Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[SirQuiamus]

Also the perspective that "if each scientist has a computer than can do X it's good enough" fails to realize that it's our existing social structures that may be holding us back.

Do you consider human social structures immutable? The internet has already changed them a lot, and we ain't seen nothin' yet.

[Reelya]

Doesn't that sort of feed into the point however?

I was pointing out that your "every scientist (individually) with advanced software" model is missing the point. Less and less is going to about some individual person having some insight and more and more is going to be about building systems where no one really knows the internals. We're already seeing a lot of this. Not only are there systems where no person understands everything, there are more and more black-box systems designed by a machine that nobody really knows why they work. It's not just that nobody has taken the time to check them out, these systems have so many variables it's impossible to comprehend everything in there.

It's sort of a religious belief that nothing is too complex for one human to understand, but that's plain bullshit. Dynamic systems with thousands of units, thousands of states per unit and complex rules of interoperation exist, and even experts are often stumped by what comes out of them. These things are optimized by machine learning and nobody really understands what they are doing internally.

[Angle]

So... We know what the problem is: Global warming. We know what causes it: Greenhouse gas. We know how to fix the problem: move away from fossil fuels. We also know what prevent us from fixing the problem: market interests. Now, how to we overcome this? There must be a way to make the protection of our species and environment more important than short term profit.

Not just market interests, there's a healthy dose of human stupidity - "I refuse to admit this is happening, or even consider the evidence, and even if it is happening, I refuse to do even so little as send my politicians a letter over it." We're getting better about it, but it's precious slow. The internet seems to be speeding it up a little, but people are still close minded foolish creatures.

In the past, changes of opinions like that would work on several generation, as each new one was more used to seeing whatever was new than the last one. We know that we don't have the time to wait for the previous generations to die, so we must find a way to change the opinion of the population. If we look in the past, we can see a few cases where there was a relatively quick change of general opinion, such as what we see in periods of war when people are influenced by massive propaganda campaigns. So that seems to be our best bet. The internet can be a good tool to spread the message, but I suspect that a good chunk of those who are in denial don't use it. We'd need to gain access to other medias, such as the television and newspapers, but those are owned by private interests, so we need to make it so that they believe that being in favor of global warming denial is less profitable than the opposite.

Yeah, that sounds like a good start. I don't think it's quite enough, though. Remember, there are dozens of these problems, just about any of which has the potential to be ruinously painful - Global Warming, plastic pollution, biodiversity loss, militant extremism, etc, etc. So we need a way to tackle ALL of these problems, while considering their effects on each other. Obviously, merely pushing ideas about in the newspaper won't do it, though it might be an important part of the solution. That's why (Here we go again!) I'm working on Agora. It's not just a method of communicating, it's a way of harnessing our collective brainpower to tackle problems we couldn't even begin to solve individually. In a way, I hope for it to be like the Super Intelligent AI we were talking about, except that it's made out of people.

And yes, I know that sounds crazy, but just you wait and see! I'll show you! I'll show you all! /humor 

Edit: I'll also point out, do you have the money for said media blitz? Do you know where to get the money? Do you know haw to target your propaganda for sufficient effect? Do you know where to find the expertise to target for said effect? etc, etc?

[SirQuiamus]

Doesn't that sort of feed into the point however?

I was pointing out that your "every scientist (individually) with advanced software" model is missing the point. Less and less is going to about some individual person having some insight and more and more is going to be about building systems where no one really knows the internals. We're already seeing a lot of this. Not only are there systems where no person understands everything, there are more and more black-box systems designed by a machine that nobody really knows why they work. It's not just that nobody has taken the time to check them out, these systems have so many variables it's impossible to comprehend everything in there.

It's sort of a religious belief that nothing is too complex for one human to understand, but that's plain bullshit. Dynamic systems with thousands of units, thousands of states per unit and complex rules of interoperation exist, and even experts are often stumped by what comes out of them. These things are optimized by machine learning and nobody really understands what they are doing internally.

Society is all about implementation, not design; and societal design-flaws are caused by insufficient data, not deficient intelligence. And what is a counter-intuitive solution to a social problem, anyway? If Multivac tells us that the optimal, utterly counter-intuitive solution is to SMASH CAPITALISM, will people believe it? If, on the other hand, Multivac tells us that the optimal, utterly counter-intuitive solution is to KILL EVERYONE, would we be willing to implement that plan? No doubt sociologists would be better off using more Big Data and neural networks in their research, but why would they listen to Multivac?

