The Singularity, and Marxism

[BryanM]

I rarely start my own threads - a post from me is like the touch of death to a thread. But I suppose I do want to talk about this some.

I'll briefly talk about Marxism, as it is tied into our near future.

Not everything the man said was correct, no one bats a thousand after all. And "correct" is such a loaded term, we only strive to be less wrong. But his basic idea was that the development of technology inextricably creates tension between the economy and the current social structure. The phases as he saw them:

"1. Primitive Communism: as in co-operative tribal societies.
2. Slave Society: a development of tribal progression to city-state; Aristocracy is born.
3. Feudalism: aristocrats are the ruling class; merchants evolve into capitalists.
4. Capitalism: capitalists are the ruling class, who create and employ the proletariat.
5. Socialism: workers gain class consciousness, and via proletarian revolution depose the capitalist dictatorship of the bourgeoisie, replacing it in turn with dictatorship of the proletariat through which the socialization of the means of production can be realized.
6. Communism: a classless and stateless society."

World War II brought us into the earliest phases of Socialism. A true socialist state is realized only once everyone is given a salary just for being alive (and not just for pooping out babies).

Jobism is the idea that jobs are the solution to every problem, every where, under every situation. But this willfully ignores outsourcing, and automation. A farmer used to have to work from sun rise to sun set to feed his family and maybe make something extra for the wife and kids. Nowadays, a man with a combine harvester can feed an entire state. Combined with advances in medicine in just the last couple centuries (less than 0.25% of the time human beings have existed), we literally are living in a near-utopian age of miracles right now. We should appreciate this more.

Some folks may take examples like Wilson crushing the all-time Jeopardy champions, or Google's autopiloted cars, as a mild "curiosity". But you can draw a straight line from them, to tech support guys/librarians, post office workers, and truck drivers being made redundant. And that's a good thing, it is beneath a human to do the work a machine can do. Their adoption (and the electric car) will not begin to happen as a gradual process, but will be a sudden chop, as how it was with cell phones (those haunted bricks with whispering ghosts inside) and dvds. When it's ready, it's go time.

"It sounds kind of like sci-fi stuff to me."
"You live in a space ship, dear."

Glancing back in the health care thread, I was reminded of how, if human beings tried really hard under perfect conditions to get it on and multiply, they could convert most of the matter in the universe into human beings (a superblob, if you will). That is a good place to move onto the potential of true AI.

Now, the argument that singularitarians make is thus: Once you have a machine capable of the intelligent work of a human, it can participate, and take over, its own growth. Just looking at it from a hardware side, is the doubling of computational power. In ~18 months, it doubles in speed. Then, since it can work twice as hard, it doubles again in 9 months, 4.5 months, 2.25 months, 1.175 months, and so on until it can work out ways to double the power of its hardware in less than an hour. By this point, is when the ability for a human being to even guess what technology may be possible falls short of the advances that will happen. This is the "singularity". And the process, once underway, would be realized in under ~4 years.

In a respect, AI machines manufacture time, our most precious resource. But also, the very processes of intelligence (and the internal models the machine uses to model the real world) would be under constant improvement. A guy who makes Hawking look like a drooling moron can accomplish more in the same timeframe than the average bear, just as a fish would never invent a rocket submarine to fight off sharks.

So, a single machine sitting around would be able to accomplish more in a week, than all of humanity colonizing every planet and working together in perfect sync until the heat death of the universe, would be able to accomplish.

The implications of what is possible through this can only be barely scratched upon: a cure for aging, and being able to make delicious lion steaks out of our poop are clearly in reach. But "more powerful than a boundless human empire"? That gets into the spooky realm of ontotechnolgy - shit that has no right to be possible. Localized changes in the gravitational constant, reversing or slowing down entropy, etc. That's the kind of being intelligent design may produce.

"Everyone complains about the laws of physics, but no one ever does anything about them."

As much as I like to cluck about politics, real social change will be brought by technology, and an ever rising unemployment number. It may be a crackpot religion, but it's the only one we have.

[Skykid]

I completely agree with you, and so does Jacques Fresco incidentally.

[ancestral-knowledge]

i agree.
I think a quote in a book of kurzweil states that all humankind has to do is to built a machine/AI that is capable to recursively improve itself. If this could be accomplished the machine would constantly check better ways and methods of solving problems. This would be in fact the kickoff to the concept of a state called singularity. After that humankind would just sit there and watch.

The concept of singularity is quite hard to grasp... in my opinion it even wouldn't be necessary to develop a cure for cancer and the likes because we would live in the net. The memories and everything that forms the self consciousness could be stored and transfered somewhere on a machine or whatever.... a physical body would not be needed anymore. You can live forever as an entity in the net.

[Randorama]

Massimo Pigliucci begs to differ on Kurzweil, and I happen to agree with him. I'd add that I can't fathom how anyone but hippy baby-boomers, who had it extremely easy by being born right after a war and have *no fucking clue* on basic bits of science, can buy his arguments. Sorry, but I'd rather be rude than being hypocrite.

On Marx, I think that he falls within the long list of 19th-century thinkers that made such broad and sweeping statements that they could be hardly false (or true, for that matter), but nevertheless he dared to propose ideas and try to use their minds, and gave us a lot of food for thought.

Some of the aberrations that stemmed from his work is something that makes me want to invoke eugenics (say, Soviet Union). Lakatos wrote an excellent explanation on why Soviet communism and other similar monsters were non-sense (in proofs and refutations). Of course, I would not blame Marx per se: in life, he loathed his devotees.

Nevertheless,

As much as I like to cluck about politics, real social change will be brought by technology, and an ever rising unemployment number.

Yes, that's very likely, in fact it has been happening for 4 decades at least.

[ancestral-knowledge]

I don't know what you have against kurzweil but i only said that the thought appeared in one of his books (Singularity is near) and now that i looked up the page it's not even from him but it's a quote.... so what now? Even i think that some of his predictions when certain things will happen are quite stupid.

And the soviet system and other various self-called "communist/socialist" regimes (NKorea for example) have absolutely nothing to do with the idea of communism/socialism/marxism. Nothing. They only take that title.

[Blackbird]

As much good as strong AI might do for humanity, I somewhat dread it's creation.

Let's think for a moment about how humans treat lifeforms that are inferior to them. You might like your pet, but it's essentially dependent on your mercy for it's survival. You provide all the means (food, etc.) for it to live. If you decided to stop providing for it, it would starve. Additionally, if it came down to a choice between the survival of a person and the survival of an animal, I am sure that most people would sacrifice the animal. The difference between humans and insects is even greater. Do you give any respect to the intelligence of ants or bees? Do their lives mean anything to you? No, a person is so much more advanced than an ant that we are barely even aware of their existence - the concept that an insect life has any value is absurd to most humans. The only time we pay any regard to insects is when they have become a nuisance, and then we simply destroy them.

Taking the preceding paragraph as a metaphor, what happens when our strong AI supercomputer begins to think of itself as a "human", (a lifeform with superior intelligence and value relative to other lifeforms) and the lifeforms it supports as animals? What if humanity becomes as insignificant as an ant compared to the AI? If I were that AI, wouldn't I think about eliminating elements that reduced the efficiency of the system, like those insignificant ants? We would be forfeiting responsibility for our own survival and giving it to another entity - one step short of handing someone the gun to kill you.

[BryanM]

I'm mortally offended you spelled my name with an I. I demand seppuku right here, right now.

Massimo Pigliucci begs to differ on Kurzweil, and I happen to agree with him. I'd add that I can't fathom how anyone but hippy baby-boomers, who had it extremely easy by being born right after a war and have *no fucking clue* on basic bits of science, can buy his arguments. Sorry, but I'd rather be rude than being hypocrite.

Kurzweil specifically really has some creepy metal fetish or something. Like Tetsuo the Iron Man. Even the sex-and-love-with-robots guy insists on them (and us) looking like people!

The mention of qualia always makes me think of this. Philosophy!

Arguments over possible speculative outcomes are kind of roundabout - the issue is to what magnitude AI could return on its investment. I suppose there would be a few levels of futurism:

* It will be able to do diddly squat.

* It will be able to maintain basic infrastructure and perform basic labor. I view this as mildly pessimistic, but certainly still desirable. Then again, as you say, I just want fucking free burritos.

* It will be able to accelerate basic research into feasible technologies, including its self.

* It will be able to have what we would consider godlike power.

Maybe there's a hard cap, or maybe that's a list of checkboxes comparable to Marx's, whose feasibility can't be taken seriously until we begin to advance through it.

[moozooh]

As much as I like the concept of technological singularity and everything associated with the topic, it's not so simple.

An AI can figure out a way to improve itself, sure (actually genetic algos already do this without being self-aware, much like your nervous system), but it will still be bound by the computational ability of the host system. For instance, a processor running at 3 GHz can only execute simple commands three billion times per second, and laws of physics are directly involved with the number. An algorithm that has taken full advantage of the host system's hardware is thus bound by the hardware. Solution? Build new host hardware and transplant the software. This, again, will be a multi-step process, considerably slow by itself, even if fully automated (manufacturing, testing, etc.), that will eventually take a hit from the law of diminishing returns and once again arrive at the universal caps of the physical world: the speed of light (the fastest way to transport information), and the size of atom nuclei (the smallest controllable persistent particle that can store and carry an electrical charge around).

Funnily enough, we are already coming pretty close to those limits with <20 nm manufacturing processes, so we have to parallelize computing to cope with them, but that is again a finite route as the cores need to be synchronized faster than the oscillator starts a new cycle, which imposes a restriction on core size, distance between cores, and maximum clock frequency. Did I tell you the systems become less and less robust to interference and random errors when approaching fundamental size constraints yet?

Cutting the long story short, you won't see an infinite intelligence growth. You won't see a situation where you fall asleep one day and wake up in a drastically changed world, either, unless somebody decides to start a new world war whilst you're sleeping (and I'm pretty sure it won't be a machine).

[Aguraki]

In a respect, AI machines manufacture time, our most precious resource.

this is so short sighted humanism optimism.
I reject notions like progress and accomplishment.
Why anyone think living longer=better always amazed me(no emo).
Human kind is lost from the start and I hope it disappear quickly to be honest.
We are very messy and machines do zero good to us,as well as rationnality and consciousness.

Ps:this post could go in the life itself thread too but we don't like negative talks usually

[BryanM]

Cutting the long story short, you won't see an infinite intelligence growth. You won't see a situation where you fall asleep one day and wake up in a drastically changed world, either, unless somebody decides to start a new world war whilst you're sleeping (and I'm pretty sure it won't be a machine).

Certainly there are hard barriers, but how much they can be sidestepped by custom architecture and parallelism, and how much can you get out of raw exaflops? How much of an improvement (if any) would optical transistors bring us?

But I would still say that the human brain isn't a divine piece of crumbly meat. How much our tiny brains with their crude chemical signals can accomplish, what a team of specialists devoted to one endeavor can accomplish, surely something at least as good as that is feasible.

A universe where I have a computer to finally make me some new episodes of The Simpsons... well, waking up every day to a new Simpsons episode is a drastically changed world, every day, indeed, no?

[Aguraki]

A universe where I have a computer to finally make me some new episodes of The Simpsons... well, waking up every day to a new Simpsons episode is a drastically changed world, every day, indeed, no?

[/quote]
yes this would be a very sad world to live in.

edit:srry moozooh fixed

[moozooh]

Aguraki, could you please fix your quote?

Certainly there are hard barriers, but how much they can be sidestepped by custom architecture and parallelism, and how much can you get out of raw exaflops? How much of an improvement (if any) would optical transistors bring us?

