Jacob Berkowitz
@jacobberkowitz.bsky.social
AI interpretability researcher
95% confident there’s just a tiny dude in my computer
(he/him) 🏳️🌈
95% confident there’s just a tiny dude in my computer
(he/him) 🏳️🌈
Reposted by Jacob Berkowitz
“is my calculator horny?“ our tech columnist asks. “i entered 5318008 into it and turned it upside down. what i saw surprised me”
“Can ChatGPT experience joy or suffering? Does Gemini deserve human rights?” our tech columnist asks. “Many A.I. experts I know would say no, not yet, not even close. But I was intrigued.”
Should We Start Taking the Welfare of A.I. Seriously?
As artificial intelligence systems become smarter, one A.I. company is trying to figure out what to do if they become conscious.
www.nytimes.com
April 24, 2025 at 6:24 PM
“is my calculator horny?“ our tech columnist asks. “i entered 5318008 into it and turned it upside down. what i saw surprised me”
Reposted by Jacob Berkowitz
Freedom of reach is the real battle in social media right now. Algorithmic feeds vs. deterministic timelines isn't just UX - it's about whether platforms or users control what gets seen. And these choices fundamentally reshape online communities. 1/6
March 1, 2025 at 6:17 PM
Freedom of reach is the real battle in social media right now. Algorithmic feeds vs. deterministic timelines isn't just UX - it's about whether platforms or users control what gets seen. And these choices fundamentally reshape online communities. 1/6
Reposted by Jacob Berkowitz
The fact that Deepseek R1 was released three days /before/ Stargate means these guys stood in front of Trump and said they needed half a trillion dollars while they knew R1 was open source and trained for $5M.
Beautiful.
Beautiful.
January 28, 2025 at 3:02 AM
The fact that Deepseek R1 was released three days /before/ Stargate means these guys stood in front of Trump and said they needed half a trillion dollars while they knew R1 was open source and trained for $5M.
Beautiful.
Beautiful.
Reposted by Jacob Berkowitz
DeepSeek’s R1 model is challenging the very foundations of the past two years of generative AI hype, wiping $1 trillion of the value of major US tech companies.
But it’s also another example that the US strategy to try to contain Chinese tech is failing — and pushing them to get even more creative.
But it’s also another example that the US strategy to try to contain Chinese tech is failing — and pushing them to get even more creative.
DeepSeek shows the US failure to contain Chinese tech
Despite chip restrictions, Chinese AI threatens the foundation of the generative AI hype cycle
www.disconnect.blog
January 28, 2025 at 3:15 AM
DeepSeek’s R1 model is challenging the very foundations of the past two years of generative AI hype, wiping $1 trillion of the value of major US tech companies.
But it’s also another example that the US strategy to try to contain Chinese tech is failing — and pushing them to get even more creative.
But it’s also another example that the US strategy to try to contain Chinese tech is failing — and pushing them to get even more creative.
Reposted by Jacob Berkowitz
someone forgot to scrub the author metadata in OPM memos... www.404media.co/opm-memos-to...
Memos to Federal Employees Were Written By People With Ties to Project 2025, Metadata Shows
James Sherk and Noah Peters appear as the authors of memos sent by the Office of Personnel Management.
www.404media.co
January 27, 2025 at 10:46 PM
someone forgot to scrub the author metadata in OPM memos... www.404media.co/opm-memos-to...