BRAND NEW "EVOLUTION" THREAD

[Siro]

Neither of you are wrong; I think you are talking past each other. I understood babe to be saying that, if a human is allowed to thoroughly study the inner workings of a program and explore how it reacts, they are usually able to devise a plan or five to defeat it (Nakamura has posted a few examples of this). You are saying that, if the computer is being treated as a regular opponent, it can out-perform a human. Both positions seem to be correct.

But we're no longer talking about chess, are we? Of course it is possible to beat a program if you have unlimited access to the source code. You can try as many combination of moves until you find some flaw in the program's response. But this is like saying a 5 year old with a time controlling device can beat any grandmaster simply by going back in time after each move, having seen how the expert will react. That's not at all what we're talking about. A computer can easily outperform any human player in a GAME of chess.

 

[Scat]

Can I get my ears clipped first so I am weaponized? This will determine greatly what animal I can kill.

 

[carolinajazz]

Always thought about this with dogs, pit bulls specifically. If one got hold of me, I always pictured myself, with a free hand, jamming my fingers and hand in through its' eye(s), ripping them out, and then going into the brain to gouge the hell out of it. I'm not sure I'd be able to fend one off any other way, without any weapons handy.

...well, if a Pit Bull bites me....it will gum you!

 

[carolinajazz]

You honestly think that a human could beat the most powerful chess playing computer? I think you and Babe are vastly underestimating how much more efficient the computer is at processing all of the alternative plays than the human mind is.

Negative! The brain is much more than a computer! Although extremely fast, computers handle only one piece of information at a time, whereas the human nervous system processes millions of pieces of information simultaneously!

How do the marvelous mechanical memories of computers compare with the memory capacity of the human brain? The RCA Corporation’s Advanced Technology Laboratories offers these comparisons, as published in “Business Week” magazine: With all of human technology the human brain still has a capacity 500 times as great as an advanced computer memory system!

The human brain contains one thousand billion billion protein molecules (one followed by twenty-one zeros). Each of these molecules can undergo many changes in its structure and afterward retain the changed shape. This changed structure may represent a new memory impression. As the molecules are replaced by wear, they duplicate themselves so that the replacement molecules are the same. By this the possible number of memories becomes indefinite, beyond comprehension!

A brain composed of such neurons can never be ‘filled up.’” Some researchers say that everything you have ever seen, heard or experienced is somewhere, somehow, in your memory!

Our active memories hold several billion times more information than a large contemporary research computer! Computers are simply left behind by the capabilities of the human mind!

......so, in conclusion....I would have taken Garry Kasparov or even Bobby Fischer, over your chess computer!

 

[Gameface]

Dogs don't weigh as much as an average human. What are their weapons? Their teeth and some fairly dull claws. So a dog bites you and starts to shake it's head. You gonna die now? You can snap that ****er's neck. You can gouge out it's eyes. You can punch it or kick it. You can body slam it. Its body is no stronger than yours. It might tolerate a bunch of pain, but it has it's limits too.

Now a pack of dogs? That's different. But if a full grown man was killed by a single dog then that man wasn't trying to live.

Watch people get attacked by a dog. They go back on their heels pretty much every time and take a defensive posture. That dog is not as strong as a man, it's just more fearless and determined. If the man is less fearless and more determined than the dog the dog is in a lot of trouble. Yeah, the man is going to come away with serious injuries. Accept that from the start and do what you can to overpower the dog.

Humans don't give themselves credit for how powerful we are. We're relatively large animals. We're fairly dexterous. We're bad *** even not counting our intelligence.

Now a mountain lion...not so much. A tiger...no. A bear, no we just don't have the size to even compete. But with weapons or a group of humans we're back in the game.

Keep in mind we are social creatures and throughout history we have traveled in groups and carried weapons. We are formidable.

