50 LLaMA models all collapse into shared 16-dimension subspaces regardless of data Not every problem in AI is a data one

The roommate cleaning problem: Alice and Bob are roommates. Alice can put up with the apartment reaching a 6/10 in messiness. Bob is only slightly less sensitive to mess. He will clean at 6.5/10. In practice, Alice does all the cleaning, since they never reach 6.5.

Be delusional enough to believe that it's possible. Be disciplined enough to prove yourself right.

delian So interesting. Here’s the paper that it came from: link.springer.com/article/10.118… “Even more dramatic, it is inconceivable that life began with just a few nucleotides. Instead, there must have been an initial step–from zero to finite–in the complexity of organisms (as measured by