Apparently, you don't know what "limited move" means
Both games have limited moves. You can control StarCraft with a mouse and keyboard, and each of those input devices only has a fixed range of inputs. The difference is only in scale. In StarCraft, you send a few inputs per second, and in Go you can take a minute.
If we need 1000x more computational power for something a human finds 2.5x more complex, do you really think that we will only need 12x more computational power for something that humans find 12x more complex?
I was talking about the number of neurons in the human brain taking part in the game decision process. I think it's reasonable to assume that the scaling in the brain corresponds with the scaling in nodes in artificial neural nets. Perceptive complexity isn't a very good measure, I think. People find common tasks, like walking, simpler than playing StarCraft, but that could be because their brains are optimized for the first task, and not for the second. I think someone playing StarCraft uses a bigger part of their brain than someone playing Go, because there's more overlap between StarCraft and daily life, so more neurons can be recruited to join in the effort. But I highly doubt the difference is more than a factor 100.
Besides, as you say Deep Blue did not use NN, and the method Deep Blue used will likely not work well for AG anyway, resulting in the 2.5x complexity increase needing 1000x more computational power.
Because Deep Blue didn't use NN, I don't think it's useful to discuss the relative complexity increase.
What makes you think a 12x (or whatever) complexity is a tractable problem?
Because DeepMind already had a 12x bigger version running before.
All the evidence I've seen points to AI scaling being O(N^m) with N being the complexity.
I doubt it. Our neocortex is only twice as big as the chimpanzee's, and our total brain is only 3x the size, but we are capable of tasks that orders of magnitude more complex.
the craft can use it's much more efficient jet engines to reach hypersonic speeds, saving the rocket engine for when it's too high to get any oxygen from the air. That reduces the weight of fuel required, as well as potentially piping, pumps, etc.
Less fuel, but you need two kinds of engines, which only adds more piping, pumps, etc, plus the dead mass of the extra engines. At least the extra fuel you carry will be burned, reducing mass as you go.
It's called redundancy, not just physical satellites, but also redundancy in ideas, technology, and methods of data processing. Also, every big nation launching their own satellites means they can pursue their own ideas, without endless committee meetings. Plus, if there's only one type of satellite, designed by a single party, then there will be even more conspiracies about how the data is manipulated.
I mean, we're talking about $100 million for a satellite that lasts a decade. That's what we pay for a football coach.
If you weren't paying as much tax, or even paying no tax, then you'd have more money you could choose to directly help fund a satellite program like this one, without politicians interfering.
Problem is that different satellites use different types of sensors, which can make it harder to compare the results from one satellite to the other. And when you calibrate the output from one to match the other, people will blame scientists for "adjusting" the data.
We went from needing 30 x 120MHz CPUs to win at Chess (Deep Blue), to 1202 CPUs and 176 GPUs to win at Go (Alphago).
AlphaGo Zero only uses 4 TPUs, and is much stronger than the 176 GPU version. It is also much stronger than the 48 TPU version that beat Lee Sedol, while only using a fraction of the space and power. If the goal is only to narrowly beat the human world champ, maybe 1 or 2 TPUs would suffice.
IOW, we used almost 1000x more resources to win at Go than at Chess.
AlphaGo Zero uses less power than Deep Blue did, and plays at a much higher level (comparing with best human players).
But the biggest problem with your analysis is that Chess is solved in a completely different way. Deep Blue didn't use neural nets, but relatively simple human-generated heuristics combined with deep brute force search. There have been some experiments with NN based chess programs, but so far, the results have not led to top-class performance. It is conceivable that a NN based chess program would require more resources than a Go program.
I agree that StarCraft would most likely require more processing units than Go, but at this point, that's all we can say.
How many more brain cells does a human StarCraft player use compared to a human Go player ? If it's a factor 12, then the AlphaGo Lee version, running on 48 TPUs may be able to do the job, given the proper algorithms.
If they figure they can get velocity faster by using lift to counteract gravity, thereby saving engine thrust to use for velocity, that could make sense
The problem is that the wings add extra mass that also needs to be accelerated, which could negate all the savings.
How do you go about solving a Starcraft game where there are infinite possibility of moves?
At any point in time, the number of inputs is limited by the availability of controls. You can press a key, move the mouse, push a joystick or other controller in a finite number of ways.
"They said we'd never achieve $FOO, and then we did. This proves we'd achieve $BAR" is a fundamentally flawed argument, regardless of what values you assign to FOO and BAR.
Except in cases where FOO and BAR are essentially the same thing, but BAR is a bit further on the scale of size and complexity than FOO, and that we can reasonably expect our development of hardware and software to be able to tackle problems with greater scale and complexity in the future.
Go and chess are full-knowledge games, you know at all points, where your opponent currently is, what his moves have been and where he can go. Not to say, there are only a very small numbers of paths you can take at any point in time.
For decades, computers have failed at Go because there are too many paths you can take. Since AlphaGo defeated Lee Sedol, people suddenly seem to think it's a relatively simple game.