 

[One Brow]

But we're no longer talking about chess, are we? Of course it is possible to beat a program if you have unlimited access to the source code. You can try as many combination of moves until you find some flaw in the program's response. But this is like saying a 5 year old with a time controlling device can beat any grandmaster simply by going back in time after each move, having seen how the expert will react. That's not at all what we're talking about. A computer can easily outperform any human player in a GAME of chess.

I'm not sure about the chess you play, and I know it's not true about the chess I play, but understanding an opponent's weaknesses and attempting to exploit them is a big part of grandmaster play, as is improving on your weak areas so opponents will not exploit them. Also, just because you got a computer to make a certain move in 5 practice games does not guarantee they will make that same exact move in a live game.

A grandmaster can use this knowledge to beat a computer for the exact same reason the 5year-old could not use time travel to beat the grandmaster; the understanding and use of long-term strategies (possibly arcing out over 50 moves) as opposed to short-term tactics.

 

[One Brow]

Negative! The brain is much more than a computer! Although extremely fast, computers handle only one piece of information at a time, whereas the human nervous system processes millions of pieces of information simultaneously!

You never hear of a dual-core or quad-core computer?

Some researchers say that everything you have ever seen, heard or experienced is somewhere, somehow, in your memory!

They are wrong. Memory doesn't work that way.

......so, in conclusion....I would have taken Garry Kasparov or even Bobby Fischer, over your chess computer!

Fine, but you would lose.

 

[Siro]

I'm not sure about the chess you play, and I know it's not true about the chess I play, but understanding an opponent's weaknesses and attempting to exploit them is a big part of grandmaster play, as is improving on your weak areas so opponents will not exploit them. Also, just because you got a computer to make a certain move in 5 practice games does not guarantee they will make that same exact move in a live game.

A grandmaster can use this knowledge to beat a computer for the exact same reason the 5year-old could not use time travel to beat the grandmaster; the understanding and use of long-term strategies (possibly arcing out over 50 moves) as opposed to short-term tactics.

You said they'd examine the 'inner workings of the program'. That to me meant looking at bugs in the algorithm. If you meant that a modern computer with a sophisticated program can be beaten after a few exploratory practice games, then I simply require more evidence. I mean, 5 years ago, they had well prepared world class players battle a top chess program running on phone hardware (to give the players a chance). And some of them still lost, while the rest barely squeezed by. A 2009 phone has about as much processing power as my toothbrush.

They no longer hold human v. Computer chess matches, by the way.

 

[One Brow]

You said they'd examine the 'inner workings of the program'. That to me meant looking at bugs in the algorithm.

Yes, that is what I meant, except the algorithms don't really have "bugs", they have weak spots and situations where they mis-evaluate a position.

They no longer hold human v. Computer chess matches, by the way.

In such matches, humans are not given the chance to examine the program, and so they tend to lose. What's your point, other than "if the computer is being treated as a regular opponent, it can out-perform a human"?

 

[Siro]

Yes, that is what I meant, except the algorithms don't really have "bugs", they have weak spots and situations where they mis-evaluate a position.



In such matches, humans are not given the chance to examine the program, and so they tend to lose. What's your point, other than "if the computer is being treated as a regular opponent, it can out-perform a human"?

The whole thing was a response to Babe's assertion that a "human with a plan can reliably beat a computer". Which is clearly not the case. What you're saying is a world-class chess player with access to the algorithm can possibly devise a plan to exploit any weaknesses he or she finds in that specific program. But I had always been talking about playing chess. My point was crystal clear from the start.

Edit: By the way, I'm conceding the point about examining the algorithm because it was never the argument I made. I don't actually know whether this is even possible in 2014 against the best programs on dedicated computers.

 

[DutchJazzer]

You never hear of a dual-core or quad-core computer?

bro it is 2014
we have six eight en 12 core consumer cpu.


dedicated chess computers more often then not have more then 1 cpu.

giving you dozens of cores.

 

[GVC]

Keep in mind we are social creatures and throughout history we have traveled in groups and carried weapons. We are formidable.

