Three snapshot applications of AI:
During early phases of the war in Gaza, the Israeli military used software to select bombing targets on a scale that would not have been possible for human analysts.1
During the Trump administration’s initial attack on the federal government, there was lots of nonsense about how Elon Musk and DOGE were using software to identify waste. House Speaker Mike Johnson commented that Musk has “created these algorithms that are constantly crawling through the data, and… the data doesn’t lie.”2
Health Secretary Robert F. Kennedy Jr. commissioned a report with the ridiculous title “Make America Healthy Again.”3 It turns out that many of the citations in the study are erroneous, including references to articles which simply do not exist. Incorrectly citing things and misrepresenting results is plausibly human malfeasance or incompetence, but totally inventing sources suggests chatbot hallucination.
The wrongs committed here are morally different, and I don’t want to suggest a false equivalence. But each of these cases provides a specimen of how reliance on AI has been used to further dangerous agendas. Yet the reliance on AI is really just a sideshow.
Israel’s ongoing war in Gaza has shown that they are willing to bomb and murder indiscriminately, outstripping even their hyped AI’s alleged ability to identify targets. Having ravaged the federal workforce across the board, securing government contracts for himself and defanging possible regulation, Musk has withdrawn from all the claims about inefficiency. And the MAHA report was only written using chatbots because that was the easiest way to generate the bullshit report that they otherwise would have had to write.
To put them in reverse order: The attack on public health, the attack on democracy, and the literal mass murder are all things that the people with power wanted to do anyway. They would have done it without AI at more or less the same pace. The AI was just the rhetoric of the moment.
It is common to say that the output of an AI chatbot is bullshit.4 In these cases (and plenty of others) the invocation of AI to justify decisions is bullshit in a further sense. The appeal to AI is a placeholder where they would have put anything, because the outcome of their deliberation was decided in advance of the reasons.5
- This article runs down some of the concerns, but it’s a big enough deal that it has its own Wikipedia page.
- Data doesn’t lie, but for the uninteresting reason that it doesn’t assert either. If data could perform speech acts, we wouldn’t need algorithms to crawl through it.
- Remember measles? Let’s bring it back!
- In Harry Frankfurt’s technical sense of expressions produced with an indifference to truth or falsity.
- This could be construed as Frankfurt-style bullshit about claims of justification. Or it could be construed as an adjacent style of bullshit, expressions produced with an indifference to whether the reasons they contain are at all convincing.