silicon socialite
Joined May 2019
- Tweets 4,751
- Following 825
- Followers 5,308
- Likes 75,011
419 Photos and videos
whether or not you agree with the AI risk cause, holly is definitely right that being weird perverts on the internet is not really serving their goals well
Jun 9
I’m not going to work to defend Eliezer and Nate’s reputations when they willingly do shit like this. Eliezer suggested this! They badly want not to be mainstream, and it’s absolute trollish vanity that does not serve the AI Safety cause. Aella does nothing valuable and corrupts everything she touches. Do me a favor and block me if you disagree.
3
1
38
2,966
seems like a good recruiting strategy though
join team orange get the good models
Jun 9
mythos will be bad ON PURPOSE on ai "frontier llm research" tasks, this is very very sad for the research community
also the fact that this is un purpose not visible to the user is crazy
1
8
492
compute in memory works great for signal processing and cryptography workloads where you can fit the whole workload in on-chip memory
it was really promising for old small-model ML workloads, but LLMs are just too damn big
Jun 2
Compute in memory will never ever ever eeeeeeeeeeevvvvvvvver work.Anyone telling you otherwise is a liar !
If anyone walks into your fancy office on Sandhill road pitching you this just hold them there against their will until I get there.
1
1
23
5,516
hit the front page of hacker news for the second time with this one!
New on the blog: How the hell is Groq raising more money?
After Nvidia hired their team and licensed their tech, Groq is trying to rise from the ashes as a neocloud -- and it might just work.
8
1,167
zach retweeted
Jun 2
oh? you're pining after a girl from brooklyn with a septum piercing and a bob? and you live in murray hill? private equity? school at umich and rye country day? wow. yeah no that's awesome hey have you ever heard of that wrought iron medieval device that cleaves people in twain?
6
10
289
15,559