I haven't read through the whole thread, but this point is key. Individuals die, populations survive. People are incredible at planning, organizing and cooperating.

 

[[size/HUGE] boobs [/size]]

What are those girlies on rail doing? If Boris is there he would jump over rail to land on tigers back. If this doesnt paralyze I will knock him out with one punch.

There was time in life I went in search of thrills like this. Siberian Forest was crazy.

 

[One Brow]

The whole thing was a response to Babe's assertion that a "human with a plan can reliably beat a computer". Which is clearly not the case. What you're saying is a world-class chess player with access to the algorithm can possibly devise a plan to exploit any weaknesses he or she finds in that specific program. But I had always been talking about playing chess. My point was crystal clear from the start.

Edit: By the way, I'm conceding the point about examining the algorithm because it was never the argument I made. I don't actually know whether this is even possible in 2014 against the best programs on dedicated computers.

In 2014, perhaps it is not possible. Outside of what I consider an arbitrary demarcation between "playing chess" and playing a game designed to exploit the weaknesses of the computer, we haven't really disagreed here.

 

[carolinajazz]

God creates man.
Man invents chess.
Man invents computer.
Man plays chess against computer.
Computer beats man at chess.
Man pulls plug on computer.
Computer unable to play chess ever again!

 

[MVP]

Man creates God.
Man invents chess.
Man invents computer.
Man plays chess against computer.
Computer beats man at chess.
Man pulls plug on computer.
Computer unable to play chess ever again!

Fixed.

 

[DutchJazzer]

God creates man.
Man invents chess.
Man invents computer.
Man plays chess against computer.
Computer beats man at chess.
Man pulls plug on computer.
Computer unable to play chess ever again!
GOD now mad.
God pulls plug on Man

fixed

 

[♪alt13]

But technology have been steadily advancing (possibly exponentially) for hundreds of thousands of years, and there is no reason to expect it to suddenly stop. The problem is that you're looking at the limitation of very specific technologies to draw broad conclusions about entire categories.

The examples you gave about flight and rocketry are the classic cases found in literature about the subject. They are instructive, but also a bit deceptive. If you take "airplanes" as a category, then yes, you can show that they reached an optimal design limit, at which point their progress slowed down considerably. But that is not the best way of looking at the situation. Transportation technologies have been advancing since the dawn of humanity. From the wheel, all the way to modern hypersonic fighter jets. Since the introduction of the plane, humans have figured out how to get into space, going farther and farther each decade. Even if you consider the advancement in airplanes, I'd say wood planes that can travel at running speed to modern Boeing composite Dreamliners are impressive for just 100 years (a blink of an eye compared to the length of human history). Individual technologies do reach a limit, but I don't think technological categories do. At least not in the short term. Like Joe said in response to your post, what if fuel-less rockets were developed? Here is a link to NASA's latest test of a fuel-less system that they tested last year:

https://www.libertariannews.org/wp-...ustProductionFromanRFTestDevice-BradyEtAl.pdf

The news really shook the world of physics when it was released a couple of months ago, and numerous teams are trying to confirm and replicate the results. But that's nothing. History is long. In 1900, the greatest minds couldn't even begin to conceive of the computers we have today. The whole concept would have sounded surreal and impossible to them (despite some attempts at simple mechanical calculating machines).

Now for computers. I design computer chips for a living. The advancements that I've seen in my short career blow my mind. I don't know if Moore's Law will continue at its current pace, but it CERTAINLY will continue at some pace. People have been saying that we've run into the limits of transistor miniaturization for a couple of decades now. But Intel just announced a new fab in Israel to start their next die shrink. They envision continuing with the same silicon die shrinks until early 20s at least. After that, we've truly run into the limits of traditional transistor miniaturization in a single plane. But silicon is far from the only option, and we're WAYS away from the computerization limits imposed by the laws of physics. In my lab, we have computing prototypes that use the spin of individual particles to perform computations (called spintronics). In theory, these can be scaled to create computers that are trillions of times more powerful than anything we have. And even that is not the end of the line. And I'm talking about classical computing. Quantum computing is a whole different game.

