Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

The Coming Technological Singularity (1993)

mindstalk.net

Post reply on HN
Print view View original

161–170 of 181 posts

Re: The Coming Technological Singularity (1993)

#161

Post by epcoa »

ben_w wrote: post #82
epcoa wrote: post #79

Earlier quoted context omitted.

Campaign team members and even candidates switch teams often enough. Weak analogy. What is the alignment of human GI, completely generalized?

> What is the alignment of human GI, completely generalized? Of any specific human to any other specific human? https://benwheatley.github.io/blog/2019/05/25-15.09.10.html Of any specific human to a nation? That's the example you replied to. Of all the people of a nation to each other? Best we've done there is what we see in countries in normal times, with all the strife and struggles within. We have yet to fully ext…

I think that's my point. The notion of maintaining an alignment, pro-human or whatever for a replicable general AI, doesn't seem to make sense. The traits of planning, learning and goal setting don't seem concordant with maintaining an alignment. I think this discussion has veered to much to anthrocentrism to be interesting, but alignment however loosely defined here isn't some constant for an individual through their life either. It can be imprecisely manipulated especially in a population by outside forces, but it can't be directly controlled.

Re: The Coming Technological Singularity (1993)

#162

Post by nsbshssh »

RaftPeople wrote: post #147
nsbshssh wrote: post #100

Earlier quoted context omitted.

There must be a name for the human bias that sees what is possible this month and doesn't believe things might be very different very soon and change quickly. AI has a tripple whamy that the models get more efficient, the chips will get faster and there will be more chips. Capitalist and government money pouring in. There is money to be made and it is national security. And to boot the cloud, big tech, cryptocurrency…

> There must be a name for the human bias that sees what is possible this month and doesn't believe things might be very different very soon and change quickly The counter to that is that AI prediction of 10 to 20 years away has been happening since the 1950's (or before?). It seems there is some foundational progress but it's very slow.

Unlike cold fusion there is real practical progress. No grid energy comes frok cold fusion. Millions of people are using AI to help get their work done.

Re: The Coming Technological Singularity (1993)

#163

Post by achierius »

trashtester wrote: post #91
Retric wrote: post #85

Earlier quoted context omitted.

You’re overstating what AlphaFold can do. Without validation the predictions aren’t dependable, but it can cut down on what’s worth investigating. There’s definitely a point where models get good enough you can trust them, but super intelligence doesn’t mean it can both come up with a model and be able to trust it without validation.

The need for validation is itself something that needs to be validated. If models can make predictions that are true >99.99% of the time, validation may only be needed for situations where even 0.01% error rates are intolerable. In any case, even if some validation is needed, AI can speed up this kind of science by at least an order of magnitude.

I think the problem is that it's not anywhere close to 99.99% -- I might be wrong but my understanding is that it's closer to 70%. So even if you can predict with perfect accuracy the cases that it gets wrong, you would still only cut down the validation cost by less than one OoM.

Re: The Coming Technological Singularity (1993)

#164

Post by crackalamoo »

achierius wrote: post #163
trashtester wrote: post #91

Earlier quoted context omitted.

The need for validation is itself something that needs to be validated. If models can make predictions that are true >99.99% of the time, validation may only be needed for situations where even 0.01% error rates are intolerable. In any case, even if some validation is needed, AI can speed up this kind of science by at least an order of magnitude.

I think the problem is that it's not anywhere close to 99.99% -- I might be wrong but my understanding is that it's closer to 70%. So even if you can predict with perfect accuracy the cases that it gets wrong, you would still only cut down the validation cost by less than one OoM.

I'm not exactly sure how you would define 99.99% or 70% correct for protein folding, but AlphaFold is extremely accurate. The original AlphaFold paper had an rmsd of 0.96 Å at 95% residue coverage, which is less than the diameter of a carbon atom (1.4 Å).

Re: The Coming Technological Singularity (1993)

#165

Post by dh77l »

lazide wrote: post #154
dh77l wrote: post #43

Earlier quoted context omitted.