FAKEDIT: A black-box society that humans cannot comprehend? That's starting to sound downright scary... 

EDIT: I'm sure you've read The Machine Stops by E.M. Forster?   

[Angle]

FAKEDIT: A black-box society that humans cannot comprehend? That's starting to sound downright scary...     

Society basically already is that for anyone who doesn't spend loads of tie studying it, and most people deal just fine.

[Reelya]

If you're using that story as an argument against AI, it's also an argument against cars, the internet or any other machine. What if the lightbulbs all stop working? How will we see? Better keep your old oil-wick lantern around just in case.

And you're completely relying on straw man catastrophic scenarios.

What is the actual case is that AI comes up with solutions we never thought up, and could never think up without the AI, like NASAs super antenna. We then apply these things which just work, without needing to know the exact principles, because they're clearly better at doing what we wanted them to do than the ones we designed. AIs are already better at designing simple things where there aren't many variables than we are. But humans are better at designing just this class of simple systems than we are at designing complex systems. So, in the long run, it's fairly certain that AIs will also be better at designing complex systems than we are.

And being designed by a machine doesn't mean these systems are more prone to error than human systems. for example I read about neural-network designed packet switching networks, and they don't suffer from the type of fatal gridlock, meltdowns and edge cases that the best human designed ones do. Because humans designed them from understanding the behaviour of one node, then extending the understanding of one node to multiple nodes interacting. Chaos theory however kicks in and makes these reductionist observations start to fail badly at some number of nodes, and under dynamic conditions which are almost impossible to visualize. The AI ones were able to apply some weird meta-node formulas to keep things running smoothly under a much wide range of conditions.

[SirQuiamus]

If you're using that story as an argument against AI, it's also an argument against cars, the internet or any other machine.

It would be silly to use fiction as an argument against anything. I'm simply pointing out that one person's utopia may be another's dystopia, and neither of them is objectively right.

FAKEDIT: A black-box society that humans cannot comprehend? That's starting to sound downright scary...     

Society basically already is that for anyone who doesn't spend loads of tie studying it, and most people deal just fine.

We can easily understand everything we know about society, but we simply cannot know everything. The AI would have to be omniscient as well as super-intelligent to devise the optimal society.


EDIT: fucked up quotes fixed.

[Angle]

Well, we should obviously make backups and all that, but I'm not too worried.

[Reelya]

> The AI would have to be omniscient as well as super-intelligent to devise the optimal society.

This is pure conjecture.

1) nothing needs to be omniscient, it just needs the same data we already have. AI has already proven better at synthesizing complex data than humans in many cases. So even existing shitty AI can improve on many things given just the data we have.

2) Neural networks and genetic algorithms for example don't contain any intelligence at all, yet they solve many problems we have not been able to.

Also "optimal society" isn't right. Sure, a "perfect solution" needs perfect data, but we don't need it perfect, just better.

And our current stuff is far from optimal. Back in Taylor's day factory owners probably thought they were the epitomy of efficiency. Yet Taylor just applies some numerical comparisons, and increases efficiency by orders of magnitude, without needing to make people work any harder. Taylors reductionist system, breaking down tasks into steps, seeing who was fastest at each step, them reassembling the steps into a new process. That's basically AI (an early genetic algorithm basically, since people will adapt each "best step" further by themselves and you can do multiple iterations of the Taylor process), the data analysis determined the design of the new process, not a person's judgement. The existing "experts" - people who did those tasks day in day out professionally for years were beaten by maths. This lead to a quantitative change - higher output, but the long term effects of that are qualitative change, changes to the way of life itself, since new products become commerically viable for the general population due to higher productivity that would have been impossible before, or have been luxuries only for the rich before that.

I think we're at a similar situation to the 19th century before Taylorism, where people think we have things pretty nailed down so computation couldn't come up with anything significantly better. But most of our systems are ad hoc systems stitched together with no real deep meaning except making things easy to comprehend in terms of the pen-and-paper beareacracy days. Even our computer databases are basically automated versions of pen and paper record keeping. Things are so ad-hoc basically, that it would be a miracle if numerical analysis couldn't overhaul our systems to be radically more streamlined and efficient.

[Eagleon]

> The AI would have to be omniscient as well as super-intelligent to devise the optimal society.

This is pure conjecture.

1) nothing needs to be omniscient, it just needs the same data we already have. AI has already proven better at synthesizing complex data than humans in many cases. So even existing shitty AI can improve on many things given just the data we have.

