Artificial Intelligence Makes People Stupid
In one of his endless apercus, G.K. Chesterson wrote that “when people stop believing in God, they do not believe in nothing. They believe in anything.”
Today, in a moment of supreme gullibility across the net, YouTube pullulates with a solemn faith in the “intelligence” of computers.
Any day now, we are going to meet the spectre of “artificial general intelligence” or even “super intelligence.”
We humans will comprehend our robots no better than we are comprehended by our cocker spaniels. If we are lucky, the machines will treat us as pets. If unlucky, as lunch.
In 1965, L. J. Good, who had been Alan Turing’s colleague fashioning AI under fire at Bletchley Palace during World War II, expounded the ultimate fantasy behind the AI movement:
“Let an ultra-intelligent machine be defined as a machine that can far surpass all the intellectual activities of any man however clever. Since the design of machines is one of those intellectual activities, an ultra-intelligent machine could design even better machines.
There would unquestionably be an ‘intelligence explosion’ and the intelligence of man would be left far behind. Thus,” he declared, “the first ultra-intelligent machine is the last invention that man need ever make, provided that it is docile enough to tell us how to keep it under control.”
In other words, he deemed artificial intelligence what theologians term an “escheating,” a “final thing:” part of an apocalyptic transfiguration of human life and cosmic destiny.
Optimized by computer scientists using techniques of intelligent design, this putative technology contrasts with human brains and bodies, which are said to have been optimized by the largely random processes of evolution. Since AI is an entirely mechanical system, it provides secular society with an eschatology for intelligent modern sophisticates who shun a belief in a deity and worship randomness.
Scientists foresee a coming planetary utility, a computer at once centralized and diffused, feeding on galactic floods of data and streams of energy to “think” and to replicate. Such a cosmic machine utterly dwarfs individual human brains confined to relatively tiny cranial cavities, bodies, and households.
The Human Brain Reigns Supreme
As Jaron Lanier, the shaggy sage of “virtual reality,” commented last year, “artificial intelligence (AI) makes people stupid.”
He did not mean that AI is so super-smart that it makes mere humans seem dumb by comparison. He meant that most computer scientists are astonishingly dumb on the crucial differences between their own minds and their microchips.
The fact is, measured by its multi-zettabyte connectome of neural connectivity, a single human brain is more complex than the entire Internet, while running on less than a billionth of the energy. But don’t worry. The inexorable advance of technology will sweep by humans soon in a heralded “singularity” like that foreseen by L.J. Good.
The Master Algorithm by Pedro Domingos illustrates many of these misconceptions. An interesting book that usefully analyzes lots of ingenious machine learning techniques, it reaches toward some overarching “master algorithm” that simulates a super-mind. It is a search for what Yuval Harari calls “homo deus.”
The assumption of the master algorithm is that if you add up all the defective and partial methods of current AI, they will complement and correct one another. You will somehow end up with a perfect mind-blend.
This error recapitulates Darwin’s key mistake: his view that genetics is a chemical blend of the characteristics of parent organisms. In blends, the errors also compound, while the distinctive advances are lost. Averages destroy information. In the same way, the computer science method of simulating creativity by adding randomness destroys information.
Dominguez fails to recognize the intrinsic differences between algorithms and thought processes: the map (symbolic scheme) is not the same as the territory (objects in reality).
No matter how fast you shuffle the symbols of the map you cannot guarantee that anything on the territory will behave the same way. They are two different domains. They are not ergodic: the same inputs do not reliably return the same outputs. The map is deterministic and able to be manipulated by AI algorithms, just as games and other regular domains are. By contrast, the territory — reality — is full of human agents and surprising outcomes, black swans and butterfly effects, creativity and invention.
As I explain in Life After Google, the greatest feat of AI is probably the Renaissance/Medallion fund of Nobel Laureate mathematician James Simons, recounted in Gregory Zuckerman’s new book, The Man Who Solved the Market: How Jim Simons Launched the Quant Revolution.
A market for currencies and securities, such as mastered by Simons and his colleagues led by Robert Mercer from IBM, is a symbol system. The Renaissance data center could master the symbols registering market changes just as Mercer’s IBM natural language translation system could capture previous patterns of speech. By telling you what has already happened, it can tell you what will happen if nothing changes.
Artificial intelligence can process symbols at immense speeds and scrutinize past patterns far faster than humans can. The Renaissance computers every second could allegedly front run assets for the equivalent of four months before any price could change.
Computer sciences fool themselves that if they shuffle symbols fast enough the symbols will somehow become real. But they don’t. So for all their unprecedented successes in beating the markets, Renaissance does not invest at all, except in their own system. They did not provide resources to test significant propositions in the market. They found patterns in past results faster than anyone else and faster than prices could change.
AI is simply not a mind. A mind can assimilate symbols and their objects and mediate between them through imagination. Rather than merely reshuffle the past, a mind can shape the future.
In the face of SEC treatment of human agency as presumptive inside trading, AI quants are taking over the financial markets. They now account for more than half of all trades. You can’t indict a computer.
But by pulling the so-called investment process inexorably into the pits of trivial symbol shuffling of past data, the result seems to render U.S. markets as dumb as a stone. Successful real human investors are deemed probable criminals. Our leading companies are attacked as scams, exploiting the public. Initial public offerings languish, and actual companies seeking funds are treated as “unicorns,” or mythical creatures.
We should really get beyond this belief system. There are many real companies out there, and opportunities are greater than ever.
In The George Gilder Report, I search for them by looking both in the US potentially and in many ventures overseas, from Tel Aviv to Shenzhen.
Join this adventure of real investment on the frontiers of human creativity.
Editor, Gilder’s Daily Prophecy