You could ask FDR and Churchill that after the fall of France and it wouldnt be too useful what they said cuz it took them almost 3 years before they openly said victory = end of the nazis and nothing else. So dont just sweep the fact that things take Time under the carpet. Its not healthy cause its like looking at tree shoots in the ground and saying but why does that not look like a tree yet. Finding gold in an une…

What are you saying exactly? The more things change, the more they stay the same?

I am saying AI has capacity to shift the natural balance in the Explore Exploit tradeoff. Human limitations we run into(listed above) lets say might allow 10% explore to 90% exploit on most probs. We might not be able to do more than that. Bacteria for example can adapt (explore) much much faster than us, cause they dont get their genes just from their parent (but from anyone via horizontal gene transfer). AI is similar which cld mean we suddenly start seeing a whole lot more Explore possible/happening than we are used too. Which would reduce the need for/amount of Exploit (baked into everything we do)

Re: The Coming Technological Singularity (1993)

#166

Post by kiba »

ben_w wrote: post #130
kiba wrote: post #129

Earlier quoted context omitted.

Nope. Land is important because everything rest on it. Even radio spectrum and orbitals can be regarded as a form of 'land'. Georgism doesn't exist in a vacuum. It wasn't like they were formulated during when the time when wealth 'meant' land. It was during the industrial revolution, possibly as a response to the problems they see in their society, problems we're still dealing with today. No longer it merely meant la…

> Nope. Land is important because everything rest on it. Even radio spectrum and orbitals can be regarded as a form of 'land'. Then you define land so broadly that the empty vacuum of space, which robots are much better suited to than us, can exploit trivially when we cannot. If you want to, that's fine, but it still doesn't need humans to be able to pay for anything.

The orbital are literally scarce resources, as are radio spectrum. If you have people just doing whatever, you'll get Kessler syndrome, especially as our orbits are filled with more satellites each year. Similarly you just can't have random folks blasting out radio signals at random.

Yes, satellites are robots. However, they have no agency. Incentive structure decides if we have kessler syndrome, which then direct humans to solve problems with robots.

So, yes, they are either directly analogous to or are literal form of land.

Re: The Coming Technological Singularity (1993)

#167

Post by kiba »

trashtester wrote: post #135
ben_w wrote: post #117

Earlier quoted context omitted.

Before the industrial revolution, even though money existed, "wealth" really meant "land" rather than "capital". While we do not today need to ask how people can afford robot lawnmowers despite being unable to find work hitching ploughs to draft horses or oxen, the fears at the time of things like this did lead to mobs smashing looms. If I have some (n) robots that can do any task a human could do, one such task must…

Wealth did tend to mean land if we go back to the middle ages. But wealth above the freeman farmer level also meant access to a workforce capable of working that land and access to (or protection from) a military force capable of defending that land. With capitalism, wealth shifted to controlling "capital", ie the "means of production". Either directly or indirectly by owning money that could (through lending) carry…

I don't know what to tell you. Georgism was contemporary to the industrial revolution. The book Progress and Poverty, as described by wikipedia "investigates the paradox of increasing inequality and poverty amid economic and technological progress"[1]. That sounds especially relevant to our time.

1. https://en.wikipedia.org/wiki/Henry_George

So, yes. Wealth means "land". Especially so in the industrial revolution.

Re: The Coming Technological Singularity (1993)

#168

Post by ben_w »

kiba wrote: post #166
ben_w wrote: post #130

Earlier quoted context omitted.

> Nope. Land is important because everything rest on it. Even radio spectrum and orbitals can be regarded as a form of 'land'. Then you define land so broadly that the empty vacuum of space, which robots are much better suited to than us, can exploit trivially when we cannot. If you want to, that's fine, but it still doesn't need humans to be able to pay for anything.

The orbital are literally scarce resources, as are radio spectrum. If you have people just doing whatever, you'll get Kessler syndrome, especially as our orbits are filled with more satellites each year. Similarly you just can't have random folks blasting out radio signals at random. Yes, satellites are robots. However, they have no agency. Incentive structure decides if we have kessler syndrome, which then direct hu…

Space is much more than circular orbits around earth, and is not a scarce resource — it's big enough that you can disassemble the earth, all the planets, all the stars, all the galaxies into atoms and give them so much padding it would still be considered extraordinarily hard vacuum. Something like 3.5 cubic meters per atom, though at that scale "size" becomes a non-trivial question because the space is expanding.

