Pinned
- Deep learning hype in one picture (NIPS conference registrations, 2002 through 2017) #nips2017
- Those who discard products as "just a wrapper on top of GPT3" would have discarded the whole software industry in 1972 as "just a wrapper on top of the C programming language". They miss the point. 👇
- FDA approved 521 medical AI tools so far. 3/4 of those products are for radiology.
- My journey from building bots in my living room to Zuck's office to personalized healthcare: 🧵
- Healthcare is so difficult. The software/AI part is easy. Connecting it to the real world is hard. You cannot just hope somebody else will do the dirty ops work for you. I've founded and sold 2 software startups before. It was just a warmup. The real game starts now!
- I don't get it. Training a LLM on "trusted" data doesn't make its output more trustable. You can make a LLM-based app safer by working at the application level, or with an adversarial model approach -- I don't see how one can make an LLM safer per se.What does the future of trusted AI-enabled Health Assurance look like? Enter Hippocratic. We're thrilled to leverage our Creation strategy, co-lead Hippocratic's $50M seed & welcome Munjal Shah & team to the GC family! From @htaneja & @AlexandreMomeni ↓ generalcatalyst.com/perspectives/o…
- - AI won't replace doctors. But doctors who use AI will replace doctors who don't. Get your headphones ready for a 🔥 discussion on @20vcFund with @HarryStebbings in a couple of days!
- Very happy to share two news today: we are launching the healthcare app that every woman deserves, and we raised 17m€ with great investors. @NablaTech nabla.com/en
- We are pleased to announce ParlAI (pronounced “par-lay”), a software framework for dialog research: parl.ai










