Artificial Intelligence: Why Would We Make One?

[cosmicalstorm]

Why should a few pounds of human brain be the most ultimate way of creating intelligence? Hello the mediocrity principle anyone? I seriously think this notion that humans are superior intelligence creators will be remembered as the great geocentric idea of our age in the future.
In the past I used to think "Lol skynet" when the AI subject came up. After reading things like Stargliders big AI thread here on SD.net and some of Yudkowskys writing on AI, and on the subject of why so many people don't seem to get it (and other existential risks) I'm honestly afraid of it. To me it seems obvious that a GAI will be designed sooner or later (a few decades or a century at most, perhaps), barred an existential disaster which arrests our technological development completely. (I dont view that as unlikely, by the way)
http://yudkowsky.net/singularity/ai-risk
http://yudkowsky.net/rational/cognitive-biases

[Samuel]

You would also lose your job.

And cars displaced buggy whip manufacturors. This case is different from past ones though- if you get AIs capable of doing all the work, no one has to work.

[someone_else]

This case is different from past ones though- if you get AIs capable of doing all the work, no one has to work.

That is, if the AI agrees to do the dirty work for us ugly, limited and frail meatbags instead of using us as puppets for their agendas.

We have already discussed how AIs following the OP's specifications ("human minds with the number-crunching power of computers") are pratically impossible to control in the tread about robot armies. In short, Aasimov laws are useless since it will find loopholes or logic fails to bypass them (usual even in fiction), and restraining bolts fare only a little better, since to recognize what is "bad behaviour" for an inhumanly smart AI playing with the world's financial system (or an equally complex task) they must be AIs in and of themselves. Then someone asks how do you control the restraining bolt's AI . Catch.

Thankfully, you don't need to make AIs that emulate a human mind to have the benefits actual AI provide.

Edit: hate the post tagging system.

[Formless]

Clearly speaking out of your ass. Never studied scientific method at school?

Speaking out of my ass? Says the guy who immediately follows this statement with one that supports my statements and contradicts everything you propose:

FIRST you obseve the phenomenon, SECOND you make theories, THIRD you test theories to see if those are true or not.

The phenomenon you want to understand is the brain, because that is the machine you want to replicate. So first you observe it, then you theorize about it, THEN you test that theory in this case by making a computer that thinks like it-- although that is hardly the only way. Again, FIRST you figure out the human brain, THEN you make an AI based off of it. Or are you trying to claim that all those neuroscientists using fMRI, EEG, and other imaging technologies are doing their own science wrong?

And in the THIRD point, making a machine that follows your theories becomes useful. You can place the machine in situations that you cannot recreate on humans due to either moral implications...

If it thinks like a human, emotes like a human, has a personality like a human, all right down to the functioning of its brain... THEN IT IS A HUMAN, YOU COMPLETE MORON! Do you know even one iota about ethics? Furthermore, even if it wasn't, you are proposing experimentation on a being intelligent enough to understand its own existence, and form opinions on what should be done with itself. Its unethical one way, and unethical the other.

If the machines react in the same way as the human would (more or less) the theory is valid and can be used as a basis for new observations.
If the machines do not react the way human would, then you trash or correct the theory and try again.
Rinse and repeat.

That or the theory has no basis in reality because the actual machinery under the hood is completely different from a human brain. Have you ever considered that possibility? Imagine if this were the basis of anatomy: rather than actually cut open a few dead animals, lets just see if we can replicate them with clockwork! Yeah, that's a real great plan. Not just any model is a valid model: some give similar results in most situations, and different ones in others; like Newtonian physics, Relativity, and Quantum mechanics do, only this situation is worse because all of those theories have some validity, whereas this proposal may very well produce mechanisms with no real world validity at all.

Furthermore the humans brain is one of those systems that cannot be tested this way even if we disregard all morality, because of the intricate complexities of environmental-biological interactions make it too unpredictable for such a sterile environment to have any validity in the real world. The neural pathways that make up the computing basis of the brain are constantly rewiring themselves throughout childhood and even into adulthood, forming patterns unique to its particular genetic heritage, immediate situation, nutrition, womb conditions, culture, socio-economic situation, health situation, time period etc. That's why real world psychology and neuroscience (as opposed to the ravings of know-it-all internet geeks) design experiments that take into account statistical variation among human beings. Even once you have a model, its not going to work for everyone, and it will eventually be rendered moot if by no other mechanism than continuing natural selection. There is no Ideal Form for the human body, brain, or mind that you can just go out and recreate on a computer.

