Skip to main content

Our golem is talking back to us

In 1964, the great mathematician and cybernetics pioneer Norbert Wiener published a small book titled God and Golem, Inc, in which he said that the relationship between the cybernetic machine and man is similar to the relationship between man and his creator. He draws from this analogy some cautionary principles, which are more relevant today as computers have become so much more powerful and robots more lifelike.

A golem in Jewish mythology was a creature made of clay into which life had been breathed. The original golem was Adam. Certain holy people were apparently able to make golems in imitation of God. However, golems created by humans were primitive beings capable of performing repetitive tasks but unable to operate independently. To Wiener, a computer is such a golem. 

The biblical creator liked to play games with his creatures, gambling with Satan for the soul of Job, for example. Men do too. Wiener points to what was then a state-of-the-art game playing machine, Arthur Samuel's checkers-playing program, which could learn and occasionally was able to beat its admittedly less-than-grand-master creator. In 2006, Deep Fritz, a machine, beat world chess champion Vladimir Kramnik. Since then, computers have been happy to play each other, as humans offer little competition.

Wiener pointed out that business and war are often compared with game strategies. Mathematical game theory has been used explicitly in analyzing global political conflicts. He warns against assuming that because machines are proficient at games, they can therefore be used to make real-life decisions. In the Cold War era, he was particularly worried about nuclear war.

"I know nothing to contradict the assumption that Russia and the United States are toying with the idea of using machines, learning machines at that, to determine the moment of pushing the atomic bomb button..." We seem to have escaped, for the present, that particular disaster, but Wall Street has embraced the use of machines in high stakes games of profit and loss. By 2010, over 70% of market trading was due to high frequency trading wherein machines use algorithms to place trades, buying and selling shares in microseconds, faster than their human masters can even monitor. The 2010 "Flash Crash" in which the Dow Jones dropped 1,000 points in 9 minutes was blamed, in part, on an increase in volatility enabled by high frequency trading.

The acid test for a golem is its inability to speak. The Sanhedrin tells of Rav Zeira encountering a golem, and when it was unable to answer him, he realized that it had been made by a sorcerer and commanded it to return to dust. Similarly, the Turing test is supposed to be able to distinguish between a dumb machine and one exhibiting intelligent behavior. If an impartial judge cannot distinguish its responses from those of a human being, then the machine must be intelligent.

By now, most people have encountered the Apple iPhone's Siri app. Siri seems able to understand normal human speech and usually responds appropriately in a well-modulated human voice. Perhaps she wouldn't fool a well-educated judge, but she is definitely pushing the envelope. Siri is holding simultaneous conversations with thousands of people at any given time, and presumably all these conversations are being parsed and analyzed somewhere deep in the bowels of Apple, Inc. Siri is learning very quickly.

"Render unto man the things which are man's and unto the computer the things which are the computer's," advised Wiener, but we are already mixing the two. Our children and grandchildren will have remote conversations with Siri's clones and descendants without knowing (or caring?) whether or not they are speaking to a human.

Blog Name