I think you're right about the unlikeliness of competition between n A.I. and humans, but I don't think you're completely right about the reasons. For one, energy efficiency is a temporary problem. Eventually, we will have fusion power with inexhaustible fuel. Who knows what else we'll have in the future. Anti-matter power plants? Power extracted it from the vacuum itself? And that's not counting the unimaginable techniques that can be invented by superhuman A.I. The other part about enhancing our own brains is spot on. We will do so as we understand more and more about what intelligence actually is, until we ourselves become the A.I. I believe that is the most likely scenario. But who knows? As our capabilities continue to climb, the power of even a single individual to inflict damage also increases. Hopefully we'll be able to survive the malicious elements in our own species.

While I can admit that the history of flight technology is an imperfect analogy your use of Chessmaster as an example of intelligence is just as if not more dubious. I will read your link later but after reading the opening statement it seems that this discovery (if verified) may prove to be more useful in theory than application. Scaling a technology that produces micronewtons of force into one that can produce meganewtons is almost certainly not going to be just "doin it bigger".

I do not doubt that it would be possible to create a human level intelligence. The fact that we are here is testament to the fact that it is indeed possible. My doubts were in response to the idea that AI may turn out to be a replacement for us. First I think that popular projections of near future computing power are likely over optimistic. As an expert can you give me a reasonably accurate projection of when these new computing technologies will be competitive with silicon transistor technology? Second I do question the desirability, competitiveness, and ethical implications of AI.

Energy efficiency is not the only issue AI faces in terms of competition but I think it is an important one. The problem of energy efficiency does not end if and when we construct fission power plants. We are already running our civilization on(and all life on earth is dependent upon) nuclear fission from the sun. The problem is not a lack of energy it is the competition for the finite resources required to transmit, store, and harness it. Forgetting about any direct competition with humans for those resources we have to remember that such a technology would be competing with holographic Monday Night Football.

Do we need more minds? Even considering the most optimistic projections for when we may be capable of creating an AI we will have 9 billion human minds already here. The same technology that could be used to create AI could be used for computational functions. I don't think we will need more minds but we will probably need many more computers.

The ethical implications I think are pretty clear. We have been thus far quite successful in limiting technologies for which we have real safety or ethical concerns. Nuclear weapons and Human cloning are both good examples. If we come close to creating an intelligent sentient AI I am confident that we will be able to enforce a moratorium while we wrestle with the implications and we may decide on an outright ban.

Despite the destruction and harm that technology has caused and the dire predictions people have been making for centuries technology has been successfully harnessed again and again to improve the conditions of humanity and expand it's population. There is no reason to assume that we will not continue this trend and responsibly manage and overcome the "malicious elements in our species".

 

[Siro]

While I can admit that the history of flight technology is an imperfect analogy your use of Chessmaster as an example of intelligence is just as if not more dubious. I will read your link later but after reading the opening statement it seems that this discovery (if verified) may prove to be more useful in theory than application. Scaling a technology that produces micronewtons of force into one that can produce meganewtons is almost certainly not going to be just "doin it bigger".

According to those studying the issue, scaling up the technology is a question of building better conductors. With a superconducting chamber, you can get a thrust of up to 630kN/kW. That's probably enough to lift a city. But I do agree that we shouldn't jump to any conclusions about practical applications until we know if this is even real. I am glad you actually looked over the calculations, so here is the theoretical basis of conductivity v. thrust:

https://nextbigfuture.com/2014/08/shawyers-2014-presentations-for.html

By the way, I do not think that a computer's ability to beat humans at chess shows that computers are more intelligent than humans. It was an example of how we've built computers that can outperform humans in very specific areas, but they are still far from being generally intelligent.