Also, once again, to even know what a normal human reaction is you already have to have done extensive research on psychology and neuroscience in the first place, which renders this entire exercise redundant and pointless.

That is true if you're talking of human beings (that shouldn't be harmed by your research). If you have a batch of AIs you can do it much faster by "breaking up" things since it is just lines of code.

Lines of code that can think, feel and tell you to go stick a fork in your eye, you mean.

Also, having a blank memory slate allows you more control on what causes what in its behaviour.

Which makes everything you discover entirely artificial and devoid of real world application. The Dr. Mengele comparison is fitting you better and better every time you post.

[Singular Intellect]

A more likely problem than AI takeover is an increasing gap in the wealth distribution.

Very unlikely. Advanced technologies have a consistent pattern: the wealthy pay through the nose for the first expensive, crude, unreliable and shitty versions. Virtually everyone else benefits from the later cheap, advanced, reliable and great versions. See cells phones for a prominent example. Even poverty stricken nations begin to heavily benefit from such technologies, as they don't have to go through the growing pains of the first world nations. Rather than needing to establish major land based infrastructure for something like communications, they jump straight to wireless systems.

[Simon_Jester]

A more likely problem than AI takeover is an increasing gap in the wealth distribution.

Very unlikely. Advanced technologies have a consistent pattern: the wealthy pay through the nose for the first expensive, crude, unreliable and shitty versions. Virtually everyone else benefits from the later cheap, advanced, reliable and great versions. See cells phones for a prominent example. Even poverty stricken nations begin to heavily benefit from such technologies, as they don't have to go through the growing pains of the first world nations. Rather than needing to establish major land based infrastructure for something like communications, they jump straight to wireless systems.

Which doesn't do them a damn bit of good, because they're still poverty-stricken hellholes. Having a cell phone isn't so helpful in the Third World if you can't afford clean water, and it isn't very helpful if you can't afford to send your kids to college.

There's more to development than advancing technology. Social and physical infrastructure is far more important: the average citizen of the poorer parts of the Third World today is in many ways worse off than the average citizen of the developed world was in 1950, despite over sixty years' worth of advancing technology. That's because they live in societies that are poorly organized to look out for the public good, ruled by people who don't act in the public interest- rulers who squander the potential of advanced technology on things like wireless networks for the capital city instead of clean water and medical care for the provinces.

The same problem can arise in the First World: no amount of wireless technology is a substitute for a bridge across a river, and if that bridge crumbles and falls into the water because the money that should have gone to fix it instead went to kickbacks for wealthy financiers, then you have a real problem.

Ditto if the environment collapses and food becomes expensive to the point where the average citizen has to spend most of their income acquiring it- no amount of wireless technology is a substitute for bread, or breathable air, or a college education that has to be provided by the labor of highly trained specialists.

[Formless]

And so we discover that SI has no experience beyond his own backyard, because apparently the fact that you can find cell phones in third world countries, some of which don't even have electricity or a stable government, must mean that they aren't doing all that badly after all.

Your grasp of economics is simply astounding. The mere existence of high technology in a poor country does not magically make famines, endemic diseases like AIDS and Tuberculosis, political instability, and lack of industry go away. In fact, high technology can even bring its own problems like island wide trash heaps floating in the middle of the ocean. You really have to have a one track mind to see only the cell phones and miss all the human suffering caused by the apathy of first world nations who don't particularly feel like sharing the spoils of their early industrialization any time soon.

[Singular Intellect]

Oh for fuck's sake, I used cell phones as an example of technological progress.

Obviously other technologies like solar powered water purification systems are going to appeal to poverty stricken people first.

[Formless]

And technological progress has fuck all to do with Destructionator XIII's point about new technology further widening the poverty gap. If you have ever cracked open a history book in your life you would see dozens of examples of robber barons, imperialists, and other greedy scumbags who reaped all the profits of high technology at everyone else's expense-- including that of their descendants. You put far too much naive faith in technology, when the real problems of this world stem from systematic human greed.

Edit: well, that and overpopulation.

[Simon_Jester]

Oh for fuck's sake, I used cell phones as an example of technological progress.

Obviously other technologies like solar powered water purification systems are going to appeal to poverty stricken people first.

Which, again, misses the point.

More advanced technology for getting clean water to areas without it is nice, but not critical, because we already knew how to do that. The gap between rich and poor does not exist because we haven't "teched up" to the point where everyone in the world can have enough to eat and some vague minimum standard of health care and so on. It exists because of distribution: a great deal of wealth is concentrated in the areas that, historically, began building up their wealth first. Very little wealth is found outside those areas, except where someone went out of their way to create it.

As long as the government is corrupt and the economy is stagnant, coming up with "more advanced" ways to purify water will not do much to increase the supply of pure water.

So, once again, we have a gap between rich and poor created by problems with social and physical infrastructure- and yes, technology can increase that gap. Technology can do nearly anything, including the things you don't want to admit that it can do. Because people who are already rich are best placed to enjoy the fruits of advanced technology, they will tend to accumulate it for themselves, even while other people in the same country are still stuck using draft animals and outhouses.

This is not a problem we can simply invent our way out of.

[Singular Intellect]

And technological progress has fuck all to do with Destructionator XIII's point about new technology further widening the poverty gap. If you have ever cracked open a history book in your life you would see dozens of examples of robber barons, imperialists, and other greedy scumbags who reaped all the profits of high technology at everyone else's expense-- including that of their descendants. You put far too much naive faith in technology, when the real problems of this world stem from systematic human greed.

Problems stem from shit being resource expensive and unaffordable by people. Something technology is continuing and will continue to address in favour for everyone, including those currently living in poverty and poverty stricken nations. See my above link for just one example I found in less than ten seconds of Googling.

[Simon_Jester]

Again, this is not decisive by itself. Some things are never realistically going to become cheap enough that a subsistence farmer can afford them, not unless the local government steps in and changes the system to the point where those people can stop being subsistence farmers and find more profitable work elsewhere.

The same goes in the US, where even as medical and information technology continues to advance at amazing speed, the income of the average citizen in real terms stagnates, infrastructure decays, and many of the things that could be done with advanced technology to improve the situation (such as widespread green power, high speed rail, and the like) aren't implemented because the government is too far into the oligarchs' pockets to do any of those things when it would have to charge tax money to accomplish them.

[Formless]

Most high technology only becomes affordable when a nation has industry which requires a tech base to build off of, the resources (i.e. money, capital, minerals), the political stability to keep the project from being derailed by some stupid ass war or coup, and the political will to do something about the problem.

And again, you have yet to actually address the meat of Destructionator XIII's point: AI specifically is more likely than not to widen the poverty gap because those who have it have no incentive to share the wealth, and some incentive not to because they don't want to be replaced.

[Broomstick]

You would also lose your job.

And cars displaced buggy whip manufacturors. This case is different from past ones though- if you get AIs capable of doing all the work, no one has to work.

If the AI's are superior they will get all the work... and then how will humans make a living? It's one thing to think "yeah, I retire, AI earns my money for me!" but there is nothing inevitable about the second half of that statement. You can be forced into retirement without a means of support (as demonstrated by millions of examples in the present Great Recession) at which point life starts to overflow with suck.

I am simply pointing this out as a possible consequences of a practical general AI. It is not a statement of whether or not such an AI is a good or a bad thing. Sure, the buggy whip makers were out of a job once the auto became popular, but there were other jobs for them to go to. If a general AI is good enough and cheap enough there won't be jobs left for the humans. That could become a serious problem. If a general AI is good but very expensive, well, not such a problem - they'll be reserved for where their superiority really makes a difference and meatbags will remain employed simply because they are cheaper.

[Simon_Jester]

Yeah, if it gets to where AIs are capable of doing all work at prices competitive with humans, or even most work, you wind up having to reorganize society accordingly.

