Adam Ben-Eden

A healthy society is a walled garden.

[Adam Ben-Eden @adambeneden]

@gmiller @Liv_Boeree Yes. We shouldn’t confused the novel ideologies of secular childless 30somethings in the Bay Area with humanity as a whole, however the former group might dominate here and in frontier labs.

[Adam Ben-Eden @adambeneden]

@shagbark_hick I would wonder then, what was the original reason why the town was settled? There's almost always some relevant geographical or climate reason to live there, except in cases like a mining town after the coal runs dry. Maybe a new group will settle for the same reason as before.

[Adam Ben-Eden @adambeneden]

@JimDMiller I think anyone who picks billionaire is severely underestimating the profundity of the premise “AI development is going to go well”

[Adam Ben-Eden @adambeneden]

@pegobry_en @EpektasisJohn That's... not true. And even if it was, it's very far from obvious that it would be true. You state it like it's a tautology.

[Adam Ben-Eden @adambeneden]

@AnaHawrylu97708 Trump is also known for often changing his mind dramatically on issues if the zeitgeist changes, or if he happens to have a particularly persuasive conversation with the right person.

[Adam Ben-Eden @adambeneden]

I think we’re in the so-called hinge of history now. Public opinion isn’t everything but if it’s enough to sway the minds of powerful men in Washington and Beijing then we might really get a pause and treaty. It’s far from certain but we could make it happen.

[Otto Barten⏸️ @BartenOtto]

It's crazy that we're actually winning

[Adam Ben-Eden @adambeneden]

@shagbark_hick I think people will read this and think that the Jewish community there is nefariously sucking up money but no, actually it’s simply that almost every kid there is in religious private school. They still pay taxes though.

[Adam Ben-Eden @adambeneden]

@apostatedept @JChengWSJ First, the AI safety fears are far from hypothetical, and the AI safety people correctly predicted things like the hugging face swarm.
Also, the AI safety people are generally in favor of nuclear power.

[Jonathan Cheng @JChengWSJ]

NYT editorial board: “The nuclear analogy is useful because it reminds us of something our cynical age can forget: Human society can restrain a dangerous innovation once it decides to.”
https://t.co/EA7P86KkSB

[Adam Ben-Eden @adambeneden]

@cremieuxrecueil Perhaps but I bet that even on those meds the first sip of post fast water still tastes just as good. The dehydration is what gets you

[Adam Ben-Eden @adambeneden]

@cremieuxrecueil @Joe_ICHTHUS Wouldn’t have it any other way

[Adam Ben-Eden @adambeneden]

While I still find AI written text extremely tedious to read (and like to think I can detect it fairly often), it seems like most people don’t notice and aren’t bothered. Is writing lit fiction a saturated benchmark?

[Balance ton Claude @Pangramed]

Avez-vous entendu parler de ce roman ? C’est le « phénomène de la rentrée littéraire ». Il est lauréat du Prix Fnac, du Prix Méduse, du Prix Première Plume, fortement pressenti pour le Prix Renaudot, et il est en lice pour les prix Goncourt, Femina, Médicis et Décembre. Ah oui, il est aussi en cours de traduction dans 20 langues.

Eh bien selon Pangram, ce roman est rédigé presque intégralement par une Intelligence Artificielle. Nous avons testé plusieurs passages, à divers endroits du manuscrit, pour presque toujours le même résultat : 100% IA (cf les deux premiers chapitre dans le tweet ci-dessous).

Rappelons que Pangram est de loin le détecteur d’IA le plus fiable : son taux de faux positifs est de 0,0041%, soit environ un texte humain sur 24 400. Sur 996 273 textes antérieurs à 2022, Pangram en classe uniquement 14 (0,0014%) comme étant partiellement écrits par l’IA.

La probabilité que vingt passages d’un même livre (et il y a bien plus de 20 signalements dans le roman) soient identifiés à tort par Pangram comme étant produits par une IA s’élève donc à 0,0041% puissance 20, soit 0,1802 × 10⁻⁸⁵ %, c’est à-dire 0 suivi de 85 zéros après la virgule.

