kinda seems to me like when people say "emergent property/behavior" they just mean "good at what it was designed to do"

Yeah definitely feels a lot like self-driving.

Precisely zilch, zero.

I mean we saw TensorFlow

100% chance someone will take this as proof of consciousness

Moved up to rank 33 on #kaggle. I'm not addicted. I can quit when I want. kaggle.com/competitions/b…

Moved up to rank 19 #kaggle kaggle.com/competitions/b…

Hot take: ML researchers are underestimating how quickly recent scaling laws may flatten out - it’s quite likely what people see as an exponential function is a sigmoid, and the harsh reality of the physics of high energy costs and power plant construction restrict the expected

Unpopular opinion: AI agents are going to do some things well, but fail to generalize enough to be useful.

Not surprising to anyone that’s actually worked on generative models.

Someone remind me how many years it was between the first GAN paper and realistic faces that still had artifacts.

dax we heard you like abstracting so we abstracted your abstraction to give you more abstraction 🤩

When is Astral going to replace mypy? 😁

So we tuning on the test set now?

I'd like to see the perfomance of a random sample from the same companies ChatGPT was given to pick from for a comparison.

what do you do when you get rate limited by your AI therapist?

I find deep research from OpenAI super useful. However, I would like to see improvement in the output format. I always want it to be in markdown but it leaves out information when asked to convert it to markdown. If anyone has luck with getting the full output in a markdown

Is making a team and paying for a seat the only way to get notifications for Weights & Biases anymore?