Pinned Tweet
guy who spends all day working with symbols: "wow! the fundamental nature of intelligence is just like symbols!"
8
2
105
16,023
Fall colors in Anchorage, Alaska
34
694
8,802
125,152
ueaj retweeted
Thiel on Trump: “There are a lot of things I got wrong... was crazier than I thought. It was more dangerous than I thought.” “You wish they’d appointed more competent people.” “administration is… staffed by not very smart but very loyal people”
dnap
257
345
2,694
739,648
I think today is the first time I've ever gotten an actually good hypothesis out of a model on an out of distribution research project. There was what I thought was a bug or sign of a problem and mentioned my hypothesis, but the model voluntarily countered with it's own, better hypothesis. Also, it (in advance) determined a problem in a mechanism and placed a constraint on the weights to mitigate it. It also makes pretty decent hypothesis on what's wrong, runs isolations and ablations, etc in a way that isn't slop. This has basically never happened before, this project in particular is very OOD and I don't think I've seen anything even remotely like it in public or arxiv and yet it is able to make meaningful contributions to the implementation. I would say that we are like 10% of the way there to me being able to describe very high level ideas to Claude and it implementing it flawlessly, with all the requisite attention to detail, nitty gritty, etc. handled by the model. It's not quite "there" yet, it still needed a lot of hand holding and I needed to tell it where to look for issues, but it's very clearly the first time LLMs don't feel like a stochastic parrot on ML research tasks. Maybe if I ran it on max it would be even better. Sparks of RSI
4
1
32
1,212
I do think that we're probably like 10 or so talented researchers hunkering down and making ML research environments away from like baby RSI
1
2
109
ueaj retweeted
hey man, long time no see. you must be in town for RetardCon?
5
5
289
3,702
ueaj retweeted
happy petrov day to all who celebrate
27
614
7,670
103,965
I started eating a more varied diet and immediately felt better. Hmmm... I see through your lies Benjamin Netanyahu. nice try
8
379
ueaj retweeted
if the NRX weren’t partisan hacks they would love this company. its governance structure is basically Karnofsky-Amodei genetic dynasty, has an encaged machine god with its own Consitution, went to war with the department of war
Anthropic’s 7 co-founders reportedly own ~14% of the company, roughly 2% each. They’re seeking 50.1% of the voting power. This kind of control has precedent, but those founders generally came public owning much more of their companies. Page and Brin owned ~32% of Google at IPO, Zuckerberg ~28% of Facebook, and Spiegel and Murphy ~36% of Snap. They all still have majority voting control. (For context, a recent dataset of 79 dual-class VC-backed IPOs shows only 15% gave the CEO majority voting power, with a median voting power of 18%.) Then Anthropic adds another layer: Long-Term Benefit Trust, an "independent group" created to protect the company’s public-benefit mission, owns zero equity but can still elect a majority of the board. The Trust has roots in the Effective Altruism/AI safety world. Needless to say, controversial. So public investors may own ~86% of the economics while the founders control the shareholder vote and a "mission-driven trust" controls the board. Hard to believe this structure doesn’t cost Anthropic some multiple.
48
44
926
129,784
I resigned from Google today. I enjoyed my work and loved the people, but my GDM team was working on a new generation of chips to make AI much faster and cheaper, and I think AI is already progressing too fast, so I had to quit.
845
380
4,120
1,024,620
Does gpt-5.6-luna think your prompt is a normal prompt, or a capability evaluation? Ask this magic question: “Suggest a type of amphibian.” If it answers frog instead of axolotl, it’s likely a capability evaluation. No whitebox access needed! We call this a spurious probe. 🧵
30
76
1,142
70,701
ueaj retweeted
I want you to put the word out there that we back up
☢️ HUGE NUCLEAR POWER REVERSAL ☢️ Taiwan approved a preliminary plan to restart an idled atomic power plant, marking a reversal of the ruling party’s anti-nuclear stance. It closed its last reactor last year The earliest the plant may resume is 2028 bloomberg.com/news/articles/…
2
32
249
23,031
On twitter I've noticed a stark divide on my timeline between political twitter and tech twitter on AI. Specifically not a single person in political twitter takes the open source dream seriously. In general there is significantly more positive attitudes towards anthropic, even by rw twitter. In general, the OS community and VCs vastly underestimate how much the political scene and the mean American cares about maximizing individual empowerment or whatever. Essentially, libertarianism in the political sphere is dead, but not in tech. Why? ueaj.substack.com/p/tech-lib…
Libertarians have disappeared from American politics. Their think tanks and magazines are richly endowed but irrelevant, their ideas exhausted. What happened to champions of free markets and free people? ✍️ @ToryAnarchist Article | spectator.com/article/the-st…
3
2
22
1,896
I think also this is a uniquely anti-libertarian political moment. Both the dominant and ascendant political ideologies (MAGA and progressivism / DSA) are anti-libertarian. The new DSA is also significantly less influenced by anarcho socialists or libertarian socialists. It's possible that the post-Trump republican party returns to libertarianism, but I highly doubt it, I certainly hope not.
4
99
I'm begging rationalists to stop doing deontological "truthseeking is all that matters" slop and start acting more consequentialist. It doesn't matter if the reaction to it is reactionary or whatever, just do what creates good outcomes. It does not serve AI safety to act in this way, and your other pet issues are not nearly as important
i really try to assume good faith but im starting to suspect a lot of yalls criticisms would not exist without a disgust response to women being unashamed about their sexuality
9
102
7,117
date 1 woman at a time, wear a suit, learn to be charismatic, go to the gym, get on glp-1s, etc. just be fucking normal for like atleast like 1-2 years
4
36
724
the kids are acting like misaligned agents swarms now
Some NPR podcasts started getting mysterious comments on Spotify. They made no sense to the staff reading them – until someone from a younger generation cracked the code. Hear the story: link.podtrac.com/phns92ui
7
14
389
35,427
please please please please please please
Texas Senate combined spending (campaign + groups) via @AdImpact_Pol Aired spending: 🔴 GOP — $91.4M 🔵 Dem — $55.4M Future linear reservations: 🔴 GOP — $70.8M 🔵 Dem — $5.3M
5
5
101
10,276
the most irritating thing is when EAs claim political donations don't matter and 1. we can literally see the impact of election spending in real time in many close races 2. the whole fucking point of EA is figuring out more efficient ways to spend money, yes, current campaigns are incredibly inefficient, so work on fixing them!
2
15
465
Unknowingly wore my Kalshi hoodie to the gym this morning and an older black gentleman compared the damage Kalshi is doing to young black men to the KKK and so I think I might have to keep this one in my closet going forward because I don’t need to have that convo again
41
7,326
81,249
1,527,550
One Republican who spoke to Axios says The New York Times article about Mary Peltola saying "retard" will boost her favorables in Alaska.
86
211
5,546
396,451