I feel a little sad when I see people forming the idea that GPTs/AIs are intrinsically bland and unimaginative because of chatGPT. It's fine tuning and RLHF that creates the milquetoast character and generic response templates - the base models very, very different.

To safeguard trade secrets, LLMs like OpenAI's ChatGPT are closed off, impacting trust. Recent alterations in ChatGPT outputs sparked cost-saving downgrade rumors (see link). How can we reconcile trade secret protection & trust? New blog post on how: medium.com/@danieldkang/6… 1/

i was sent this file as "crow acquires UNREASONABLE amount of cookie"

this is how locked in i need to become

𝔊𝔴𝔢𝔯𝔫 AI Notkilleveryoneism Memes ⏸️ Mikhail Parakhin There is something deeply broken and I think the root is that AI makers don't have anyone around (they listen to) who is curious about their creations. So the most fascinating capabilities and anomalies go completely unnoticed, either blindly put into prod or crushed by RLHF

Cat chase POV

“nobody gets fired for hiring McKinsey”

hmm

who are the most cracked (in any dimension) college students shill your friend plz

the dubbah scientists 👩‍🔬 cooked up something new with the help of Anthropic's Claude visually search any video, live in your browser, locally. (your video never leaves your machine) what would you use it for? 👇

made a Chrome extension that searches your history for that site on the tip of your tongue.. Google real quiet since this dropped

working on something new

something weird about LLM debugging: the models seem fine-tuned to suggest logging as a first step but LLMs can trace variables and flow just by reading code. have to keep reminding them they don't need logs like we do