Within a few years, we'll have an AI beat a human at StarCraft too. You'd better mount your goalposts on wheels.
Self-awareness is not needed for many useful applications.
The correct approach to AI will only be possible once we understand how a biological brain is capable of those things
No. In most cases, when you try to understand something, it's smart to try to build something as soon as possible, so you can test your ideas. When you build an artificial neural network, you can do all kinds of experiments, and take measurements, and get a much better understanding.
Also, our biological brain is severely constrained by requiring that it can be built and operated from turkey sandwiches, rather than metals and electricity. Already, there's a growing number of applications where an artificial neural net outperforms a human brain.
US is around the middle of the pack, producing value of $2,291 per ton of CO2 emitted. China is one of the worst, at $435/ton
The numbers are distorted because a lot of the US/Eur manufacturing is outsourced to China.
A Chinese factory makes a widget for $5 that gets sold in the US for $35. All the CO2 produced to create the widget is counted as China's emissions towards the $5, while the US claims $30 added value for zero CO2 emissions.
If you measure performance, that just means that a good worker only has to put in 4-6 hours a day to achieve the performance required to keep the job. In the remaining time he can read slashdot, or get a second job.
Six thousand years of moral philosophy would disagree with this statement.
That's because philosophers are looking for an objective truth that doesn't exist. However, people are pretty good at coming up with a pragmatic solution that works in practice, even if it's inconsistent.
think everyone should think the same way.
We don't. Ask people if abortion is okay, or if it's okay to eat dogs, or if you should be allowed to kill animals for fun. There's plenty of disagreement, but generally we manage to come to a compromise. Artificial Intelligence is just one more case, and will be treated in a similar way. If enough people agree that some rights should be given to AI, they will be given.
Slashdot Mirror
Hey, now give me a reason for putting anything in space at all, ever.
GPS, telecommunication, earth observing sciences, weather forecasting, mapping, espionage, to name just a handful of obvious reasons.
what is the point of exploring the moons of Jupiter if we plan to never ever leave earth,
It's interesting to see what's out there.
Of what use is a newborn baby?
They are fun to make. And they don't cost billions a piece.
Apparently, you don't know what "limited move" means
Both games have limited moves. You can control StarCraft with a mouse and keyboard, and each of those input devices only has a fixed range of inputs. The difference is only in scale. In StarCraft, you send a few inputs per second, and in Go you can take a minute.
If we need 1000x more computational power for something a human finds 2.5x more complex, do you really think that we will only need 12x more computational power for something that humans find 12x more complex?
I was talking about the number of neurons in the human brain taking part in the game decision process. I think it's reasonable to assume that the scaling in the brain corresponds with the scaling in nodes in artificial neural nets. Perceptive complexity isn't a very good measure, I think. People find common tasks, like walking, simpler than playing StarCraft, but that could be because their brains are optimized for the first task, and not for the second. I think someone playing StarCraft uses a bigger part of their brain than someone playing Go, because there's more overlap between StarCraft and daily life, so more neurons can be recruited to join in the effort. But I highly doubt the difference is more than a factor 100.
Besides, as you say Deep Blue did not use NN, and the method Deep Blue used will likely not work well for AG anyway, resulting in the 2.5x complexity increase needing 1000x more computational power.
Because Deep Blue didn't use NN, I don't think it's useful to discuss the relative complexity increase.
What makes you think a 12x (or whatever) complexity is a tractable problem?
Because DeepMind already had a 12x bigger version running before.
All the evidence I've seen points to AI scaling being O(N^m) with N being the complexity.
I doubt it. Our neocortex is only twice as big as the chimpanzee's, and our total brain is only 3x the size, but we are capable of tasks that orders of magnitude more complex.
the craft can use it's much more efficient jet engines to reach hypersonic speeds, saving the rocket engine for when it's too high to get any oxygen from the air. That reduces the weight of fuel required, as well as potentially piping, pumps, etc.
Less fuel, but you need two kinds of engines, which only adds more piping, pumps, etc, plus the dead mass of the extra engines. At least the extra fuel you carry will be burned, reducing mass as you go.
Which seems wasteful to me
It's called redundancy, not just physical satellites, but also redundancy in ideas, technology, and methods of data processing. Also, every big nation launching their own satellites means they can pursue their own ideas, without endless committee meetings. Plus, if there's only one type of satellite, designed by a single party, then there will be even more conspiracies about how the data is manipulated.
I mean, we're talking about $100 million for a satellite that lasts a decade. That's what we pay for a football coach.
If you weren't paying as much tax, or even paying no tax, then you'd have more money you could choose to directly help fund a satellite program like this one, without politicians interfering.
Nice idea, but isn't going to happen.
Europe launched the Cryosat-2 satellite in 2010.
Problem is that different satellites use different types of sensors, which can make it harder to compare the results from one satellite to the other. And when you calibrate the output from one to match the other, people will blame scientists for "adjusting" the data.
