In December, an artificial intelligence programme has demonstrated a considerable ability. "Alpha zero", such as the algorithm of the company deep mind, killed himself within hours of chess and then played against "Codfish", the best chess program in the world, in the whole of chess knowledge of mankind. "Alpha zero" taught "Stockfish" a crushing defeat: out Of 100 games alpha zero "won" their 28, the rest ended in draws. The professional world was impressed with how quickly the "alpha zero" superhuman abilities had gained. The former world chess champion Garry Kasparov followed the development of artificial intelligence. We met him this week in the framework of the "Applied Machine Learning Days" at the ETH Lausanne.
Mr Kasparov, in 1997, they lost against the IBM chess computer Deep Blue. What is the role of chess today, when it comes to the development of artificial intelligence?
Many scientists use chess to test your ideas. It is a comparatively simple System and, therefore, very practical. To fly like the fruit fly Drosophila, a model organism for genetic research, chess is a model system in the field of artificial intelligence.
After "alpha zero" had swept the chess program "Stockfish" from the Board, they wrote in the journal "Science", they were happy about that "alpha zero" is exactly how a dynamic, open-style care. We see in "alpha zero" human creativity at work?
no. Words like "dynamic" or "risky use" in connection with a machine, be very careful.
If you look at the history of mankind, then it happens repeatedly, that a new disruptive technology has destroyed whole industries, before then new jobs have been created. It's called progress.Garry Kasparov
you have written it yourself: In your eyes in the style of the "alpha-zero had a" dynamic, risky, and aggressive.
of Course, alpha zero is a different machine than the traditional chess program "Stockfish". However, alpha zero has only one eye: the chances of winning. If "alpha zero" chess pieces makes sacrifices and never-before-seen, brilliant trains, then, because it would like to beautiful and creative play. It is the only power, because it sees in this type of game benefits. In contrast to the conventional chess, computers, alpha zero has gained "" however, a much deeper insight into the game of chess. I was enthusiastic about the game of "alpha zero".
Actually, it means that the higher the level at chess, the more boring and technical the game is.
this is a myth that has cleaned up the "alpha zero". We have seen that chess can also be at the highest level is extremely attractive.
Why alpha zero is much better in chess, Go or other Games than previous programs?
Conventional computer programs raw use of Computing force and need a lot of data, on the basis of which they calculate their moves. In the case of systems such as "alpha zero" is different. These Algorithms need only the rules of the game. Then they play against themselves, to generate their own data and learning very fast. This is a gigantic step forward.
alpha zero "the perfect chess program"?
no, I would not say that the data generated by the program, are perfect. But they are different than the chess knowledge of mankind. We should accept the fact that the data, the "alpha zero" is generated, the human Knowledge are superior. We are now machines, the use of our information, not our accumulated chess knowledge, but build on the basis of our instructions to your own database, their own Knowledge.
Our role in relationship to the machine will be different, but a crucial one.Garry Kasparov
What is the Benefit of these programs?
they help us to see things from a different angle that was previously hidden.
Where except in chess, we could benefit from?
Practically everywhere, where we manage closed systems such as, for example, the rules of chess, or to define the rules for the Board game Go. That is where these machines Shine.
Could so help tools such as "alpha zero" to solve the big unsolved questions of science or to discover new active ingredients for the medicine?
of Course. But, once again, It is Central that we are able to define closed systems with parameters that provide the framework for such a program as well as the rules in chess.
The new challenge for researchers would be to identify these closed problems, and machine-oriented to formulate rules for it?
Exactly. Our role in relationship to the machine is a different, but crucial: Our ability to find the right angle of view for the machine on which you should concentrate.
There are in fact Jobs that are at risk. No Question About It.Garry Kasparov
There are fears that the learning machine could run one day become independent and run Amok.
This fear of artificial intelligence is material for Hollywood. With reality this has nothing to do with it. There is not the slightest indication of how it should be possible that these machines could break out of their closed System. These doomsday scenarios of black painters are nonsense.
Some people worry that their Job in artificial Intelligenzzu lose.
Yes, there are in fact Jobs that are at risk. No Question About It. Because I want the people fool. I understand if those who are affected are not happy. But in the end, we will benefit humanity by artificial intelligence.
in Concrete terms?
we Take the much-discussed example of medical diagnosis – for example, in the radiology Department. If intelligent machines are in the Interpretation of medical images better than humans, they can get some services. Then well-paid Jobs will be at risk. I agree that this can be a Problem. At the same time, the analysis of medical images is cheaper. So, while some Jobs will go in Switzerland and in other Western countries lost, we can save millions of lives in Africa or India, because there many people is a diagnostic option available, you could not pay for before.
but There are also negative applications of artificial intelligence. Think of Fake News, to Monitoring on the basis of Big Data, the influencing of elections. For such Algorithms can also be used.
always Been a certain amount of people have used innovation to harm other people or to destroy something. But the Problem is not the technology or the science itself. Science and technology are neutral, agnostic, neither moral nor immoral. I'm worried, but not because of the artificial intelligence, but because of the bad people, because of dictators, gangsters and terrorists. You can abuse this in the free world advanced technology, to harm us. This is a social and a political Problem, not a Problem of artificial intelligence.
we Are facing great social upheaval?
If you look at the history of mankind, then it happens repeatedly, that a new disruptive technology has destroyed whole industries, before then new jobs have been created. It's called progress. So, yeah, artificial intelligence, and the whole digitization pose challenges. But we will benefit in the end. I've met with Deep Blue more than 20 years ago. I think I have the right to say: This is not the end of the world.
Created: 30.01.2019, 18:12 PM