# Getting Exaflop computers by 2020:

sonhouse
Science 04 Mar '15 13:57
1. sonhouse
Fast and Curious
04 Mar '15 13:57
http://www.extremetech.com/computing/155941-supercomputing-director-bets-2000-that-we-wont-have-exascale-computing-by-2020

So they figure 1 Exaflop will = 1 human brain but takes 20 to 100 megawatts of power.
2. DeepThought
04 Mar '15 17:20
I don't agree that a conventional computer could ever be equivalent to a human brain. By conventional computer I mean a Turing machine. While they might be computationally more powerful, we have things like intuition, we are able to leap to the correct conclusion and then work out why we are right. Quantum machines may be able to do that, but that is not what they are talking about.
3. sonhouse
Fast and Curious
04 Mar '15 20:381 edit
Originally posted by DeepThought
I don't agree that a conventional computer could ever be equivalent to a human brain. By conventional computer I mean a Turing machine. While they might be computationally more powerful, we have things like intuition, we are able to leap to the correct conclusion and then work out why we are right. Quantum machines may be able to do that, but that is not what they are talking about.
I saw another estimate go 1 to 10 exaplops. So I doubt a 1 exaflop machine will equal a human. And besides, humans come in such a wide range of intelligence themselves, so maybe a 1 exaflop comp could simulate a dude with an IQ of 50, how many exa's would it take to equal Einstein or Newton?

If they show, say a 10 exaflop machine making scientific breakthroughs men hadn't thought of, THEN I would say they would be exceeding human ability, even the best of us. Till then, Whoopee, we simulated Bill O'Reilly.....
4. Soothfast
0,1,1,2,3,5,8,13,21,
04 Mar '15 21:04
Originally posted by sonhouse
I saw another estimate go 1 to 10 exaplops. So I doubt a 1 exaflop machine will equal a human. And besides, humans come in such a wide range of intelligence themselves, so maybe a 1 exaflop comp could simulate a dude with an IQ of 50, how many exa's would it take to equal Einstein or Newton?

If they show, say a 10 exaflop machine making scientific breakt ...[text shortened]... xceeding human ability, even the best of us. Till then, Whoopee, we simulated Bill O'Reilly.....
A Trash-80 could simulate a Bill O'Reilly and have the capability of adding small integers.
5. 04 Mar '15 21:24
Originally posted by sonhouse
I saw another estimate go 1 to 10 exaplops. So I doubt a 1 exaflop machine will equal a human. And besides, humans come in such a wide range of intelligence themselves, so maybe a 1 exaflop comp could simulate a dude with an IQ of 50, how many exa's would it take to equal Einstein or Newton?

If they show, say a 10 exaflop machine making scientific breakt ...[text shortened]... xceeding human ability, even the best of us. Till then, Whoopee, we simulated Bill O'Reilly.....
"So I doubt a 1 exaflop machine will equal a human."

So do I.
It only takes a handheld calculator from 80ies to calculate a square root with six decimals that no (but a few) humans can do.
6. sonhouse
Fast and Curious
05 Mar '15 12:491 edit
Originally posted by FabianFnas
"So I doubt a 1 exaflop machine will equal a human."

So do I.
It only takes a handheld calculator from 80ies to calculate a square root with six decimals that no (but a few) humans can do.
Then again Kasparov swore up and down a computer would never beat him....

Now a frigging cell phone can give him a good game.
7. 05 Mar '15 18:50
Originally posted by DeepThought
I don't agree that a conventional computer could ever be equivalent to a human brain. By conventional computer I mean a Turing machine. While they might be computationally more powerful, we have things like intuition, we are able to leap to the correct conclusion and then work out why we are right.
Which is not nearly as magical as you seem to think.

Quantum machines may be able to do that, but that is not what they are talking about.
Ordinary Turing machines can do that. What they lack is computing power and programming. But we will get there.

The biggest problem right now, is that computing has focused on what is essentially a single threaded model. The human brain achieves its massive computational power by doing multiprocessing on a massive scale. The problem is that there just aren't that many applications for that in the computer world so not much effort has been put into developing it except on a small scale in graphics cards. Instead the focus has been on single threaded computing as fast as possible.
8. DeepThought
05 Mar '15 19:04
Which is not nearly as magical as you seem to think.

[b]Quantum machines may be able to do that, but that is not what they are talking about.

Ordinary Turing machines can do that. What they lack is computing power and programming. But we will get there.