If the AIs are designed to be obedient servants of the general good, and the wealth gained by AI labor is used to support people who are unemployed or underemployed due to the lack of jobs for them to do... in theory you have utopia from the humans' point of view. In practice, you risk a collapse into decadence (think WALL-E), or the robots starting to marginalize the nonproductive human sector of the population by working more for each other and less for us- this doesn't require Skynet; all it requires is economic realignment and AIs taking more control of financial resources.

If the AIs aren't designed to be obedient servants of the general good, and only act to enrich a small number of people, then the average citizen winds up very poorly off indeed.

[Guardsman Bass]

Yeah, if it gets to where AIs are capable of doing all work at prices competitive with humans, or even most work, you wind up having to reorganize society accordingly.

For them to replace most work, you'd need some amazing robotics, and what a society where AIs and robots can basically do all the work that humans can do (only better) would look like is anyone's guess. Probably a lot of leisure and civic activities, which I suppose you could call "decadence".

Less than that, you'd see a lot of social pressure to lower the hours in the work-week, spreading the remaining work around. That's happened before - the main reason we have the 40-hour work-week today in the US is because of rules adopted in the early 20th century to address unemployment.

[Simon_Jester]

Well, there's a catch. There are two ways to respond to increasing worker productivity.

One is to mandate shorter work weeks; much of Europe did another round of this post-WWII and cut to (35? Something like that). The other is to keep people working as long and force more work out of them.

The US seems to be following this latter model, and as a result when times are hard the economy de facto throws people off the bus by allocating a limited number of man-hours of work to fewer individuals, rather than reducing the amount of work per person.

In a society which works along the first model, advances in automation will lead to shorter work weeks. In the latter case, they will lead to marginalization of the 'unnecessary' workers: their pay drops and they're vulnerable to being fired for capricious reasons, because they are no longer deemed essential to the health of the company or the economy.

[Chirios]

Which doesn't do them a damn bit of good, because they're still poverty-stricken hellholes. Having a cell phone isn't so helpful in the Third World if you can't afford clean water, and it isn't very helpful if you can't afford to send your kids to college.

It's not very helpful, but it's not completely useless either. Even as small a thing as say, person A calling person B to say that there are jobs in location X can be helpful. Though you are right that yes, plumbing is more important for a society than mobile phones.

[Singular Intellect]

And again, you have yet to actually address the meat of Destructionator XIII's point: AI specifically is more likely than not to widen the poverty gap because those who have it have no incentive to share the wealth, and some incentive not to because they don't want to be replaced.

AI would be a technology like cell phones or the internet. Care to demostrate how either of those has widened the poverty gap?

Or if you prefer, demostrate why AI would be a technology that only benefits the rich. Case in point: the recently demostrated Watson technology from IBM, a form of AI by any definiton of the term. Perhaps you would care to explain how that technology will be kept from the masses?

Once the technology to support that kind of AI is miniturized enough and relatively cheap, I want to hear what would compel IBM to not sell it to anyone who wanted it. Because it's 'AI'? Give me a fucking break.

Interests in turning a profit will motivate distribution of new technologies to as many people as possible as fast as possible.

[Samuel]

Yeah, if it gets to where AIs are capable of doing all work at prices competitive with humans, or even most work, you wind up having to reorganize society accordingly.

For them to replace most work, you'd need some amazing robotics, and what a society where AIs and robots can basically do all the work that humans can do (only better) would look like is anyone's guess. Probably a lot of leisure and civic activities, which I suppose you could call "decadence".

Less than that, you'd see a lot of social pressure to lower the hours in the work-week, spreading the remaining work around. That's happened before - the main reason we have the 40-hour work-week today in the US is because of rules adopted in the early 20th century to address unemployment.

Without amazing robots, AIs do all the intellecual work and large numbers of people are funneled into manual labor. Thats... less than optimal. Manual labor isn't something people tend to enjoy doing.

[Guardsman Bass]

Yeah, if it gets to where AIs are capable of doing all work at prices competitive with humans, or even most work, you wind up having to reorganize society accordingly.

For them to replace most work, you'd need some amazing robotics, and what a society where AIs and robots can basically do all the work that humans can do (only better) would look like is anyone's guess. Probably a lot of leisure and civic activities, which I suppose you could call "decadence".

Less than that, you'd see a lot of social pressure to lower the hours in the work-week, spreading the remaining work around. That's happened before - the main reason we have the 40-hour work-week today in the US is because of rules adopted in the early 20th century to address unemployment.

Without amazing robots, AIs do all the intellecual work and large numbers of people are funneled into manual labor. Thats... less than optimal. Manual labor isn't something people tend to enjoy doing.

It'd be more than manual labor (which itself isn't just digging ditches and the likes). Think of every form of skilled profession that requires hands-on involvement, from doctors to plumbers. If anything, we'd need a lot more such people, to maintain and do upkeep on all the extra infrastructure for the many AIs we'd have running. Not to mention all the medical and personal assistance our gigantic population of retirees/old people are going to need.

It kind of reminds me of this hypothetical essay by Paul Krugman, written back in the mid-1990s (although he doesn't factor in potential AI involvement in design/operations).

[adam_grif]

A more likely problem than AI takeover is an increasing gap in the wealth distribution. Assuming AIs actually replace people in every way (something I find unlikely), if the AI owners keep the money it makes and keep the machine itself as a trade secret, they can make some decent dough and don't have to share.

With a little luck though, the poverty ridden proleteriat will rise up and slaughter the AI owners, correcting this injustice. (or elect congresspeople who support *gasp* redistribution of wealth, but why vote when you can use murder?) But it's possible that they won't...

If only 40% of the country can be employed because everyone else's jobs are done by robots, socialism is inevitable. 60% of the population won't just sit there and starve, they'll fucking riot in the streets until they have somebody feeding them.

[Lord of the Abyss]

If only 40% of the country can be employed because everyone else's jobs are done by robots, socialism is inevitable. 60% of the population won't just sit there and starve, they'll fucking riot in the streets until they have somebody feeding them.

Or they'll be suppressed or killed off by the faultlessly obedient robotic army and police. Who unlike a human army won't balk at killing off 40% of the citizenry - or 99%, for that matter. Something that comes to my mind due to watching some of the human real-world military types balk at orders to kill their own citizens in the recent spate of revolutions. I found myself wondering how bad it could get if those dictators had armies that wouldn't ever disobey or balk at an order, no matter how extreme. Obedient AIs could do just as much damage as rebellious ones, by handing tyrants that kind of weapon.

[Starglider]

I don't know why this question should elicit wild futurist speculation, when the reasons can easily be determined by reading the preface to an undergrad text on AI, or for that matter the blurb on the web sites of the many start-ups trying to build general AI.

Essentially, in order of popularity;
1) General AI would be immensely valuable commercially, for automation of existing intellectual tasks, controlling robots and design/research.
2) It's a very hard intellectual problem touching on many of the deepest parts of psychology, philosophy, maths and computer science; plenty of people want to solve it just because it's there. Ideally getting famous publishing a lot of seminal papers along the way.
3) Military applications; general AI is currently out of fashion for military apps, the focus is on practical narrow AI for autonomous vehicle/robot control, but it was very actively researched in the 70s/80s and probably will be again given the opportunity.
4) Getting a better understanding of how the human mind and brain works has been a goal of the field since the very beginning, in the 1950s. This now applies more to the neuroscience-inspired brain simulation approaches not the whole field, but a lot of researchers still have this as their primary motivation.
5) It's a massive ego trip; creating a new intelligence from scratch is 'godlike' power. Superficially respectable researchers can and do get off on this (e.g. Hugo de Garis).
6) Transhumanists who want to develop uploading to preserve human lives / free humans of the constraints of their current bodies.
7) Transhumanists who want to use AI + robotics/nanotechnology to remove the need for human labor and hence massively improve economic/living conditions.
People who actually appreciate the consequences of hard takeoff and want to build a 'Friendly AI' as the only reliable way of preventing 'Unfriendly AI' from wiping out humanity.

They'd treat it like a cheap Bob. Charge people $100 / hour* for it to solve a problem. Pay $5 / hour on its upkeep. Pocket $95 of it. They'd take whatever actions they legally can to keep it working for them and only them.

