Keep on to blur preview images; turn off to show them clearly
Taste of what LLM-driven scientific discovery will be like: with 12 minutes of thinking, GPT-5 Pro suggested repurposing a known drug to treat an untreatable food allergy. Same exact result was found by an (at the time unpublished) peer-reviewed study. And models still improving.
President & Co-Founder @OpenAI
RT @morganb: It’s real and it’s spectacular.
🏗️ Love to build stuff (@runwayco, @sandboxvr, @postmates, @zynga) people love. 💸 Investor @amplitude_hq, @mercury, @owner, @elevenlabsio, @meetgamma ++
RT @soleio: I’ve designed software, backed startups and helped founders. Now I’m writing fiction. This piece is from a collection of fable…
🏗️ Love to build stuff (@runwayco, @sandboxvr, @postmates, @zynga) people love. 💸 Investor @amplitude_hq, @mercury, @owner, @elevenlabsio, @meetgamma ++
broader context here:
Asst professor @MIT EECS & CSAIL (@nlp_mit). Author of https://t.co/VgyLxl0oa1 and https://t.co/ZZaSzaRaZ7 (@DSPyOSS). Prev: CS PhD @StanfordNLP. Research @Databricks.
though, as ever, once a capability is achieved with a system, it’s easy to distill it back into model weights:
broader context here:
One reason I’m more bullish than ever on “LLMs” is that essentially *no one* in the field is a DNN forward pass maximalist or a self-supervision maximalist anymore. It’s just not the same meaning as 2020-22 LLMs.
though, as ever, once a capability is achieved with a system, it’s easy to distill it back into model weights: