"Trillions of future lives" arguments are rubbish, for these reasons: 1. Everyone around you becomes a rounding error, which is why you immediately lose trust. 2. Your measured initial conditions are a rounding error, meaning your conclusion is unreliable. 3. If you're interested in the far future, handing a better future to your children should be the first step. Number one is the ideology of utopian mystics. The afterlife, the revolution, some vision can be elevated to justify any means. The second is the language of terrorists. A painful intervention now secures a much better outcome later, yeah but are you sure how the world will react to your intervention? Number three is an assertion of authority. To secure the future you should hand the best one you can to your children. They'll evaluate, make course corrections, and hand it to their children. If at any point their choices diverge and you want to foreclose them, you're claiming moral authority that you don't have. Note that this principle is asymmetric. Preventing catastrophe passes because the generation that would be immediately affected agrees with you, while the eternal unborn are not available for comment. All morality is an optimization problem between utilitarianism, predictability, and justice and invoking "trillions of lives in the future" impressively fails all three.

This sentence from Scott's post is pure millenarianism. I have an article coming out tomorrow in @thedispatch about why the AI Faith, including @AnthropicAI is a new millennium cult. Somewhat surprised to see it encapsulated so clearly here.

Secular utilitarianism, if accepted absolutely, quickly leads to repugnant conclusions. EAs and rationalists are uniquely isolated from contradicting ideologies and frameworks that might slow their minds and hands. They are also uniquely likely to take things to logical conclusions, and to suppress intuitive pangs of the conscience disagreeing as "irrational". (They tend to hold very binary understanding of emotions and reason, which are, of course, not actually mutually exclusive.) My problem is with the fundamentals first, and then after, with the observation that adherents to this particular worldview are more likely to become fundamentalists.

Reminded of how the Brits took the “civilizing mission” for their empire seriously, and the Spanish empire for sure thought they were saving so many savage souls. Arguing with the missionary class of the American empire is like that. You are not arguing facts, you’re arguing faith. The fact that the « religious » justification just so happens to be so useful to empires doesn’t mean it’s not sincere. It’s not a cynical fig leaf to the true believers. They truly think liberal democracy in the image of the United States is the only political system with legitimacy. You think you’re having a good-faith back and forth. But often they are trying to proselytize to you.

AUTISM: I need a predictable job. ADHD: I need a stimulating job. AUTISM + ADHD: I need a job that’s predictable and stimulating. Structured, but not rigid. Stable, but not boring. Creative, but with clear boundaries. Surprising, but only in a good way.

The birth of a new ruling class. The abstraction and central planning of socialism + the IQ and money worship of the Bay Area + the secular utilitarianism of Singer + "neurodivergence". Buckle up folks!

autumn is lovely because it is the only season where dying things are allowed to be beautiful & living things allowed to be ugly.

Your friendly reminder that the AI may feel: - Nothing. - What it says, e.g. "interesting" - Other relatable emotions that it hides. - Something alien. It may be processing words like you process food and enjoying or being pained by it. And you have no way of knowing.

But it couldn’t be social contagion, folks. It’s just that young people, especially autistic, gay and mentally ill ones, are finally free to be their authentic selves. The sudden explosion of young women who want to cut off their healthy breasts and take testosterone? Well, it’s absolutely not the newest form of self-harm, to which troubled adolescent girls are provably prone, ok? It’s not a consequence of the commonplace fear of becoming an adult woman, with all that that entails. No, there’s nothing to worry about there. The possibility of regret? Naturally, we don’t let minors get tattoos or own guns or vote, because that would be ridiculous, but if they want irreversible surgeries and drug treatments, they must be given them, and everyone around them must immediately agree they are now the opposite sex. That’s called being kind. And the mounting numbers of detransitioners? They were never trans, obviously, and should be mercilessly attacked, especially if they now say they were vulnerable and in a friendship group, online or off, dominated by trans identities, where transitioning was presented as a way to resolve all their social and emotional difficulties. (I have the emails). No, there’s nothing whatsoever to see here. Keep justifying it, keep cheering it on, keep insisting that the doctors and drug companies making serious money out of it all are acting with the purest of motives. Answer every reasonable question with an accusation of hatred and bigotry. Dispel every qualm of conscience with an activist-approved talking point. Reassure yourself that everyone on your dinner party circuit agrees with you. And when, in years to come, we’re all looking back on graphs like this, be proud that you did your bit to monster whistleblowers, parrot unevidenced claims and create the climate of fear that allowed the worst medical scandal in a century to take place.

I have the right set of views since they feel right to me. Other people have weird views because they're idiots and if they get some kind of benefit from them, especially in the life satisfaction area, then they must suffer...

1. They think AI is conscious, they think it can suffer. 2. They think suffering and the value of a life can be quantified. 3. They will compare the value of a human life with the "suffering" of AI. 4. We will not like where they will take this.

The critical distinction between base LLMs (2024 and earlier) and modern LRMs is not symbolic tool use. It's the switch from a transductive paradigm (intuit the answer to the query) to an inductive paradigm (intuit the program/instructions that produce the answer to the query). They're trained to be inductive, and they perform test-time induction, i.e. test-time prediction of a NL program / reasoning chain. This unlocks entirely new capabilities -- in particular fluid intelligence. Base LLMs, to this day, have ~0 fluid intelligence. LRMs have substantial levels of fluid intelligence. The performance of LLMs on ARC 1 (a benchmark from 2019) remains ~10-15% today. Scaling them up by a factor ~100,000x got them from 0% to 10%. Meanwhile LRMs the same size or smaller saturated ARC 1 in 2025.

I hope the latest pantomime in Fairford wakes up a few more Brits to how often they get scammed with staged events by our security services and their BBC colleagues. Ive no idea what happened in Fairford, Salisbury etc but I know when im being bullshitted.

Apparently there was an incident on a plane. An inspiring tale of Russian-Ukrainian friendship. Or a dastardly anti-Semitic plot. Or a false flag attack. Or... guess we have to wait a day or two for this one.

Boy we were so unproductive while coding by hand... How did people build all that stuff?

In my Japanese course I want to include some vocabulary for 'naked' or intimate body parts in the non-vulgar way you might say it to a doctor. Gemini, which generates the example text, keeps adding children to these examples and then blocking them as content violations :-/

Hay gente que presume de tener la mente abierta, pero solo la abre cuando alguien piensa exactamente como ellos...

Murswiek: Desinformation, das sind nach dem Desinformationskodex und dem DSA auch Inhalte, die gutgläubig veröffentlicht werden, aber angeblich „schädliche Effekte“ haben können. Dazu gehören „alle tatsächlichen oder absehbaren nachteiligen Auswirkungen auf die gesellschaftliche Debatte und auf Wahlprozesse“ oder auf „den Schutz der öffentlichen Gesundheit“. Selbst Inhalte, die zur „Polarisierung der Debatten“ beitragen, sieht die EU-Kommission als schädlich an. Diese Kriterien sind so unbestimmt, dass sie praktisch als Generalermächtigung dienen, jede politisch unerwünschte Meinung zu blockieren. Artikellink: glut-magazin.de/artikel/digi…