A Harvard Business School presentation on artificial intelligence by Professors Karim Lakhani and Marco Iansiti this week was fascinating but creepy. While the technology is empowering numerous new businesses, most of those businesses seem to be in China. Gradually, as the capabilities of artificial intelligence combined with Big Data sank in, I came to a chilling realization that in the wrong hands, the technology may be very dangerous indeed. We had better make sure our defenses are raised against this possibility.
The presenters gave a number of examples of AI in action, which certainly stimulated thought. The first example was a reproduction Rembrandt painting – they loaded all the existing Rembrandts into a computer and asked the AI algorithm to produce a Rembrandt of a 30-40-year-old man, and it produced a picture that was quite convincing. However, I’m no art expert, but I’m not sure the AI result was definitively a Rembrandt – it was a little flat.
That is however not surprising; the great man himself employed a studio full of assistants to finish his paintings and add the easy bits. Hence if you load all the Rembrandts into a computer, you will have work that is part-Rembrandt, part “school of.” This painting was definitely “school of” and might have been “Rembrandt on a bad day” but you certainly wouldn’t have rated it as a top-quality Rembrandt masterpiece. Similarly, the little AI-Mozart that I have heard never sounds like the best Mozart, but more like some random 18th Century tiddly-pom.
The second AI example they gave was a Chinese insurance company’s program for counting pigs. The insurance company provides farm insurance on some large percentage of China’s 850 million pigs, and naturally wants to check how many pigs are in a particular farm, and whether the same pigs are being shuttled from farm to farm to defraud the insurance company (I once knew someone whose family pulled that trick for EU sheep subsidies). The clever AI could not only count the pigs, but also use pig facial recognition to determine which pig was which. Clearly, this is a highly valuable capability, which can replace lots of human pig counters; being AI it has a learning capability and positive economies of scale (the more pigs it sees, the better it gets at recognizing them). Of course, China being China, one is forced to recognize that the same capability would work on people….
The final example the presentation gave was Ant Financial, the Chinese payments system that has expanded into a financial services conglomerate with 1.4 billion customers and only 10,000 employees. (Only Chinese marketing people could name a consumer-oriented financial services company after that most un-libertarian of insects. “Pussycat Financial” would at least have pretended that the company’s customers would be allowed to flourish independently, each in their own way!)
Ant Financial clearly benefits from the increasing economies of scale that AI can provide. It also uses AI algorithms and Big Data on its customers to generate new product offerings for its gigantic customer base. Naturally, AI can also be used to predict whether the customers will accept those offerings, and what might tempt customers into new offerings that might prove risky for them but carry higher commissions.
Up to a point, this is harmless and good marketing. However, it comes close to the Chinese government’s “social credit” scoring, which is used to determine who can travel, who gets particular job opportunities etc. Given Ant’s own close connections to the Chinese authorities, its data-gathering must be highly concerning for residents of China, and at least moderately concerning for the rest of us.
That is the problem with the current level of AI – “Weak AI” as it is called. It allows control-freak humans to gather information on citizens at a level never previously possible, then use that information for social control. We see it even in the United States, where the social media companies – Google, Twitter and Facebook – have ceased to be neutral platforms for the populace to express themselves, and have become machines whereby the “woke” Silicon Valley executives who run those companies can control the information allegedly free Americans send to each other.
President Trump announced last week that he would ban the purchase of the Chinese social networking services Tik-Tok and WeChat, because of their ability to collect information on U.S. consumers and in WeChat’s case on their conversations and funnel that information to the Chinese commissars. That is a useful protection, but it will last only as long as President Trump does. A Biden administration would doubtless eventually wake up to the problem, but very probably too late.
Naturally, it is less concerning that a bunch of leftist ex-hippies in Silicon Valley control the information flow than that the Chinese government has the potential to do the same. But really, it makes very little difference to me: my philosophical outlook differs almost as much from that of Silicon Valley as it does from that of Beijing, so my First Amendment rights can as easily be infringed by Silicon Valley’s gentle dweebs as by the jackbooted thugs of Beijing. In my own case, I can avoid such control by avoiding almost completely going on social media, but then, I am 70 years old and live a hermit-like existence here in Poughkeepsie. Were I 50 years younger and interested in interactions (ideally with the opposite sex, you never know!) an entire absence from social media in today’s world would seriously cramp my style, so I would surely get tempted.
“Weak” AI does not have the potential to become a Hitler itself; it is nowhere near powerful enough – it lacks the “intelligence,” artificial or otherwise. It is simply an unpleasant tool that the world’s Hitlers can make bad use of, and that therefore grants them a new and undeserved springtime. Strong AI, the Nirvana to which Silicon Valley looks forward in a decade or two, is another matter; that has the capability to become a Hitler itself, without humans needed to direct its tyranny.
I cannot help believing that the entire technological progress of the last 25 years has been misguided. Almost 20 years ago, I wrote for UPI a two-part essay “The Business of Playing God” that looked at the possibilities that then appeared to be approaching rapidly in the field of genetic engineering. Alas, spurious ethical objections to human genetic engineering and the monstrous flow of capital and talent in a different direction have prevented those possibilities from coming to fruition. I thought at that stage we were only around a decade from human cloning and from an even more exciting possibility: the artificial creation through genetic engineering of human beings with superior intelligence. Alas, those possibilities are no closer today.
Skeptics of genetic engineering, particularly in the George W. Bush administration, raised fears of the damage that could be done by genetic engineering technologies in the hands of bad guys. Yet those threats to freedom have already appeared, brought to us by AI, with no genetic engineering needed. The idea that the People’s Liberation Army could advance upon us with a regiment of human clones, or that an evil genetically engineered Einstein could destroy the world, were always far-fetched. The genetically engineered clone or genius humans would be human, subject to all the same desires and moral impulses as the rest of us, and as difficult to order about as a non-engineered army or genius.
Human intelligence is not a very useful tool for tyrants, because the humans will fight against the tyrants for their own freedom. Genetically engineered humans thus offer us a huge prospect of improving the world, allowing us to take advantage of their superior capabilities for scientific and other advances in a huge number of areas. You don’t need very many genetically engineered geniuses, provided they are geniuses.
Artificial intelligence, on the other hand, offers us few possibilities of genuine improvements to human thought, because it is not independently sentient – its Rembrandts, unlike those produced by genetically engineered humans, are mere copies of the originals. However, AI is an appallingly powerful force multiplier for those, whether Hitlers, Chinese commissars or Google geeks, who want to control our speech, thoughts and actions.
Let us therefore reverse the direction of scientific research, to produce more life-enhancing genetic engineering, and fewer Orwellian AI threats. Thereby we can produce nighttime for Hitler and the Chinese commissars, and springtime for Shakespeare, Mozart and Einstein.
(The Bear’s Lair is a weekly column that is intended to appear each Monday, an appropriately gloomy day of the week. Its rationale is that the proportion of “sell” recommendations put out by Wall Street houses remains far below that of “buy” recommendations. Accordingly, investors have an excess of positive information and very little negative information. The column thus takes the ursine view of life and the market, in the hope that it may be usefully different from what investors see elsewhere.)