Marc Andreessen’s recent article “Why AI Will Save the World” is—sadly—an example of how polarized society and public discourse have become, particularly on such high-stakes topics as the opportunities and risks posed by the proliferation of AI. While I largely applaud (and mostly agree with) his optimism about the tremendous potential of AI to improve virtually every aspect of human life and flourishing, some of his arguments against AI regulation are not only absurd, but threaten to unnecessarily poison this important debate.
In Andreessen’s view, those in favor of any kind of new regulation fall into either of two categories: Scared cowards, whose irrational fear that AI would pose an existential threat to humanity is blown vastly out of proportion—the baptists—and those who stand to benefit financially from stricter regulation—the bootleggers. Unsurprisingly, this analogy refers back to the attempts to prohibit alcohol consumption in the US in the 1920: The baptists back then were legitimately concerned, albeit confused and misguided, religious groups and the bootleggers were… well, those who made a fortune from peddling black market liquor.
The second group in today’s debate, according to Andreessen, is made up of individuals and institutions wanting to rake in public funding for AI safety research. If your lab recently applied for a grant to further our understanding of the AI goal alignment problem, for example, you’re a duplicitous bootlegger in Andreessen’s book. On the other hand, if you’re honestly worried about something like the effects that AI-generated noise and misinformation will have on democracy, you’re a gullible, but harmless, alarmist whom we shouldn’t pay any more attention to than the guy on the street corner with his handwritten “The End is Near” sign.
Andreessen himself picked this particularly condescending comparison to religious extremists and violent mobsters. And that begs the question: What role should we assign someone like him in this analogy?
Lo and behold, besides baptists and bootleggers there was a third group lobbying 1920s regulators on the topic of prohibition, namely the liquor companies themselves. They argued that any form of regulation would harm the economy, that it would infringe on civil liberties, and that, ultimately, it was each and every person’s individual responsibility to engage responsibly with the substances they produced. Their goal, needless to say, was to kill off all attempts to control the sale or consumption of alcoholic beverages, even to children, so that their target market would remain as large as it could be, and access to that market would not be not obstructed by legal hurdles.
Marc Andreessen’s venture capital firm, a16z, is managing about 35 billion dollars in assets. Many of the companies they are invested in either have AI at the core of their business model, or at least have a lot to lose should lawmakers decide to tighten the screws about what they can and cannot do with that technology. And while there are many objectively good arguments to not, for example, ban all AI research altogether, even the smallest changes in the regulatory environment will mean hundreds of millions gained or lost for Andreessen and a16z.
As he himself acknowledges in his very article, there is hardly an absolutist standpoint in this debate, nor is there on any political issue of real importance. There are no completely objective rights and wrongs, and nearly everyone who voices an opinion also has some sort of agenda. Therefore, we should judge each argument based on its own merit, and not on the—obvious or hidden—interests we ascribe to the people who brought it forward. I’m not going into them today, but trust me, I’m fully on board with many of the points Andreessen brings up, because they are sound, logical, and reasonable. But the murkiness of Andreessen’s stance feels unsetting to me: He puts himself in this non-existent absolutist position from which he assigns either bad intentions or mere stupidity to anyone who argues in favor of AI regulation. But he completely fails to acknowledge where he himself is coming from, and where his interests lay. He is part of the Ayn Rand-inspired libertarian elite of Silicon Valley. He belongs to a small group that stands to profit the most from an unrestricted, free ranging market economy. He has billions of dollars to gain from lawmakers dragging their feet on AI regulation.
All of these facets of Andreessen himself, interesting as they may be, should not blind us to the fact that many of his arguments against regulation are pretty good. But the least I would expect form a civilized public debate would be an acknowledgement that some the counter-arguments, particularly those brought forward by credible AI safety researchers around the globe, also have their merits and not just straw men raised by scientists in order to accrue public funding.