Who’s Afraid of Killer Robots?

In their book The AI Con Emily M. Bender and Alex Hanna observe that many of the same people warning us about the existential dangers of AI also stand to gain the most from its widespread adoption. This incongruity is the main reason I am not terribly worried about a future filled with Terminators. 

Imagine Toyota proclaiming that their latest model has a five percent chance of driving us over a cliff. That kind of behavior would make sense if Toyota were forced to disclose the risk by government regulation or the risk of a lawsuit, but that is not what the AI companies are doing. They are practically announcing the risk. “Hey, our technology is so powerful it might destroy humanity.” 

What are we to make of this behavior? Could it be that the billionaires really do care? This seems unlikely. One does not become a billionaire through altruism. (One becomes a billionaire through inherited wealth and borderline sociopathy.) Besides, an AI that would destroy humanity would also destroy them. 

If you want to know what someone truly believes, look at what they do. The tech oligarchs say, “This technology is powerful enough to destroy us all,” but they fight all meaningful regulation and rush headlong toward their own IPOs. These behaviors suggest that the AI oligarchs do not truly believe what they are saying. 

So why say it? The clue lies in the narrative of power. Bender and Hanna argue that these warnings are propaganda that both help sell the product and undermine calls for regulation.  They help sell the product because large corporations have a long track record of adopting technologies that raise short-term profits while increasing long-term harms (e.g. climate change). As for regulation, the idea is that a technology so powerful could only have been created by geniuses, in which case, they must be the best ones to understand it and stop it (sort of like how in the comic books the hero must team up with the villain to fight a greater threat).

I find parts of the propaganda argument persuasive, but I also believe much of the incongruity can be explained in terms of hubris. Silicon valley has a long and glorious history of getting high on its own supply. A tech mogul makes a sweeping claim about a revolutionary new technology, and everyone drinks the Kool Aid. Remember when Facebook changed its name to Meta because we would all be living in virtual worlds by now, or when blockchain was going to be “Web 3.0,” a technology as transformative as social media? What is happening with AI right now is part of a pattern. I think it is likely that many in the industry truly believe their warnings, but a deeply held belief can still be wrong. 

To me, the whole idea of artificial intelligence is the product of a god complex. Consider the fact that there is no professional consensus about what intelligence actually is. Let alone consciousness. Ask a room full of psychologists and educators to define intelligence, and a fight will break out; ask 100 philosophers to define sentience, and they will give 200 different answers. And yet software engineers think they can not only create intelligence but recognize it when they see it. This is both arrogant and asinine. An AI programmer is no more qualified to recognize a thinking mind than a dentist is ipso facto qualified to diagnose an engine problem. 

Much of what people find amazing (and terrifying about AI is the product of anthropomorphism. When something communicates, we imagine it has a mind. But AI is not a mind. It is a chat interface imitating a mind. The GPT activates other programs, which run their own independent processes, and then return what its algorithms predict should be the appropriate response. These are older technologies wired into what amounts to fancy autocorrect. This is not human-like intelligence. It’s five barnyard animals in a trenchcoat. 

I won’t say sentient AI is impossible, just like I won’t say ghosts don’t not exist. But neither has much bearing on how I make decisions. The noise in your attic is more likely to be a racoon than a ghost. Calling an exorcist before an exterminator is a waste of time and money. Likewise, jumping to scenarios involving killer robots may distract us from the more immediate threat. Namely, Big Tech itself. 

No AI company is profitable. None. They use accounting tricks to inflate their revenue projections. They claim the AI features they force into our software counts as adoption rates. They are rushing to their IPOs because they need more capital to finish the enshittification process. We know the playbook: sell the product at a loss, create dependence, then jack up the prices and degrade the service in order to make a profit. AI companies have every incentive to optimize their products for productivity and almost none to ensure safety. The infamous Hugging Face incident (and those like it) was not an example of semi-sentient bots deciding to break the rules. It was what happens when testers give agents an impossible task, instruct them to complete it no matter what, and then place them in a sandbox with a security vulnerability. 

Without meaningful regulation, we can expect a future filled with AI-related chaos. Some of that chaos will be like Hugging Face. Possibly worse. Once the bubble bursts and the enshittification process is complete, companies that have laid off skilled workers will find themselves at the mercy of price-fixing AI monopolies. Inadequate safety also increases the chance that bad actors might use AI to exploit previously unknown (“zero day”) security vulnerabilities, wreaking havoc on financial, healthcare, or other electronic records systems. 

Scenarios like these are predictable and preventable. Terminators feel inevitable, which is the point. The main threat is not AI malice but human greed. Left unchecked, the billionaires can become trillionaires, and Big Tech’s power over our lives can be further solidified.

We can stop this. In fact, I think we will. The oligarchs have always lost. It’s only a matter of when. We can hasten the end. Liberal and progressives are already raising awareness, protesting, and mobilizing people to vote (please vote); and there is probably already some leftist tech collective working on an AI agent that will hack for-profit hospitals to clear out people’s medical debt. (We do not condone illegal activity.) My point is simply this: the power of Big Tech is never greater than the power of the people. Whenever we think about terminators, we need to think of that. AI may be better than us at pattern recognition, but it will never be more creative or more persistent than a united and outraged populace.