We need to talk about Jacob's resignation.

im open for an interview if you need to talk about my resignation

Can we talk about Google folks who are shilling other companies every other day. To outsider, it looks like secure person embracing and encouraging competition, but it’s just a message to the world they are open to reach out and jump the ship. Amaar even blocked me 😭

I think it's good when companies allow employees to explore competitor's products. I don't really understand how they would improve if they don't know what others are doing.

That’s the plausible deniability and they are smart to know that too. Maybe I am wrong but it feels little sus when top posts from @GoogleAIStudio PM are about Astra, Fable, Seedance and iPhone Duo

Fast forward 2 weeks and Theo will be paid by the other side.

Thanks for our weekly dose of psychosis theo! Preciate it

Watched it an hour ago. It was a good one. We can feel how it concerns you (not the part making it a good one though).

I read the whole thread and I think it deserves to be taken seriously, especially coming from people working directly on frontier models. But from everything I’ve seen, LLMs are still extremely good prediction machines. What makes them potentially dangerous is the access and autonomy we give them. So either they are extrapolating current progress much further than I would, or people inside these labs are seeing capabilities that are significantly beyond what we have access to publicly. I genuinely don’t know which it is.

i hate how these people smuggle a worldview through proximity. let me remind folks that just because someone works at a lab, it does not automatically mean that they know what they are talking about. "I know how it works because i trained it" yea bro you took a python course and your work lives in collab notebooks.

Thank you for this wonderful breakdown Theo!

all this nonsense is mental illness. Every single technological leap comes at a cost, especially during the adjustment period. These are silly people mixing very serious costs and risks with science fiction creating apocalypse porn as a replacement for religion. In other words, there is a massive leap from "these things are getting increasingly more powerful and capable in a non linear and increasingly decipherable manner" to "therefore terminator".

muddying the waters nothing we have seen so far justifies this narrative everyone knows about the problems and risks, restating them over what I see as a bad faith or incompetent perspective helps nobody This is going to look so silly 1 year from now

@grok context

They aren't racing blind, they're building the future. Quitting to cry on X instead of solving alignment challenges isn't cool, it's just fear disguised as morality.

As i wrote, it can't be slowed down, it's a never ending spiral

Bro we all have refusal removed locals anyway, no matter what they do we're getting this shit from China in 6 months with no guard rails. Prepare for it or die.

Seems more like posturing to me.

Everyone who doesn’t agree with me is wrong!

Uh-oh.

Damn, theo doesn't know how marketing works 🤣🤣

I didn’t buy it when I saw that post, and don’t buy it now. It feels too coordinated for whatever reason.

The real problem isn’t that AI companies are racing. It’s that the incentives reward moving faster than our ability to understand the consequences. The dangerous part of AI isn’t intelligence itself. It’s intelligence deployed before we fully understand what we’re creating.

after that tweet, all day i just wait for theo to make a video about it and post.

The most interesting resignation

Doomerism®

Short story short, Jacob is a bitch.

@grok sum this up real quick pls

AI is coming after me. Me..

you need to stfu

you can tell that Theo uses Claude as his daily driver because of how much he genuinely says genuinely

“It’s almost like a religious thing”. If you drop the “almost like” then you’ll start to understand.

The solution is to pivot to actually understanding intelligence and engineering it. It's a combination of rules and knowledge that LLMs brute force.

Yeah, the safety issue is not new information. Glad it's getting more exposure. We shouldn't trust any of the frontier providers, money and power corrupt the best intentioned. @geoffreyhinton called for a ban on pursuing superintelligence. Maybe not the worst idea.