🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Humans seeing sparks of AGI in LLMs x.com/Yoda4ever/stat…

Someone at firstpost wrote an actual news article based on my shitpost 😂😂

Linuxopsys sudo mount /dev/box /mnt cat /dev/null > /mnt

Smoke-away sir the GPUs are melting

Stanford CS25 - Transformers United So much fun catching up with these Transformer lectures. There is a lot of content I'm already familiar with but I always love reviewing stuff to build on my understanding of complex concepts and learn new ones along the way. I find that

*Animate math derivation* 💯

This isn't a fractal, but I think this plot has a phase transition.

When BTechs from IIT join US colleges for MTech programs, they have to unlearn and learn a lot. The engineering curriculum in India needs to be revised and the practical + project should be much more rigorous. Right now BTech programs are run like 4 year cultural fests.

Miguel Piedrafita ✨ C is memory safe. Your code is not

Your training loss admits a trivial solution 🙄

I wrote a tutorial on diffusion models for undergrad and grad students. I tried my best to give intuitive explanations for complicated equations. Your feedback is much appreciated Thanks to those who suggested various reading materials to me arxiv.org/abs/2403.18103

MSFT has open sourced MS-DOS github.com/microsoft/MS-D… Under MIT license and correctly made it read only. It will be interesting to see what hobbyists will do with the forks. - DOS for ARM - a renewed life for “dos games” platform - what else ..?

Boy I miss these. He's 200% right, of course. Straight off the vine, nothing quite like it.

Overheard my mom on the phone with her sister: "Did she complete LLM training?" After the initial shock realized she was enquiring about my cousins Master of Law graduation.

editing and committing code to master directly on github. from your phone.

these guys think hardware means building a juicer, friend. early in my PhD, I did 500 nm UV lithography to study photon recycling in semiconductors; experiments would only work in winter because of weird thermal expansion/humidity effects. this is hard stuff.

Really interesting result of Twitter's thumbnail compression here: when you see this tweet on your timeline, a lot of the lines appear green; when you expand the tweet, there are fewer; and when you expand the image itself, you discover that there are no green lines at all

This is a fundamental problem for machine learning (ML) because ML models need to be trained on input-output pairs ie problem-solution pairs for language models. To make them output "I don't know", you need to collect all problem-"I don't know" pairs which is not possible.