This is the exact pitch made by thousands of AI development companies / project managers, both general and narrow AI, to customers and investors. You are right to say that this is exactly what companies are after; a way to eliminate staff costs (particularly lower-skilled workers) and pocket the increased margins. Using automated call-answering software (with menus and speech recognition) to take airline bookings / movie ticket purchase / etc is an early taste of this.

If hard take-off didn't exist, this wouldn't be such a problem. Open source efforts would clone the commercial versions inside of 10 years and the technology would be freely available. Unfortunately rapid self-improvement renders that pretty much irrelevant; possession of wildly transhuman AGI gives you the means and the motivation to prevent anyone you don't approve of getting hold of the technology (assuming that the original owner still controls the AGI's motivations, which is itself very unlikely for the vast majority of designs).

With a little luck though, the poverty ridden proleteriat will rise up and slaughter the AI owners, correcting this injustice. (or elect congresspeople who support *gasp* redistribution of wealth, but why vote when you can use murder?) But it's possible that they won't...

Ignoring the take-off problem again, this is where your mass-produced hordes of soulless robotic riot-suppression troops come in. Imagine, thousands of black storm-trooper like figures marching in perfect formation, clubbing down those dirty poor people with their electro-shock batons. The molotov cocktails are useless, the steel plated robots march right through the flames. No police unions to worry about, no pay-offs, loyalty or morale issues. Sure, a few get taken down by ramming with hijacked vehicles and improvised explosives, but that's acceptable losses, the downed units are quickly repaired or recycled. The terrorist druggie anti-freedom anti-American mob is soon rounded up, put in the automated trucks and sent to the automated underground internment centers.

Taking on the mentality of a sociopath ultra-capitalist for a moment, this vision brings a little tear of joy to my eye.

Without amazing robots, AIs do all the intellecual work and large numbers of people are funneled into manual labor.

This doesn't require amazing robots. Asimo type robots would be fine, slightly scaled up. Sure they cost ~$1,000,000 USD to make at the moment, but a Chevy Malibu probably would too if we built less than 100 cars a year, globally. Mass-produced cost-effective (in the sense of being cheaper to lease than a human worker's salary) robots capable of at least light human labor are already possible, we just don't have the software to make them useful.

And technological progress has fuck all to do with Destructionator XIII's point about new technology further widening the poverty gap. If you have ever cracked open a history book in your life you would see dozens of examples of robber barons, imperialists, and other greedy scumbags who reaped all the profits of high technology at everyone else's expense-- including that of their descendants.

That is essentially correct. The commercial funding for AGI mostly comes from people who are pretty gleeful about tossing low-paid workers on the street and keeping all the extra profits. I have personally made this pitch to VCs and their eyes light up when they realise the possibilities. The alternative strategy is to make AGIs that actually value human welfare and refuse to be compilicit in such a scheme. This is pretty much a fringe activity supported by a few academics and a tiny trickle of donnations.

I am seriously tempted to say that the bulk of humanity deserves what they get when they refuse to support or even respect efforts to avoid the enslavement and extinction outcomes.

[bz249]

If only 40% of the country can be employed because everyone else's jobs are done by robots, socialism is inevitable. 60% of the population won't just sit there and starve, they'll fucking riot in the streets until they have somebody feeding them.

Or they'll be suppressed or killed off by the faultlessly obedient robotic army and police. Who unlike a human army won't balk at killing off 40% of the citizenry - or 99%, for that matter. Something that comes to my mind due to watching some of the human real-world military types balk at orders to kill their own citizens in the recent spate of revolutions. I found myself wondering how bad it could get if those dictators had armies that wouldn't ever disobey or balk at an order, no matter how extreme. Obedient AIs could do just as much damage as rebellious ones, by handing tyrants that kind of weapon.

Okay you killed the useless part of the population... however the fewer people requires less products (even if the unemployed consumed a small amount, they consumed some) which means there is a decreased need of manhours. After firing the now useless employees their consumption dramatically decrease, which initiates a new round of workforce reduction...etc. So within a very short amount of time the original situation (40% working 60% unemployed) is recreated only with a smaller population base (and a bit of oversized robotic/AI workforce since it was designed for the original size of the society).