2) Neural networks and genetic algorithms for example don't contain any intelligence at all, yet they solve many problems we have not been able to.

Also "optimal society" isn't right. Sure, a "perfect solution" needs perfect data, but we don't need it perfect, just better.

And our current stuff is far from optimal. Back in Taylor's day factory owners probably thought they were the epitomy of efficiency. Yet Taylor just applies some numerical comparisons, and increases efficiency by orders of magnitude, without needing to make people work any harder. Taylors reductionist system, breaking down tasks into steps, seeing who was fastest at each step, them reassembling the steps into a new process. That's basically AI (an early genetic algorithm basically, since people will adapt each "best step" further by themselves and you can do multiple iterations of the Taylor process), the data analysis determined the design of the new process, not a person's judgement. The existing "experts" - people who did those tasks day in day out professionally for years were beaten by maths. This lead to a quantitative change - higher output, but the long term effects of that are qualitative change, changes to the way of life itself, since new products become commerically viable for the general population due to higher productivity that would have been impossible before, or have been luxuries only for the rich before that.

I think we're at a similar situation to the 19th century before Taylorism, where people think we have things pretty nailed down so computation couldn't come up with anything significantly better. But most of our systems are ad hoc systems stitched together with no real deep meaning except making things easy to comprehend in terms of the pen-and-paper beareacracy days. Even our computer databases are basically automated versions of pen and paper record keeping. Things are so ad-hoc basically, that it would be a miracle if numerical analysis couldn't overhaul our systems to be radically more streamlined and efficient.

Ok, so you admit that numerical analysis, design by neural network and GA, etc. do not indicate intelligence... so why do we need AI again?

Making everything more efficient is not going to solve the underlying problem, that we expand to fit our vessel like any other species. There's also the fact that simple ideas are often the best. Complex systems are not inherently more applicable to our problems.

Come to think of it, maybe the solution is to introduce some competition? A bit of man-machine warfare could do the trick, but so could uplifting a bunch of species that complement our needs but keep us in check by virtue of social (mis)agreement. Anyone working on enhancing the intelligence of plants? I'm completely serious, it seems about as realistic as introducing intelligence into a nonbiological system without also including those pesky social and metabolic pressures guiding its development that make us "bad" at sorting statistical outcomes for large sets of numbers.

[WillowLuman]

The internet and its effects are not quite as unprecedented as we'd like to think. Which is not to say that it's nothing new, but there are parallels to its effects in recent history.

Should we just start a separate environmental issues thread? The discussion here is interesting but centered more on the singularity than on those.

[penguinofhonor]

.

[Reelya]

Ok, so you admit that numerical analysis, design by neural network and GA, etc. do not indicate intelligence... so why do we need AI again?

Neural networks, genetic algorithms etc are what is defined as the field of AI. What they lack is consciousness. The concept of a singularity doesn't necessarily imply we get a "person in a box" out of it. So asking for that is a red herring.

The main point of the singularity concept is that these sorts of things aren't linear at all. Say you build a machine that can read and semantically understand college-level textbooks. That would be a very useful machine. Making the machine will take a long time. But once you have that machine, you can basically pour all the books into it in no time at all. That is a discontinuous growth in what that machine is then capable of. The implications of having the first such machine that is capable of doing that will have massive flow-on effects through every field of human thought.

But asking why AI would be useful, you might as well ask why any human smarter than average is useful. Think about possible future AIs. Is the potential limit of AI smartness less than an average  human, equal to an average human, or greater than an average human. There's nothing inherently special about the average intelligence of a human. The limit of machine intelligence is very unlikely to exactly fall there. The limit is likely to be vastly different to the limit for a human. And given that it's potentially possible to build a computer with many more components than a human brain, the limit is probably above the normal limit for a human. So, asking what good will AI do is like saying what's the point of having smarter-than-average people, when average people are perfectly sufficient.

[Angle]

and I'll point out, again, Singularity doesn't necessarily mean AI - it just means exponential growth of intelligence. We already have that. It's just that in this case, it's driven by having more people with more time and more resources and more knowledge so they can come up with better ideas and communicate them to other people faster and more effectively.

[Reelya]

Well, it's more like we have exponential growth in total computation power now, but that's just raw computing power. The real benefit comes when we get better automated software generation that can create much more powerful software models. Because each generation of that will leverage the hardware so much more. And software can be created and deployed much faster than hardware. Hence, this sort of thing will accelerate the effect of powerful hardware.