Love the Analogy from @karpathy of "LLMs are ghosts of internet-scale code" in his @dwarkesh_sp podcast.
- They're brilliant at re-creating what's common (boilerplate, CRUD apps), but get stuck on what's new.
- They'll try to "correct" your novel/new implementation with a standard one from their past life.
This is why learning to code, understanding system thinking and design principles makes you 10x more productive.
Sci Hub is a "pirate" website with 88M research papers freely available.
Using it is not advisable. We should all try to make billion-dollar acdemic publishers richer.
Here's a thread on integrating Zotero with Sci Hub to get free articles.
PLEASE DON'T DO THIS 😉
At #ICML2023, Chay (@wrong_whp) will present Hiera, a hierarchical vision transformer that is fast, powerful, and simple. Code+models at: https://t.co/barBBsup7N
If interested, please come to the oral presentation on Tue 25 Jul 5:30pm HST or poster #219 on Wed 26 Jul 2pm HST.
This has been my Whatsapp status for well over a year now!
The person who told me about it in an interview is now the 17th World Champion!
Congratulations Ding Liren for inspiring the world with your deep thinking.
Late night thought:
If LLMs or AI trained on our words are a true threat, and if the most powerful models will be trained on the latest data…
Then maybe spreading love and empathy is the best AI safety method. Every time we say something positive, we make it likely for future models to predict something positive.
Maybe too simple, but I don’t care. I love you all.
Came up in DMs w @BrandonGoldman