The biggest problem right now, is that computing has focused on what is essentially a single ...[text shortened]... in graphics cards. Instead the focus has been on single threaded computing as fast as possible.[/b]
Oh yes it is. Human's can take on NP hard problems and get it right. No matter what the processing power of a Turing machine it simply cannot compete with a human on some sorts of problems.

Bear in mind that we have significant philosophical differences. You seem to regard the world as deterministic. I do not. So from your point of view a quantum machine is a Turing Machine. In my world view, because I'm not tied to determinism, a quantum Turing machine is qualitatively different from a conventional one.
9. sonhouse
Fast and Curious
05 Mar '15 21:591 edit
Originally posted by DeepThought
Oh yes it is. Human's can take on NP hard problems and get it right. No matter what the processing power of a Turing machine it simply cannot compete with a human on some sorts of problems.

Bear in mind that we have significant philosophical differences. You seem to regard the world as deterministic. I do not. So from your point of view a ...[text shortened]... ied to determinism, a quantum Turing machine is qualitatively different from a conventional one.
And that would be because a quantum machine IS a different animal. You might call it a massively parallel machine because of the superposition of numbers and logic effects, it can be an or gate, and gate, nor gate, Xor gate, a not gate, and so forth all at the same time. Sounds like massive parallel to me. Not only massive parallel but simultaneous massive parallel. You can't get that in a conventional computer even if it was the size of the solar system using up all the energy of the sun for regular logic gates one molecule in size it wouldn't be able to do that. It would be dam good at simulations thoughđź™‚

So far it is still theory, I don't think anyone has actually built a quantum computer yet. For one thing you might need hundreds of qubits and now we are working oneseys and twoseys. I saw one piece where they had 9 qubits all acting to correct one another's mistakes so there certainly is progress but that is still just a few qubits, about 200 short of a real machine.
10. 06 Mar '15 09:32
Originally posted by DeepThought
Oh yes it is. Human's can take on NP hard problems and get it right. No matter what the processing power of a Turing machine it simply cannot compete with a human on some sorts of problems.
A Turing machine can't compete yet merely because they are neither fast enough nor programmed to do so. But there is nothing whatsoever mathematically stopping them from doing so.

Bear in mind that we have significant philosophical differences. You seem to regard the world as deterministic. I do not. So from your point of view a quantum machine is a Turing Machine. In my world view, because I'm not tied to determinism, a quantum Turing machine is qualitatively different from a conventional one.
In this case it isn't really important whether the world is deterministic. A deterministic Turing machine can do everything and more that a human brain can do - using basically the same techniques. And to a large extent, we already know what those techniques are, we just haven't got the hardware to implement them on that scale.
11. sonhouse
Fast and Curious
06 Mar '15 11:42
Originally posted by FabianFnas
"So I doubt a 1 exaflop machine will equal a human."

So do I.
It only takes a handheld calculator from 80ies to calculate a square root with six decimals that no (but a few) humans can do.
When a computer writes a best seller novel or wins the Noble prize for some fantastic new theory, THEN maybe we can say computers are more intelligent than humans.
12. 06 Mar '15 13:06
Originally posted by sonhouse
When a computer writes a best seller novel or wins the Noble prize for some fantastic new theory, THEN maybe we can say computers are more intelligent than humans.
When a human can decrypt a megabyte size document using a 160-bit secret key, THEN maybe we can say humans are superior to computers...
13. sonhouse
Fast and Curious
06 Mar '15 13:36
Originally posted by FabianFnas
When a human can decrypt a megabyte size document using a 160-bit secret key, THEN maybe we can say humans are superior to computers...
Different skill sets. It's like saying when a human can beat a Corvette in the quarter mile, we can say humans are superior to cars.
14. 06 Mar '15 13:481 edit
Originally posted by sonhouse
Different skill sets. It's like saying when a human can beat a Corvette in the quarter mile, we can say humans are superior to cars.
Exactly. I bet you haven't written any best seller novels nor won any Nobel prizes. Why demand that of a computer?
Most human skill sets are dependent on human education as much as intelligence, and even then we all end up with different skills. The first intelligent computers will have a very different education from us and thus will have a very different kind of intelligence and a very different skills set.
The first computer that does write a best seller novel will probably not be able to see or walk or do many of the things we do with our brains. It will have a very specialized intelligence targeted at writing novels. It will probably have read more novels than any human ever has, and will be able recall them better too. It will be very different from us.
15. DeepThought