Bad For Good

OpenAI CEO Sam Altman’s comments to Politico during an interview are stirring up controversy. I also disagree with how he laid things out, but perhaps for a different reason.

The fastest way to lose an argument is to unknowingly accept a hostile premise. And yeah, I think Mr. Altman screwed up. Because he doesn’t think big enough about what ubiquitous automation should be able to do for people. Here’s the comment that people have become riled up about:

“I think there’s a lot of daylight,” Altman told Decoded co-author and POLITICO senior tech reporter Brendan Bordelon. Asked specifically where he diverges from Anthropic and its CEO Dario Amodei, Altman said, “we believe that the world should accept some bad things happening for the benefits of this technology and people having the agency.”

And here’s something that he said a bit later in the interview:

“I wouldn’t take a trade of saying, ‘We’ll make sure there’s no major hacks, there’s no misuse of this technology, there’s zero scams, there’s zero all the other bad things that will happen,’” said Altman. “Because I think people will do tremendously — orders of magnitude more — good stuff than bad stuff.”

For me, the problem is the assumption that the benefits of generative automation won’t be universally accessible seem to be baked into the statements. If there are parts of the world in which the best available use case for generative automation is perpetrating more effective theft or fraud against people with more resources than themselves, what, exactly, has the technology done for them? Given that a person who has a realistic opportunity to materially improve their lot in life via the “good stuff” the technology is purported to be bringing would prefer that to being bothered with the risks of perpetrating “bad stuff,” if someone lacks access to the ability succeed via doing “good stuff,” one can make the case the technology has failed them. Is pervasive automation worth the costs if it can’t make the pie bigger by increasing the size of everyone’s individual slices? Mr. Altman’s statements are based on a presumption that the technology can’t fundamentally alter the incentives of people. That it can’t create better, safer and more socially acceptable ways of looking out for one’s material well-being.

Of course, one can make the case that people will do “bad stuff” because they’re fundamentally bad people, and the tradeoff is going to be empowering them along with the fundamentally good people. Okay, then. But for the person who now finds themselves assailed by bad people consistently using generative automation in endless attempts to dispossess them of whatever “good stuff” they’re able to obtain for themselves (or that bad people incorrectly think they’ve obtained), how better off are they in the end? It’s not likely to be the case that the added benefits will necessarily cover the costs of required added security measures. So some people are going to be left holding the bag.

Sure, generative automation will solve a lot of problems for a lot of people. But Mr. Altman’s statement seems to assume the fact that there are going to be people whose problems are either a) unaffected or b) made worse by the technology, and that this is a necessary trade-off. To me, that presupposes that the technology is simply incapable of creating enough abundance that the winners will be ready, willing and able to share with the losers, while still seeing themselves as winners. And if technology never manages to live up to the hype, that starting point may indicate that things won’t turn out as optimistically as might be hoped.

And “all the other bad things that will happen” is papering over a lot. If a person believes that creating a highly effective biological weapon is a solution to a known problem, either because they feel they have a direct use for one, or (to paraphrase Robert Frank) because it’s bad not to have bioweapons when the other side has bioweapons, then that is just as reasonable a potential outcome as major hacks, scams and other “misuse” of the technology. Because “misuse” is little more than a use case that an outside evaluator doesn’t like.

In the end, the world should not have to accept more major hacks, misuse of the technology, scams, et cetera, because one of “the benefits of this technology and people having the agency,” should be the ability of people to not rely on these things to lead good lives.