I do not doubt that it would be possible to create a human level intelligence. The fact that we are here is testament to the fact that it is indeed possible. My doubts were in response to the idea that AI may turn out to be a replacement for us. First I think that popular projections of near future computing power are likely over optimistic. As an expert can you give me a reasonably accurate projection of when these new computing technologies will be competitive with silicon transistor technology? Second I do question the desirability, competitiveness, and ethical implications of AI.

I don't have any disagreements with the first part of the statement. I think that our existence is a proof that A.I. is an achievable goal, much like how it is a proof that true nanotechnology is possible. Popular projections had been spot on up until about 10 years ago. Then problems of quantum tunneling and heat dissipation (which you talked about in a different post) proved to be significant obstacles, and growth of computing power slowed a bit since then (it is still incredibly fast though). As the technology stands today, without any major adjustments to silicon photolithography, current pace of exponential growth in computing power will continue at least until 7nm die shrink sometime in the early 20s (that's the smallest size to be successfully demonstrated). Chances are, the process will be adjusted to allow a couple more die shrinks, down to 4nm process in the mid to late 20s. By then, computers should be around 1000x as powerful as today. That is when we'll reach the limits of miniaturization of silicon-based integrated circuitry.

It is very difficult to predict which paradigm will take over once that happens. Currently, there are many competing technologies vying to be the next computing substrate. None of them is anywhere near ready at the moment. I don't think any of the really amazing technologies will be ready by the mid 20s either. I think we'll either transition to using non-silicon semiconductors, or find ways to build transistors along the z-axis on silicon wafers. Both of these methods should be feasible by then (since they can be built now, but are not commercially viable), and will allow the extension of Moore's Law for another decade or two. After that, we better have a radical solution ready. Graphene will be amazing if it can be manufactured cheaply, perfectly, and with an introduced band gap for performing computations (the problem with graphene currently is that it is too persistently conductive for computation). Single particle computation will probably be widely available towards the end of the century.

Energy efficiency is not the only issue AI faces in terms of competition but I think it is an important one. The problem of energy efficiency does not end if and when we construct fission power plants. We are already running our civilization on(and all life on earth is dependent upon) nuclear fission from the sun. The problem is not a lack of energy it is the competition for the finite resources required to transmit, store, and harness it. Forgetting about any direct competition with humans for those resources we have to remember that such a technology would be competing with holographic Monday Night Football.

We're talking about fusion power, not fission (which is what is currently available). Fusion produces vast amounts of energy. Much more than any fission reactor. And it uses hydrogen isotopes to produce the power. I remember reading that the oceans contain enough fusion fuel to power the planet for hundreds of millions of years. And space is packed full of hydrogen. So it is unlimited power, without any pollution. It would, without a doubt, solve any foreseeable energy problems. I need to emphasize again that I'm not bringing up space resources, which will definitely be mine-able by the middle of the century. Even without those, fusion would solve all our energy problems. It would also solve all our water problems since energy required for desalination becomes irrelevant.

Do we need more minds? Even considering the most optimistic projections for when we may be capable of creating an AI we will have 9 billion human minds already here. The same technology that could be used to create AI could be used for computational functions. I don't think we will need more minds but we will probably need many more computers.

Why not have more intelligence of all kinds? More humans, more A.I., more animals engineered by humans to be intelligent, more dumb object infused with computational power, etc. The more the merrier, as long as we can expand past the boundaries of the planet.

The ethical implications I think are pretty clear. We have been thus far quite successful in limiting technologies for which we have real safety or ethical concerns. Nuclear weapons and Human cloning are both good examples. If we come close to creating an intelligent sentient AI I am confident that we will be able to enforce a moratorium while we wrestle with the implications and we may decide on an outright ban.

Despite the destruction and harm that technology has caused and the dire predictions people have been making for centuries technology has been successfully harnessed again and again to improve the conditions of humanity and expand it's population. There is no reason to assume that we will not continue this trend and responsibly manage and overcome the "malicious elements in our species".

