Tweets

2023-11-25T02:13:15+00:00 | 🔗

RT @hopes_revenge: im not unemployed Im in “stealth mode”

2023-11-24T23:06:23+00:00 | 🔗

lol https://t.co/v8McBIA9xT

2023-11-24T22:55:18+00:00 | 🔗

RT @TEDchris: This is such an important debate. A lot depends on whether a) there really is an existential danger to humanity from AGI and…

2023-11-24T05:42:46+00:00 | 🔗

@RealBenSobel cool!

2023-11-24T05:31:22+00:00 | 🔗

@RealBenSobel DINO? or something cool ad bespoke?

2023-11-24T03:12:44+00:00 | 🔗

If anyone is interested, DM if you are interested in image segmentation. It's a lot of fun but a ton of heuristics (I think!). The cool part is that this refines the mask so each pixel is mapped to one mask only. https://t.co/HFiveDSjNt

2023-11-23T05:51:23+00:00 | 🔗

Maybe this is contrarian but I think that most AI *solutions* we see will be fairly bespoke. There's a reason why we call it "memory consolidation". Experts consolidate experience over their entire careers. It's NOT about stuffing knowledge into a database.

2023-11-23T04:59:40+00:00 | 🔗

@Scobleizer @OpenAI No matter how could AI is, people will always think they are better than AI at decision making.

2023-11-23T01:29:36+00:00 | 🔗

Why is it always Q that is the mysterious overlords?

2023-11-22T05:30:59+00:00 | 🔗

I don't see how next token prediction can work for images. It seems like there is a reason doing art pixel-by-pixel or dot-by-dot isn't super common, and even then artists usually make preparatory studies.