Tuesday, April 12, 2016

The problem of AI ownership

Blue Mountain Supercomputer Over at Marginal Revolution Alex Tabarrok makes the optimistic case that artificial intelligence will benefit the middle class. His argument is based on comparing the possible development of AI to the current increase of educated people in countries like China and India. The problem is that this analogy suffers from a serious flaw: the issue of ownership.

The people currently getting an advanced education in China and India are, of course, autonomous individuals. But any true AI developed by a company would be owned under current IP law. In addition, any work product of the AI would be owned by the company.

Any smart company that makes a true AI would instruct it to design an even better AI, which would in turn be instructed to design a better AI, and so on until you quickly have AI much better and smarter than any human. Within a rather short period of time, one company or a small group of companies/investors could end up owning over 90 percent of the intellectual capacity on Earth. That is counting all artificial and human sources of intelligence. That is an insane amount of money and influence in the hands of a few which requires almost no labor or help from the masses.

While a rather imperfect analogy, the closest comparison is countries with huge amounts of easily extracted oil and gas. In these places it takes relatively few human employees to product huge amounts of wealth. In Norway it has worked out well, with the country using its oil money to broadly spread prosperity. Yet many other countries show that an easily controlled source of incredible wealth can give a small group of elites the freedom to do terrible things. Look at Saudi Arabia, where oil money has made it possible for the leaders to deny the populous democracy and given them the economic freedom to oppress women. Or Turkmenistan, where oil wealth has helped make possible the creation of a truly oppressive regime with often bizarre rules, like changing the word for "bread" to be the same as the president-for-life's mother's name.

I'm personally optimistic about what AI could theoretically allow for. Potentially it could produce an incredible spike in science, technology, wealth and free time to benefit all. But my optimism for the technology is greatly hindered by my pessimism regarding our politics. I fear that the American government is not well designed to quickly make the major changes to IP law, tax policy and welfare that would be required. Even if they did, I also fear companies might just move their AI servers to other countries with less benevolent governments.

My first sci-fi novel, Cobalt Slave, is now on sale at Amazon. Please consider leaving a review.