Watch an RL model trained on billions of cycles figure out the best way to peel a banana
_sonith · x · 2026-10-07
A fun demo thread: the author shares footage of an RL model trained on billions of cycles figuring out "the best way to peel a banana." A counterintuitive, entertaining AI demo.
More from Fun
- "OpenAI solves math while I answer Stripe billing pages — basically the same" — generativist · 2026-10-07
- "Convert all 722 preprints into 3blue1brown videos. Don't make mistakes" goes viral as a prompt meme — willcb · 2026-10-07
- AI turns OpenAI's 199-page zeta proof into a 2-minute 3D animation claiming first zero-free strip past 7/8 — imjustnewatai · 2026-10-07
- DevDayer flashes OpenAI's DevDay Gameboy to get RuneScape running on it — JasonBotterill · 2026-10-07
- From idea to working 'dots photo booth' in 48 hours at OpenAI DevDay after-party — gabrielchua · 2026-10-07
- Paul Allen overheard Gates plotting to dilute his stake during cancer treatment—kept shares, worth $40B — SumitGup · 2026-10-07