There are no guarantees in life. I think technology is our best hope for immortality (literally and figuratively). But you're right, it comes with a steep price tag, and there are many difficult discussions to be had.

Thanks for the thoughtful response.

 

[♪alt13]

According to those studying the issue, scaling up the technology is a question of building better conductors. With a superconducting chamber, you can get a thrust of up to 630kN/kW. That's probably enough to lift a city. But I do agree that we shouldn't jump to any conclusions about practical applications until we know if this is even real. I am glad you actually looked over the calculations, so here is the theoretical basis of conductivity v. thrust:

https://nextbigfuture.com/2014/08/shawyers-2014-presentations-for.html

By the way, I do not think that a computer's ability to beat humans at chess shows that computers are more intelligent than humans. It was an example of how we've built computers that can outperform humans in very specific areas, but they are still far from being generally intelligent.



I don't have any disagreements with the first part of the statement. I think that our existence is a proof that A.I. is an achievable goal, much like how it is a proof that true nanotechnology is possible. Popular projections had been spot on up until about 10 years ago. Then problems of quantum tunneling and heat dissipation (which you talked about in a different post) proved to be significant obstacles, and growth of computing power slowed a bit since then (it is still incredibly fast though). As the technology stands today, without any major adjustments to silicon photolithography, current pace of exponential growth in computing power will continue at least until 7nm die shrink sometime in the early 20s (that's the smallest size to be successfully demonstrated). Chances are, the process will be adjusted to allow a couple more die shrinks, down to 4nm process in the mid to late 20s. By then, computers should be around 1000x as powerful as today. That is when we'll reach the limits of miniaturization of silicon-based integrated circuitry.

It is very difficult to predict which paradigm will take over once that happens. Currently, there are many competing technologies vying to be the next computing substrate. None of them is anywhere near ready at the moment. I don't think any of the really amazing technologies will be ready by the mid 20s either. I think we'll either transition to using non-silicon semiconductors, or find ways to build transistors along the z-axis on silicon wafers. Both of these methods should be feasible by then (since they can be built now, but are not commercially viable), and will allow the extension of Moore's Law for another decade or two. After that, we better have a radical solution ready. Graphene will be amazing if it can be manufactured cheaply, perfectly, and with an introduced band gap for performing computations (the problem with graphene currently is that it is too persistently conductive for computation). Single particle computation will probably be widely available towards the end of the century.



We're talking about fusion power, not fission (which is what is currently available). Fusion produces vast amounts of energy. Much more than any fission reactor. And it uses hydrogen isotopes to produce the power. I remember reading that the oceans contain enough fusion fuel to power the planet for hundreds of millions of years. And space is packed full of hydrogen. So it is unlimited power, without any pollution. It would, without a doubt, solve any foreseeable energy problems. I need to emphasize again that I'm not bringing up space resources, which will definitely be mine-able by the middle of the century. Even without those, fusion would solve all our energy problems. It would also solve all our water problems since energy required for desalination becomes irrelevant.



Why not have more intelligence of all kinds? More humans, more A.I., more animals engineered by humans to be intelligent, more dumb object infused with computational power, etc. The more the merrier, as long as we can expand past the boundaries of the planet.



There are no guarantees in life. I think technology is our best hope for immortality (literally and figuratively). But you're right, it comes with a steep price tag, and there are many difficult discussions to be had.

Thanks for the thoughtful response.

Honestly I haven't read the pdf yet(just the first 2 paragraphs)but it looks interesting and will definitely read it in full in the next few days.

I actually meant fusion(brain fart). That paragraph makes no sense otherwise. The point that I was trying to make is that limitless ability to generate electricity through fusion does not necessarily mean limitless ability to transmit and harness that electricity. The fact that we are even talking about mining asteroids is a clear indication that the materials needed to take advantage of a limitless supply of electricity are going to become much more expensive.