Not to brag, I may have achieved a breakthrough in LLM technology by discovering how well Claude reacts to the command: 'carcinize this' 🦀

Prompt: “Fashion photoshoot of sneakers inspired by Van Gogh,' selected from the first image shown in every case. 2023 versus 2024.

oneusefulthing.org/p/doing-stuff-…

AI has a jagged frontier. It is good at some things that seem very hard and bad at some things that seem really easy for humans.

One of the most consistent things that people are surprised by is how bad AI is at word games (“end sentences with a vowel,” “count words”) 👇 is why

Given the expected announcements from Apple Monday, here is what I wrote a few months ago about the likely near future of AI, with on-device AIs talking to cloud AIs.

We will see complex organizations of AI interacting with each other to solve issues. oneusefulthing.org/p/an-ai-haunte…

At current growth rates, AI runs out of easy-to-access high quality data by 2028, depending on how aggressive training is.

There are techniques that may extend this (eg synthetic data) and possibilities for using less-easily-accessed data. (Google & Meta are sitting on a lot).