Which reminds me of a blog post I want to write.

> Similarly you just can't have random folks blasting out radio signals at random.

That's literally what the universe as a whole does.

You may not want it, but can definitely do it.

> Yes, satellites are robots. However, they have no agency.

Given this context is "AI", define "agency" in a way that doesn't exclude the people making the robots and the AI.

> Incentive structure decides if we have kessler syndrome, which then direct humans to solve problems with robots.

Human general problem solving capacites do not extend to small numbers such as merely 7.8e20.

For example, consider the previous example of the moon: if the entire mass is converted into personal robots and we all try to land them, the oceans boil from the heat of all of them performing atmospheric breaking.

And then we all get buried under a several mile thick layer of robots.

This doesn't prevent people from building them. The incentive structures as they currently exist point in that direction, of a Nash equilibrium that sucks.

Humans do not even know how to create an incentive structure sufficient to prevent each other from trading in known carcinogens for personal consumption even when labelled with explicit traumatic surgical intervention images and the words "THIS CAUSES CANCER" in big bold capital letters the outside.

If anyone knew how to do so for AI, the entire question of AI alignment would already be solved.

(Solved at one level, at least: we're still going to have to care about mesa-optimisers because alignment is a game of telephone).

Re: The Coming Technological Singularity (1993)

#169

Post by lazide »

dh77l wrote: post #165
lazide wrote: post #154

Earlier quoted context omitted.

What are you saying exactly? The more things change, the more they stay the same?

I am saying AI has capacity to shift the natural balance in the Explore Exploit tradeoff. Human limitations we run into(listed above) lets say might allow 10% explore to 90% exploit on most probs. We might not be able to do more than that. Bacteria for example can adapt (explore) much much faster than us, cause they dont get their genes just from their parent (but from anyone via horizontal gene transfer). AI is simi…

How would an AI ‘explore’? Does it have hands?

Re: The Coming Technological Singularity (1993)

#170

Post by ben_w »

kiba wrote: post #167
trashtester wrote: post #135

Earlier quoted context omitted.

Wealth did tend to mean land if we go back to the middle ages. But wealth above the freeman farmer level also meant access to a workforce capable of working that land and access to (or protection from) a military force capable of defending that land. With capitalism, wealth shifted to controlling "capital", ie the "means of production". Either directly or indirectly by owning money that could (through lending) carry…

I don't know what to tell you. Georgism was contemporary to the industrial revolution. The book Progress and Poverty , as described by wikipedia "investigates the paradox of increasing inequality and poverty amid economic and technological progress"[1]. That sounds especially relevant to our time. 1. https://en.wikipedia.org/wiki/Henry_George So, yes. Wealth means "land". Especially so in the industrial revolution.

Industrial revolution: roughly 1760 onwards if you count agricultural revolution as part of it.

Middle ages: 500-1500

Henry George, 1839-1897, is indeed part of the industrial revolution. @trashtester and I were both comparing what happened before the industrial revolution to what happened in it.

The valuation of land before still wasn't Georgism: the land was assumed to have productive output, and if someone didn't pay taxes based on the assumption, perhaps you couldn't because the land wasn't that productive, that's their problem.

And you know what else happened in the industrial revolution? Karl Marx and Adam Smith, the former placing workers rather than land at the root, and the latter placing capital rather than land at the root. As with HG, neither liked rent-seekers, and they were both more influential in the "solutions" than Henry George. (Not that he wasn't, they were just more).

Not that Henry George could possibly have foreseen even mere Earth orbitals as "land", let alone disassembling the moon into a swarm of robots that outnumber humans by more than humanity outnumbers a single human and which don't need to land on Earth which is good because if they did we'd all die just from them landing. Can't blame him for that, the difficulty seeing clearly this far ahead is why some call it "the singularity" (though I prefer "event horizon").

Post reply on HN