Bill Gates is starting to pump out some end-of-year book recommendations, and he identifies The Coming Wave by Mustafa Suleyman as his “favorite book about AI”. Here are a few quotes (from the Gates article):

…what sets his book apart from others is Mustafa’s insight that AI is only one part of an unprecedented convergence of scientific breakthroughs. Gene editing, DNA synthesis, and other advances in biotechnology are racing forward in parallel. As the title suggests, these changes are building like a wave far out at sea—invisible to many but gathering force. Each would be game-changing on its own; together, they’re poised to reshape every aspect of society…

In my conversations about AI, I often highlight three main risks we need to consider. First is the rapid pace of economic disruption. AI could fundamentally transform the nature of work itself and affect jobs across most industries, including white-collar roles that have traditionally been safe from automation. Second is the control problem, or the difficulty of ensuring that AI systems remain aligned with human values and interests as they become more advanced. The third risk is that when a bad actor has access to AI, they become more powerful—and more capable of conducting cyber-attacks, creating biological weapons, even compromising national security…

So how do we achieve containment in this new reality? …he lays out an agenda that’s appropriately ambitious for the scale of the challenge—ranging from technical solutions (like building an emergency off switch for AI systems) to sweeping institutional changes, including new global treaties, modernized regulatory frameworks, and historic cooperation among governments, companies, and scientists.

When it comes to AI, economic productivity, and job loss, it seems obvious that the answer is to take a portion of the economic value added by AI and reinvest it in services and benefits for the people adversely affected. Easy peasy right? And politically very difficult, at least in the U.S. “Value added tax” and “universal basic services and/or income” are words you could use to describe such programs, but we need to come up with better words and strategies if we are going to successfully describe these concepts to voters and neutralize the powerful interests who so far have been successful obstacles to these practical, somewhat obvious policies. The advantage of a VAT is the broadest possible tax base pays it in small increments over time rather than all at once, and therefore it is resented much less than filing an income tax return. If AI can truly increase economic productivity, then phasing in a VAT over time as productivity increases could be a way to increase quality of life for the greatest number of people possible. Throw in some automated counter-cyclical infrastructure spending along with the usual monetary policy adjustments, and you might have something. AI itself might be able to manage a system like this effectively in a way that is truly win-win for everyone.

It’s hard to be optimistic at this point in history about “historic cooperation among governments, companies, and scientists”. Still, maybe we have hit rock bottom on this and the coming trend will be up at some point.

The discussion of biological weapons and bad actors is chilling. Think of the ideologies that lead people to rationalize mass suicide and mass murder of civilians in events like 9/11 and the Oklahoma City bombing. The people who perpetrated those acts would certainly have used nuclear weapons if they had them handy. They will use biological weapons in the future if they can get their hands on them, and as the article points out it will be easier to get their hands on them and much harder to detect who has their hands on what. I don’t have an answer on this other than surveillance. Surveillance of AI, by AI perhaps? It sounds dystopian, but maybe that is what is needed – AI designed to be pro-human and pro-social looking for that needle in a haystack which is bad humans using bad AI to try to do something really terrible.