Most AI gains we've seen are in domains with fast, checkable answers (math, code), where RL on verifiable rewards works great. Judging complex systems means watching them run for months. That feedback loop is bound by physics: you can't speed up time. Solvable, just not quickly.
Replying to @Gystenghast
i am surprised you can agree they go from doing things that take minutes to "multi-week coding" in a year and not bat an eye at that
18
As an expert of a sort, it's painful to watch these. AI safety and regulation is good, but AI can't even build and maintain a simple app over time. How is superintelligence around the corner? We don't even know how to stop Microsoft Outlook from crashing with the things.
.@paulfchristiano thinks that on the current trajectory (absent any slowing down), there is a 1 in 3 chance that we could have superintelligence powerful enough to overthrow the government within a year. (He articulates a concrete scenario).
He argues that the labs should pause further algorithmic progress until they can secure their weights and research secrets. Otherwise China could simply steal the superintelligence in the middle of the intelligence explosion, and the labs would no longer be able to slow down even if they saw their models going off the rails.
He also explains how AIs will be trained during the intelligence explosion, and why even a system trained entirely inside a data center could still have the capabilities profile such that humans would have very little chance of competing with it or controlling it.
And he discusses the plan for alignment - basically, making a superintelligence that actually does what we ask the way we mean it. Fixing reward hacks and building better graders will help up till the next few model generations, but he thinks that no single current approach has better than even odds of letting us control superintelligence. The hope is that the AIs figure out the solution to this problem during the intelligence explosion.
Paul is head of the Alignment Research Center. Previously, he was head of AI safety at the US AI Safety Institute. Before that, he led alignment at OpenAI, where he invented RLHF. He’s recently joined OpenAI’s board (though here he is speaking as a private citizen).
2
2
101
It seems hard to explain that there is infinite work. Truly, people do not understand the limitless desire of mankind and vastness of the universe. "What will we do if AI takes all the boring jobs we spend our lives doing?" Like, I dunno, _literally_ anything you can imagine.
My mind sometimes wanders to a really unsettling conversation I had a few months ago
Essentially, this person kept asserting that AI creating post-scarcity conditions would render most people useless, and I kept trying to ask "what do you think people are for?"
10
I work with a dog rescue and all the childless old people in it have tons of friends and animals around at all times.
113
Replying to @AllanatrixQ
I'm not sure what other people are doing because my job is now much more research based and technical than it ever was. Before AI, 90% of my work was writing test classes and trying to argue senior devs out of whatever it was the last blog or YouTube video told them was good.
32
> That’s bullshit and loser thinking.
New: you think you know Proton, but you really don't.
I spoke to 8 current employees, got Slack chats, meeting audio, and an AI manifesto written by its CTO.
“That’s bullshit and loser thinking,” Yen wrote. “Stop thinking always of how to be slower, and focus more on how to be faster.”
CEO Andy Yen critiques AI in public. But inside, he is pushing everyone to use AI, even those with security concerns 404media.co/manifestos-fear-…
32
Gyst retweeted
the gift of alcohol and other drug experiences is to reveal to you that your sober state is no less strange nor surreal than your intoxicated one. in either case you are a phantom floating through physical sensation, and while you may be more cogent, you’re equally dislocated
The true art piece here is that they will grift money out of their supporters to get bailed out of any real consequences, proving actual fraud doesn't get punished.
A Minnesota prosecutor charges conservative influencers James O’Keefe and Cam Higby with felony counts of unlawful voting and identity theft after they misrepresented themselves at polling places using the names of registered voters. nbcnews.com/politics/2026-el…
1
25
I think this misrepresents addiction for most people. People think you stumble onto magic drugs and get hooked. It misses how often addiction is self-medicating. Many people could do heroin tomorrow and not get addicted. Long-term, it sticks because it solves a problem you have.
1
2
27
I've got bad news for you about how all humans think about each other.
if you weren’t raised as a woman you wouldn’t understand ngl
its the feeling that your identity is tied to your gender. you will always see me as a woman and not who i actually am. i am never just your friend i am always a possible love interest.
115




