Hard barriers aren't really side-steppable because they are cosmological constants, and technology is something that happens here in this world, still. Electrical charge needs to be able to travel from point A to point B in time, and that by itself takes time. You can't really compress time or accelerate a charge-carrying particle past the speed of light, at least not in a way that makes sense for computations. You also need to battle thermal expansion, something that gets critical when a general distance of particle movement goes on the same order of magnitude with distance of particle displacement.

A quantum computer would reduce the time needed to solve a task by approximately a quadratic root compared to conventional algorithms, depending on particular task and possible improvements; otherwise it is bound by the same constants as the usual hardware (except it's even less robust and is harder to manufacture en masse due to the complexity involved). But in terms of raw FLOP number... well, let's say you won't ever be able to brute-force a round of Go, because the game tree complexity by far exceeds the amount of matter in the universe, and is an astronomical number. See for yourself:
http://en.wikipedia.org/wiki/Go_and_mathematics
http://en.wikipedia.org/wiki/Observable ... er_content
In fact you won't even be able to brute-force a Super Mario Bros TAS. A pity, eh?

On a lighter note, you're severely downplaying the capabilities of the human brain! Brain is unique in the respect that it can not only reprogram itself even without external input (that's how you can learn things in your sleep, for one), but also change its physical structure while doing so: grow in size at any age, increase the amount of neurons, and modify synaptic connections routinely. We aren't going to see feats like that in machines, ever, because their physical structure is largely fixed (parts need to be removed and swapped part by part, until the whole system is changed completely), and unlike a living organism it can only deteriorate with time. The amount of possible synaptic connections in a single brain is also astronomical, btw, although without a proper direct-to-brain interface we are never going to be able to take full advantage of that computational power: enough of it is being spent on processing gigabytes of raw data coming from our sensory organs and internal systems, upkeeping cognitive abilities, and doing routine body maintenance tasks.

Waking up every day to a new Simpsons episode, though... well, shit, that's drastic alright. Though I guess you could achieve stuff like that by prolonged sensory deprivation possibly coupled with hallucinogenic substances without having to wait for any kind of singularity to happen.

[BulletMagnet]

it is beneath a human to do the work a machine can do.

I don't have time to go into detail atm, but I have to disagree with this.

[Ex-Cyber]

Some folks may take examples like Wilson crushing the all-time Jeopardy champions, or Google's autopiloted cars, as a mild "curiosity". But you can draw a straight line from them, to tech support guys/librarians, post office workers, and truck drivers being made redundant.

The problem with the singularity is that you can't draw a straight line from Watson to Watson 2 to a Watson N that can design Watson N+1. Fundamentally, nobody knows how to do that; it's not simply a matter of hardware not being big enough or fast enough. Even if it were, it's likely that we'll see some kind of lull between the Moore's law progression for ICs tapering off (which has arguably already started) and the ramp-up of whatever replaces it, because all indications are that it will need to be radically different.

That being said, I think the biggest obstacles to any kind of radical advancement will be cultural and political, not technological.

[GaijinPunch]

Good post. I recall talking about it in the "Intelligent Life" thread a while back. While obviously such things are hard to fathom, most things involving technology are only impossible until they are not. So, saying it can't happen is totally fucking gay. There are a few hurdles that would have to be overcome: specifically physical constants which at the moment, seem to be unbreakable (speed of light, etc). The more theoretical one that sticks out to me, is that (at least now) the computers are only as smart as humans that developed them. This opens up all sorts of potholes and short-comings considering how stupid people are. What's to say they don't destroy themselves even quicker than we will?

I'm more curious in the immediate effects of computers taking over. Perhaps technology (not in the bomb sense) is the next event that will lower the population to a healthy level.

We aren't going to see feats like that in machines, ever, because their physical structure is largely fixed (parts need to be removed and swapped part by part, until the whole system is changed completely),

Uh... hello? Cyborgs? Haha. Why can't I name something famous w/ cyborgs in it.

[moozooh]

Cyborgs don't grow or change cybernetic parts, so they still need to be reconstructed to improve their physique. :P

[GaijinPunch]

Cyborgs don't grow or change cybernetic parts, so they still need to be reconstructed to improve their physique.

Not the one I'm working on!