Quantumfreakonomics
No bio...
User ID: 324
Quantumfreakonomics
3hr ago
The discourse didn't make sense to me until I realized that the pro-acceleration+anti-Anthropic faction consists entirely of venture capitalists who want unregulated open weight models so they can use them to found B2B SaaS startups. Just a complete failure to grasp the implications of the situation we are in. If they fully understood what it means for Anthropic (or OpenAI) to solve alignment and build superintelligence then they would be the ones begging the government for a slowdown.
Quantumfreakonomics
2d ago
A few clusters off the top of my head that hate each other to varying degrees:
-
The Emily M. Bender/Timnit Gebru woke academic cluster.
-
The Coefficient Giving(aka Open Philanthropy) mainstream EA cluster.
-
The MIRI/Lightcone Infrastructure cluster.
-
Current OpenAI employees.
-
Current Anthropic employees.
Quantumfreakonomics
2d ago
Depends how far we are into AI integration into the economy. If the weights have been exfiltrated, we'd have to shut down the whole internet until we can figure out what is going on. If the AI controls its own supply chain, then it's straight-up Terminator 2 and we lose.
Quantumfreakonomics
2d ago
The simplest explaination is that the speakers agree with Osama Bin Laden's overall worldview but think that his tactics aren't very productive (obviously they can't admit this outright).
Like, imagine yourself in the mindset of an open-minded devout young Muslim man and read Bin Laden's Letter to the American People. Your response is probably something like, "well I am not about to fly a plane into a building, but he makes some good points." He's worried about the same things you are: Israel, secularization, sexual debauchery, consumerism, the plight of the Ummah. When you think about it, the tragedy of 9/11 isn't really Islam's fault, because America started it.
Quantumfreakonomics
2d ago
Coefficient Giving just launched a new funding project for AI Safety orgs. They are currently begging for applications. Shoot your shot.
Quantumfreakonomics
2d ago
Suppose that AI risk is a hoax. Suppose that the alignment problem is easy.
Then Anthropic plows ahead, builds AGI, and takes over the world. Despite not falling for the Democrat “hoax”, we still end up with a Democrat one-world government.
The only winning move for Trump is to stop Anthropic from doing frontier AI research and/or nationalize the labs.
Quantumfreakonomics
2d ago
Allegedly, one of the ways OpenAI and Anthropic poach talent is to offer top researchers unlimited tokens to use on their unreleased frontier models.
Quantumfreakonomics
2d ago
Yes. Amazingly, this fact alone makes them better than every other AI organization in existence (with the possible exception of Redwood Research and METR).
Quantumfreakonomics
2d ago
The number 1 rule of AI safety is don’t build the AI that takes over the world. The fact that Anthropic and OpenAI think that the best way to do this is to build superintelligent AI and take over the world has much more to do with Silicon Valley startup culture and utilitarianism than AI safety.
MIRI seems to be the only AI organization that understood the assignment.
Quantumfreakonomics
2d ago
Right now Trump is reTruthing Fox News clips framing AI safety as a global one world government Democrat control-seeking scheme
Oh my God, we are in the stupidest timeline. Mr. President sir, AI safety isn’t a global one-world government Democrat control-seeking scheme, Anthropic is a global one-world government Democrat control-seeking scheme. That is the thing that they are building the AI to do. Your job is to stop them so that they don’t take over the world, or worse. They are willing to risk the lives of every single person on Earth if it increases the chance of their global AI cult ruling what used to be human society.
Quantumfreakonomics
4d ago
I strongly suspect that network effects are ≈ the only thing that matters about a dating app. Your best bet is using Hinge and looking for subtle class clues.
Quantumfreakonomics
4d ago
It's pretty impressive how the Catholic Church was able to survive as an institution for almost 2000 years, far longer than anything else currently extant in the Western world.
That's about all I can come up with.
Quantumfreakonomics
5d ago
Annals of Great Friday-at-6:00PM Announcements: "Caroline Ellison Has Joined Manifund."
For those needing a refresher, Caroline Ellison was Sam Bankman-Fried's lieutenant in charge at Alameda Research, the hedge fund he used to embezzle FTX customer funds. She pleaded guilty to felony wire fraud and conspiracy charges. Manifund is an effective altruism grantmaking and infrastructure provider spun-off from Manifold Markets.
I think this is definitive evidence that no, the Effective Altruism community did not properly internalize the lessons from the FTX crisis. This is going to get a lot of bad press, and it will be easy to criticize the decision as bad optics, but this is not simply "bad optics". Aella's Twitter feed is bad optics. This is bad practice. It is bad practice to hire people with bad decisionmaking records for decisionmaking roles. It is bad practice to establish a norm of soft-landings for people culpable in major scandals.
I had sort of expected Caroline to remain a semi-public figure associated with EA/AI/rationalism after her release from prison, but I didn't expect her to get hired for a financial infrastructure job. She should be blogging for pennies on Substack or working in the bednet mines.
Quantumfreakonomics
6d ago
I have read offical Iraq War after-action reports that basically concede that the United States would not have invaded Iraq if they hadn't been the only country in the world to not send official condolences after 9-11.
Quantumfreakonomics
6d ago
It is most likely not misconduct at all, and certainly not serious misconduct. The jury instructions include, "when deciding whether to believe a witness and how much importance to give a witness's testimony, you must look at all the evidence, drawing on your own common sense and experience of life." If someone is a nurse, and they think that gives them special insight as to which side's psychiatric expert to trust, that's fair game.
Quantumfreakonomics
6d ago
The core physical idea is:
59Co + neutron flux + energy = vaporized 60Co
Cobalt 59 and energy are dirt cheap compared to neutron flux. Cobalt 60 can be manufactured beforehand by placing 59Co in a nuclear reactor for a few months, or it can be generated on-site by surrounding a nuclear warhead with cobalt 59.
Quantumfreakonomics
6d ago
I mostly stopped using AI to generate humorous images after they finally fixed the "generate a cartoon sponge character" loophole.
I don't know. "High-end escorts fuck him for free," doesn't exactly make Nate look bad.
More options
Context Copy link