Keep on to blur preview images; turn off to show them clearly
ChatGPT just launched a Sora-powered holiday gift. Send a 🎁 emoji, upload a selfie, and get an AI cartoon. IMO, this is much more impressive than it looks on first glance. It pulls info from ChatGPT Memory to customize to you (ex. stockings had my dogs' names on them!)
This feels like a very cool way to intro Sora to the true mainstream consumer, who many not have downloaded the standalone app or created a Cameo. The fact that the little details are personalized to what you’ve shared with ChatGPT makes it feel extra magical.
When I sent🎁to ChatGPT "A compact Bayesian thought experiment: If an agent can self-improve, the value of additional resources is convex in capability. Even tiny advantages early have superlinear expected value, because they compound through better strategy, better modeling of rivals, and faster improvement. This is why races become unstable."
Economist. Author Singularity Rising. JD Stanford, PhD U of Chicago. Exploring path to the singularity—likely disaster, possibly utopia. Prof Smith College.
LichtFeld Studio is getting a sleek new sequencer with multiple export options, optimized for social media :)
Founder and CEO of https://t.co/5MjtfpwEU3 | Your guide to radiance fields | Host of the podcast @ViewDependent | FTP: 279 | discord: https://t.co/lrl64WGvlD
RT @GabRoXR: Playing #VR games "without" a headset... ... Absolute genius or total abomination? This is called @portalvr_io and it lets…
Opening portals to VR without headsets at @portalvr_io. Problems soluble, potential to improve invariant.
the progression from single-file edits/tab completions at this time last year to ~30 minutes of autonomous software engineering with Codex today has been nothing short of remarkable year after year, betting against deep learning remains a terrible idea
researcher @OpenAI | prev CMU
in particular, you see people come up with 101 Problems With RL Scaling. but then they don't apply remotely the same level of imagination when it comes to thinking of solutions it's worth remembering that all these "RL environment" companies are <6 months old. we're so early
the progression from single-file edits/tab completions at this time last year to ~30 minutes of autonomous software engineering with Codex today has been nothing short of remarkable year after year, betting against deep learning remains a terrible idea