@VelveteenAmbush's banner p

VelveteenAmbush

Prime Intellect did nothing wrong

3 followers   follows 1 user  
joined 2022 September 05 02:49:35 UTC

				

User ID: 411

VelveteenAmbush

Prime Intellect did nothing wrong

3 followers   follows 1 user   joined 2022 September 05 02:49:35 UTC

					

No bio...


					

User ID: 411

Running circles around a shape-rotator like Sutskever is child's play for him.

Actually, reportedly, it was Anna Brockman crying and begging Sutskever to switch his allegiance that seemed to clinch it. Ilya had officiated Greg and Anna's wedding, held at the OpenAI office. Another point for Hanania's theory that women's tears win in the marketplace of ideas.

A member of your board praising your competitor

Yes, this would be very unusual and blameworthy when "board" means "board of directors of a traditional C Corp." But OpenAI is a nonprofit and this was a nonprofit board. It was set up that way purposefully to allow the directors to slow OpenAI down if they felt it necessary for their mission. I'm glad that Sam prevailed, and I want them to accelerate at least for the time being, but the common assumption that "the board" was supposed to act to further OpenAI's commercial interests (as opposed to its mission) is wrong.

Best since the previous Sam was deposed in the previous November

he had reprimanded Ms. Toner for the paper

Replacing the toner usually fixes any issues you're having with the paper in my experience

that are pledging to leave for Microsoft

Read carefully. The most important word in the letter is "may." Not will.

I think most of the employees are going to stay, Shear will remain CEO, and Sam is going to end up in a small but potent research group in Microsoft. As to how long he'll stay... I can't imagine it will be long, a startup-guy billionaire like him at Microsoft would be like a tree trying to grow at the bottom of a cave.

If you were in the marketing department at AB and someone said “hey, why not send a one-time promotional can to this influencer that she’ll only market to her (highly woke) progressive following and that our core audience will never even hear about?” what would you say to convince them of how badly things would go?

I'd say "we're a lifestyle brand and this would directly contradict our brand image, so it's bad practice, like playing with fire."

Honestly, I don't get it. I empathize with wanting to expand your brand image beyond your core audience, but you grow your territory by expanding it around the periphery -- not by parachuting in on the other side of the world.

Framing this as a big-brained move seems to take more complexity than the obvious explanation: the marketing department was culturally completely disconnected from the core audience and wanted to do something that made them feel good instead of their audience, so they cooked up some arguments to do it.

And some (like me) think that the King has poor taste, admitting mediocrities genuinely unsuited to the responsibilities of nobility who will therefore bend their mediocre talents toward scapegoating society for their mediocrity and attempting to undermine and erode it out of spite.

And what if Abbott escalates back, and blocks the Coast Guard? Escalating is a great move whenever you're sure the other side won't counter-escalate. When you aren't, it's dicey.

Sanctuary state is one thing, “come have your anchor baby in California” advertisements across Central America is another.

They're the same picture.

Most other populations manage to survive without corporate sponsorship.

Corporate sponsorship is all that separates us from malthusianism, under which most extant populations would collapse.

the high comorbidity of psychiatric disorders that does not seem to be abating as societal acceptance improves

You did propose to set this aside, so I acknowledge that I'm not actually disputing anything necessary to your point, but I nonetheless feel obligated to mention that while these comorbidities are associated with gayness, the causality isn't obvious.

Another interesting analogy is left-handedness. Left-handed people are more likely to have mental health comorbidities than right-handed people. But it would be kind of odd to suggest that left-handedness is unhealthy, and certainly odd to suggest that left-handed people should use their right hands to avoid mental illness. Which isn't to suggest that the latter is (analogously) what you're saying about gay people, but in these conversations the label gay or homosexual (among men) seems to vascillate between the inclination and the practice.

I agree with your broader point, but... cheap indoor lighting and washing machines were transformative.

It's the part where the growth stays at the level it's at despite there being no other recent transformative breakthroughs in the past 20 years.

Zoom out to a scale of millennia, using human population as a proxy for wealth. Here's a representative chart (ignore the dotted projections for this exercise). We are on the insane upward spike at the very rightmost edge. What sustains that meteoric rise? The answer is periodic transformations, such as the steam engine, indoor plumbing, electricity, the washing machine, commercial air travel, the personal computer, the internet and the smartphone. Your question seems to assume that this rate of improvement is just background radiation, some sort of fundamental base rate against which innovations should provide additional upside, but in fact it is sustained by these breakthroughs.

Krugman deserves all the scorn that he gets for that quote; the fax machine was useful, but nowhere near on the scale of the internet, and not nearly the same economic engine.

Does a Clinton or Biden really ensure it?