We went from needing 30 x 120MHz CPUs to win at Chess (Deep Blue), to 1202 CPUs and 176 GPUs to win at Go (Alphago).
AlphaGo Zero only uses 4 TPUs, and is much stronger than the 176 GPU version. It is also much stronger than the 48 TPU version that beat Lee Sedol, while only using a fraction of the space and power. If the goal is only to narrowly beat the human world champ, maybe 1 or 2 TPUs would suffice.
IOW, we used almost 1000x more resources to win at Go than at Chess.
AlphaGo Zero uses less power than Deep Blue did, and plays at a much higher level (comparing with best human players).
But the biggest problem with your analysis is that Chess is solved in a completely different way. Deep Blue didn't use neural nets, but relatively simple human-generated heuristics combined with deep brute force search. There have been some experiments with NN based chess programs, but so far, the results have not led to top-class performance. It is conceivable that a NN based chess program would require more resources than a Go program.
I agree that StarCraft would most likely require more processing units than Go, but at this point, that's all we can say.
How many more brain cells does a human StarCraft player use compared to a human Go player ? If it's a factor 12, then the AlphaGo Lee version, running on 48 TPUs may be able to do the job, given the proper algorithms.
If they figure they can get velocity faster by using lift to counteract gravity, thereby saving engine thrust to use for velocity, that could make sense
The problem is that the wings add extra mass that also needs to be accelerated, which could negate all the savings.
And what's the practical use of having thousands of manhours of zero-G ?
It's not actually infinite, just exponential.
Indeed. And Chess and Go are exponential too. The difference is the branching factor and depth of the tree you need to consider.
How do you go about solving a Starcraft game where there are infinite possibility of moves?
At any point in time, the number of inputs is limited by the availability of controls. You can press a key, move the mouse, push a joystick or other controller in a finite number of ways.
"They said we'd never achieve $FOO, and then we did. This proves we'd achieve $BAR" is a fundamentally flawed argument, regardless of what values you assign to FOO and BAR.
Except in cases where FOO and BAR are essentially the same thing, but BAR is a bit further on the scale of size and complexity than FOO, and that we can reasonably expect our development of hardware and software to be able to tackle problems with greater scale and complexity in the future.
Go is a full knowledge, limited move type game.
Not really. You don't know what your opponent is planning.
Go and chess are full-knowledge games, you know at all points, where your opponent currently is, what his moves have been and where he can go. Not to say, there are only a very small numbers of paths you can take at any point in time.
For decades, computers have failed at Go because there are too many paths you can take. Since AlphaGo defeated Lee Sedol, people suddenly seem to think it's a relatively simple game.
Within a few years, we'll have an AI beat a human at StarCraft too. You'd better mount your goalposts on wheels.
because it will never be self-aware
Self-awareness is not needed for many useful applications.
The correct approach to AI will only be possible once we understand how a biological brain is capable of those things
No. In most cases, when you try to understand something, it's smart to try to build something as soon as possible, so you can test your ideas. When you build an artificial neural network, you can do all kinds of experiments, and take measurements, and get a much better understanding.
Also, our biological brain is severely constrained by requiring that it can be built and operated from turkey sandwiches, rather than metals and electricity. Already, there's a growing number of applications where an artificial neural net outperforms a human brain.
No. Even Deep Blue sucked at Go.
US is around the middle of the pack, producing value of $2,291 per ton of CO2 emitted. China is one of the worst, at $435/ton
The numbers are distorted because a lot of the US/Eur manufacturing is outsourced to China.
A Chinese factory makes a widget for $5 that gets sold in the US for $35. All the CO2 produced to create the widget is counted as China's emissions towards the $5, while the US claims $30 added value for zero CO2 emissions.
If you measure performance, that just means that a good worker only has to put in 4-6 hours a day to achieve the performance required to keep the job. In the remaining time he can read slashdot, or get a second job.
The best NN have comparable performance on the MNIST dataset as humans (around 0.2% error rate).
despite effectively having much higher-resolution
MNIST images are normalized at 28x28 pixels per digit.
Exactly. What if China's AI program fails, and we've developed all this great technology for nothing ?
That said, as I mentioned earlier, the probability of AI ending up hostile is extremely low though.
Unless someone designs it to be hostile so they can attack another country.
Six thousand years of moral philosophy would disagree with this statement.
That's because philosophers are looking for an objective truth that doesn't exist. However, people are pretty good at coming up with a pragmatic solution that works in practice, even if it's inconsistent.
think everyone should think the same way.
We don't. Ask people if abortion is okay, or if it's okay to eat dogs, or if you should be allowed to kill animals for fun. There's plenty of disagreement, but generally we manage to come to a compromise. Artificial Intelligence is just one more case, and will be treated in a similar way. If enough people agree that some rights should be given to AI, they will be given.
Satellites don't measure surface air temperatures. They measure troposphere.