Lycaos King

That one seemed pretty obvious to me. Angle of the hairline, sharper shadows on the nose to give it a different shape. Smaller eyes and head overall (technically looks a bit larger, but farther away). Eyebrows are larger and rougher. Mouth is more prominent, philtrum is sharper. Angle of the jaw changes.

That's what I got in about 45 seconds of looking it over. It was an interesting exercise. Thanks for sharing that link.

Strictly speaking there is no such thing as "natural selection" or "fitness" or "adaptation" or even "evolution". There are only patterns of physical objects, which increase or decrease in frequency over time in ways that are only loosely modeled by those terms. 

But it's practically impossible to talk about physical systems without fudging a bit of teleology in, so I don't think it's a valid objection.

How would this method distinguish between apparently and actually optimized features? In an evolutionary example, for instance, what's the difference between a bird with:

a large beak that was optimized to consume certain kinds of food

a large beak that was the result of a genetic bottleneck that resulted from a series of accidental deaths culling small beaks from the gene pool (neutral drift).

a large beak that is the result of a single generation mutation that superficially resembles an environmental adaptation, but is, in actuality, unfit.

A large beak that... (read more)

(You're correct. I was using fictionalist in that sense.)

I think the equivocation of "Theorem X is provable from Axiom Set Y" <--> such-and-such thing is Good; would be the part of that chain of reasoning a self-described fictionalist would ascribe fictionality to. 

As I understand it, it's the difference between thinking that Good is a real feature of the universe and Good being a wordgame that we play to make certain ideas easier to work with. Maybe a different example could illuminate some of this.

Fictionalism would be a good tool to describe ... (read more)

Thanks for the reply.

I don't disagree with Eliezer's position for the most part, I just don't see where he lays out a coherent foundation for why he believes certain things about human values. (Or maybe I'm just being uncharitable in my evaluation and not counting some things as "real arguments" that others would.)

By objective and discoverable, I meant in the sense of the values understood in and of themselves without reference to humans in particular. Obviously you can just model human brains and understand what they value, but I meant that you can't lear... (read more)

"Why would any supermind want something so inherently worthless as the feeling of discovery without any real discoveries?"

"No free lunch.  You want a wonderful and mysterious universe?  That's your value."

"These values do not emerge in all possible minds.  They will not appear from nowhere to rebuke and revoke the utility function of an expected paperclip maximizer."

"Touch too hard in the wrong dimension, and the physical representation of those values will shatter - and not come back, for there will be nothing left to want to bring it back.... (read more)

No. Yudkowski is a moral fictionalist, but he has never (to my knowledge) ever justified his position. Granted I haven't read his whole corpus of work, but from what I've seen he just takes it as a given.

Thoughtfulness, pro-sociality, and conscientiousness have no bearing on people's ability to produce aligned AI. 

They do have an effect on people's willingness to not build AI in the first place, but the purpose of working at Meta, OpenAI, and Google is to produce AI. No one who is thoughtful, pro-social, and conscientious is going to decide not to produce AI while working at those companies, while still having a job. 

Hence, the effect of discouraging those sorts of people from working at those companies has no net increase in Pdoom.

If you want to avoid building unaligned AI, you should avoid building AI.

Who says you contribute to the pool at the same rate you'd contribute to your own children? Surely other people in the pool would have different priorities than you, wouldn't they? What if there are N people in the pool and you contribute 1/5N to the children in each pool?

Add that to the fact, that maybe you only have one standout chromosome, and you could easily see a situation where genetic analysis of the population in your family + your pool shows a sudden disappearance of 90% of your genes with a proliferation of 5% of your genes. Is that equivalent t... (read more)

Maximizing the amount of your genetic material in the (near) future is my null hypothesis. I don't think it's totally accurate, but in the absence of a good understanding of which parts of our genetic material produce the non-quantifiable traits we care about: things like the shape of one's smile, personality, taste in food, overall "mood", then I expect people to be reluctant to trade off genetic density at rates greater than ~25-60%

The alternative extreme hypothesis would be a "parent" who wants to maximize their "children's" traits to the point where they'd prefer 0% genetic inheritance if the resultant child would be superior in some respect.

That comparison misses something crucial, which is the density of genetic material passed on. Each generation represents a dilution of the first parent's genetic material with non-kin, but also has the potential for increased numbers of descendants at each generation. By the time your family would be producing your great-grandkids, they'd have the potential to have 2 dozen or more of your direct descendants.

With chromosomal selection you're trading off a massive amount of genetic saturation: essentially getting the percentage genetic inheritance of a great... (read more)

Regarding DragonMagazine: It would often publish content for Dungeons and Dragons that was of a more hurried and slightly lower quality. This led to it being treated as a sort of pseudo 3rd party or beta source of monsters and player options.

People in online communities would frequently talk about options being "from Dragon Magazine" or "Dragon content" in order to forewarn people of content that may not have been given a thorough pass on editing/game balance. As such that phrase was very prevalent in online forums for D&D discussion, which as I understand it, would show up a lot in the training data.

Most people on this website are unaligned.

A lot of the top AI people are very unaligned. 

While it's probably true that copyright/patent/IP law generally in effect helps "preserve the livelihood of intellectual property creators," it's a mistake IMO to see this as more than merely instrumental in preserving incentives for more art/inventions/technology which, but for a temporary monopoly (IP protections), would be financially unprofitable to create. Additionally, this view ignores art consumers, who out-number artists by several orders of magnitude. It seems unfair to orient so much of the discussion of AI art's effects on the smaller group of

Am I the only person who thinks AI art still looks terrible? I see all these posts talking about how amazing AI art is and sharing pictures and they just look...bad? 

Write semi-convincingly from the perspective of a non-mainstream political ideology, religion, philosophy, or aesthetic theory. The token weights are too skewed towards the training data.

This is something I've noticed GPT-3 isn't able to do, after someone pointed out to me that GPT-3 wasn't able to convincingly complete their own sentence prompts because it didn't have that person's philosophy as a background assumption.

I don't know how to put that in terms of numbers, since I couldn't really state the observation in concrete terms either.

When Dath Ilan kicks off their singularity, all the Illuminati factions (keepers, prediction market engineers, secret philosopher kings) who actually run things behind the scenes will murder each other in an orgy of violence, fracturing into tiny subgroups as each of them tries to optimize control over the superintelligence. To do otherwise would be foolish. Binding arbitration cannot hold across a sufficient power/intelligence/resource gap unless submitting to binding arbitration is part of that party's terminal values.

"This is to help you, yes you, stop spinning stories where everyone is competent and things are done for sensible reasons.."

I'll take that and throw it right back your way. You will never be able to predict the actions of authority figures if you assume them to be incompetent instead of malicious. When malice is the best fit curve for the data, you should update your model. The purpose of school shooter interventions is to exercise authority and keep people afraid, not to prevent school shootings. Same for NPIs. Paxlovid is illegal because its legality would result in a decrease in power for authorities.