Biden has certainly pulled every lever he could reach to send as much money as he can to universities. His initial student loan bailout had a cost of hundreds of billions before SCOTUS spiked it...

And here's the bailout for news organizations, framed as allowing them to collectively bargain to collude against tech companies.

So that's university jobs and news media jobs...

North Koreans seem worse off than South Africans (than all Southern Africans) on any metric except violent crime rate, for which little data exists but which we can probably assume is lower in the DPRK. The DRC and Liberia have higher GDP/capita by some estimates.

I don't dispute your metrics, but there is a consistent effect where North Koreans who successfully defect to South Korea end up missing their home. Naturally this is blamed on "discrimination" but the defectors I have seen interviewed in the documentaries I've watched reminisce about villages where everyone knew one another and had a defined role in a tight knit community that lasted their whole lives. Not that this excuses the starvation or state-sponsored grotesqueries, but personally I would not find it an easy decision if forced to choose whether to be born into an average life in North Korea or in South Africa.

You could procure the Pope's death, and your own selection in the ensuing Conclave, I suppose. Sounds hard, but coups usually are.

I tend to agree that the old religions have been rendered obsolete, more or less empirically; science has reduced them to Russell's teapot. Some people still believe in them, genuinely, but that's more of a feature of their personal psychological ability to believe things as a result of cultural overhang or because they want them to be true than of any epistemological strength of the belief systems, the latter being similar in some sense to QAnonism.

Science taketh, but science also giveth, and thanks to the empirical advances of machine learning and the retreat of the soul in prevailing theories of cognition, there is now plenty of room for new religions. The Simulation Hypothesis is fertile ground for spiritual entrepreneurs to build neo-gnosticisms. Roko was the John Edwards of Yudkowskianism. Reports abound of the emotional tortures of EA types who have heard his brimstone sermon, and I trust their sincerity. Scott's The Hour I First Believed is a more sophisticated and pro-social synthesis.

Right, if that's the idea, the proponent should do the budget math to figure out what it would cost to raise the entire rest of the world up to first-world standards, because it's obviously a fantasy even if you assume that wealth can be delivered via wire transfer irrespective of the human capital in the recipient country.

There's a lot of superstructure now that is intended to prevent further such changeovers...

Up to what standard? I had understood that residents of impoverished countries actually become more likely to emigrate as the countries become wealthier.

Well, they needed a plaintiff if they wanted the policy change. The absence of live controversies doesn't necessarily indicate a lack of impact. The chilling effect of the threat of prosecution (and its downstream effect on other policy debates) is what they were trying to lift.

Agreed. If we were designing the Constitution from scratch, I think it would actually be kind of difficult to come up with a convincing argument as to why we should have a case and controversy requirement for constitutional impact litigation in particular. The sweep of a SCOTUS decision is so vast in proportion to the individual plaintiffs that I'm not sure what practical effect it has other than to slow SCOTUS down -- but SCOTUS can do that directly via its ripeness doctrine, and it doesn't necessarily constrain SCOTUS if it wants to hasten its review (I believe it could choose to grant cert directly on appeal from a pre-trial motion). For the most part, it's easy enough to find or manufacture a case when there is the will to mount a challenge. There are the occasional scenarios where it isn't clear that anyone could have standing to challenge a particular policy, even in theory, but that strikes me as more of a bug than a feature, insofar as I'm not sure why we should want that particular class of scenarios to be immune to judicial review.

I think the particular cases that end up before SCOTUS are mostly just a quirk of fitting a square peg (Marbury v. Madison) into a round hole (a constitution that purports to forbid advisory opinions), and my baseline assumption is that those specific petitioners/cases are extreme outliers -- either manufactured or highly selected as vehicles -- and unlikely to be representative of the practical effect that the litigation is intended to produce. Such is the nature of systems that are evolved rather than designed, I suppose.

I think LLMs with an "agent wrapper" should be enough. AutoGPT is a primitive wrapper. It can't do much with the LLMs of today, but it wouldn't surprise me if the LLMs of tomorrow OR a more sophisticated wrapper around GPT-4 suffices to bootstrap a generally intelligent agent.

I don't think they'll be able to recursively self improve right away... modern foundation models take a LOT of dollars and time to train, so a "just barely AGI" model won't be able to execute a takeoff. But in principle it could still be a full agentic AGI.

That isn't at all contrary to the claim, regardless of the claim's truth.

"The Facebook algorithm accidentally ordered the genocide of the Rohingya in Burma in order to drive clicks" is sufficiently truth-adjacent that I no longer believe this.

(head cocks)

"Oh, yes, we have a huge team working on AI misinformation and AI racial bias to avoid incidents like that, that is indeed exactly what AI safety means and we are leaders in the field."