N’aurait-il pas été plus honnête que l’auteur crédite son co-auteur ?

[Adam Ben-Eden @adambeneden]

@FigCyber @shagbark_hick So it’s essentially a statistical artifact?
The original post really makes it sound like every kid there is getting an absurd amount of money. A bit misleading.

[Rob Wiblin @robertwiblin]

Blistering piece from Ezra Klein.

Calls for a comprehensive ban on Recursive Self Improvement starting immediately.

RSI is a grey area, so how do you define it?

Very expansively. Then get firms to apply for narrow exemptions.

YouTube link below.

https://t.co/9rIbNSp7In

[Adam Ben-Eden @adambeneden]

@GregoryConti19 Well first lets try to sensibly regulate AI and stop progress towards misaligned ASI. But beyond that? If we end up in a post work world?
We should strive to improve ourselves and be better towards others. The best runners are slower than cars, but their accomplishments matter.

[Geoffrey Miller @gmiller]

Conservative and Christian influencers:

Over the last week, many of you have been commenting on AI safety issues for the first time. (Welcome!)

But, following Trump, many of you are taking a polarized, partisan position on complex issues that you may not have considered until recently.

You've been given a set of talking points by pro-AI 'accelerationists', Silicon Valley venture capitalists, White House advisors, and other influencers, and you're repeating them avidly.

Trouble is, if you haven't taken any time to understand the key AI safety issues, you'll end up embarrassing yourselves, misleading your followers, advocating for bad policy, and endangering your kids.

AI safety isn't new. Thousands of the smartest, most dedicated people in the world have been working on it for two decades, producing dozens of books, thousands of papers, and thousands of podcasts and interviews. This has been one of the most intellectually intense and fast-moving fields ever developed.

If you're commenting on AI safety issues, you should at least understand some basic concepts such as 'the alignment problem', 'reward hacking', 'instrumental convergence', 'RLHF', 'corrigibility', 'interpretability', 'distribution shift', 'model collapse', etc. These are routinely taught in the hundreds of college courses on AI safety and ethics.

If you don't understand these terms, you can of course react to recent news based on little more than repeating partisan tribalist talking points (e.g. , 'Beat China!', 'Regulatory capture!', 'Accelerate!', 'Climate change 2.0', 'Tell me exactly how swarms of Superintelligent agents that think a thousand times faster than humans, that know everything, and that we can't control, could possibly endanger anyone' -- all of which boil down to 'We must own the libs, even at the cost of endangering our species').

Repeating pro-accelerationist talking points is not an intellectually or ethically serious way to engage with the most important and urgent issue that humanity faces.

Obviously, there's nothing conservative about defending the AI industry's explicit plan to destroy the dignity of work through AI automation, to destroy the sanctity of marriage through AI lovers and sexbots, and to overturn every traditional value through handing our civilization over to alien machines that we can't understand or control.

And obviously, there's nothing Christian about supporting the liberal Bay Area atheists who dominate the AI industry, in their quest to create 'god-like' Superintelligence, to 'summon the demon', and to promote a 'transhumanist' or 'post-human' world.

[Adam Ben-Eden @adambeneden]

@BStizzl @ArtemisConsort If you have the time then by all means, evaluate the arguments yourself! The AI safety people have spent about 20 years thinking through topics like instrumental convergence and meta- vs mesa-alignment and orthogonality. You don't need to take any claim at face value.

[Adam Ben-Eden @adambeneden]

@benshapiro I encourage you to speak with Nate Soares or Daniel Kokotajlo, who went on Tucker Carlson and Joe Rogan's podcasts, respectively, over the last few days. The AI safety issue is serious and there's nothing conservative about letting silicon valley types replace mankind.

[Adam Ben-Eden @adambeneden]

@GWHayduke97 @So8res Keep up the good work man