We are a geeky company and Go is one of the most played games among our colleague. During the last few days there were some pretty exciting discussions on the recent development of AI. What does Alphago winning against Lee Sedol mean? We asked FACT-Finder CEO Carsten Kraus, our founder and active Go-player about the importance of Alphago not for the future of technology but for society.
AI has won against mankind in many different games before. What is the new quality?
Carsten Kraus: The new quality here is that AlphaGo has learned most by itself. While the victory of #deepblue against Kasparow was a victory of genius programmers plus superior calculating power, AlphaGo has achieved its improvement since October mainly on its own: By playing games against itself and learning. Such is the power of Deep Neural Networks.
There are other DNN victories that illustrate how autonomous those machines have become: such as Space Invaders, where the DNN learned only by “looking” at the screen pixels and experimenting with keystrokes.
Go was “known” to be unbeatable by computers, because it mainly relies on abstraction patterns – such as real life does. Nobody could grab those patterns, thus they could not be taught to computers. It is not that nobody tried: In 1985, the Taiwanese industrial Ing Chang-Ki. Ing Chang-Ki and his Ing Foundation offered 40 Million Taiwanese Dollars (at the time >1 Million US$) to a software that reaches 1 Dan until the year 2000. The prize was never won: in 2000 – 4 years after the victory of deepblue against Kasparow in chess – computer Go was still at the level of a weak amateur player.
A few months ago, AlphaGo already won against the European Champion. The victory against Lee Sedol looks like only a small improvement.
Carsten Kraus: In fact the learning curve was tremendous. In Go, the differences in strength are huge: If the European Champion played 100 even games against Lee Sedol, he would probably not win a single one.
Is Artificial Intelligence already superior to human intelligence?
Carsten Kraus: In all its complexity of abstraction layers, Go is a game with simple rules, simple actions and quick feedback. Feedback is essential for DNNs to learn. Real life is much more complex, there are various kinds of inputs and outputs, and feedback can take time and be ambiguous rather than “win/lose”.
So, this is all just science with no practical meaning?
DNNs have already provided breakthroughs in a number of practical applications, the most prominent one being speech recognition – without DNNs, Siri would never have been possible.
What will be next?
Carsten Kraus: A “game” which is very complex, but has very simple input and output and a clear measurement of success, is the stock market. Man-designed Algorithmic Trading is already there, but DNNs will take over soon. Combine deepbrain’s algorithmic learning with the text understanding of IBM’s #Watson, and computers outsmart Warren Buffet within the next few years. Humans will never take back control again.
Then Google instead of Warren Buffet makes the money. Most people are not trading at the stockmarket. Why should we care?
Carsten Kraus: Money rules the world. If you control money, you can control people, industries, and countries. (See novel “a trillion dollars” by Andreas Eschbach http://www.amazon.com/One-Trillion-Dollars-Andreas-Eschbach-ebook/dp/B00LLBBXSA). Nietzsche claims that our commonplace moral values are not “human nature”, but have been learned over centuries when punishment for stealing, betraying, killing was extremely harsh. I believe Nietzsche is right, and that automatically learning DNNs may come to completely different moral values. The software may find out that it can optimize its results by playing foul, by feigning a sale first or driving an industry close to ruin before taking over majority shares. This will certainly influence employees lives. For the software, it’s still a game.
Will computers ever be creative?
I believe so. The style AlphaGo played is considered “creative” by top Go players. However, it is still a long way from Go creativity until computers autonomously invent. Autonomous (secondary research) scientific findings will be easier for them than autonomously inventing even simple things like a new kind of bottle opener. For real life they have to understand a lot of things which our human brains have, unwittingly, learned within the first two years of our life. (See my speeches on Semantic Search vs Keyword Search, e.g. https://www.youtube.com/watch?v=vSWLafBdHus)
When will computers take over the world?
I think 2035. But even when we cease to be the dominant intelligence on the planet, it need not mean we also give up being the dominant species. Now, wise political decisions are required. Soon.
Thank you for the interview!