"AI will be able to defeat you, personally, very soon."

AI isn't going to end civilization, but that claim is a great way to make it seem more powerful than it is.

Link: Anthropic Researcher Says Pretty Much All Humans Will Become "Permanently Unemployable" After Its IPO, by Joe Wilkins in Futurism

Tao Lin, an Anthropic safety researcher, in a recent interview recounted by Futurism:

“I think that it’s very likely that pretty much all humans on Earth will become unemployable,” the researcher said. “I want to tell people that AI will be more capable than you, personally, very soon. It will be able to defeat you, personally, very soon.”

This is such a pessimistic, one-dimensional view of humanity. It treats our lives as a series of flat value exchanges, reducing our efforts to scalar measures of productivity that can’t capture all that is valuable about the world. AI is not better than you.

But AI doesn’t have to be better than you to take your job. It only has to be good enough, and cheap enough, that someone decides that it’s profitable to replace you, just as McClatchy just did to its newsrooms. That’s an entirely human decision.

AI is genuinely useful for some core tasks, and it will become more useful. It will change some jobs and remove others; it will also create new ones. It will be transformative in the same way that spreadsheets were transformative.

But it can’t write a poem that speaks to the heart, or an essay that captures a topic so deeply that you understand the human implications beyond the immediate mechanics. As Futurism points out, there’s so far very little evidence of the widespread AI-driven jobs collapse Lin is predicting. And other doomer predictions are even more out there; it will not recursively self-improve until every last molecule of biomass has been consumed by its nanobot horde.

The danger isn’t AI: it’s the people who deploy it without understanding what it is and isn’t capable of, and the people who seek to devalue human work for their own benefit.

We are not safe from people who decide to lay workers off in favor of an LLM. We are not safe from people who overestimate AI’s capabilities and use it for tasks where it isn’t suitable. We are not safe if someone puts AI in front of our nuclear weapons, if our natural resources are consumed by data infrastructure, or if it’s negligently designed and negligently deployed. Those risks are real, but they’re human. We are safe from a science fiction super intelligence that has so far not manifested.

It will not be able to defeat you, personally, any time soon. But telling you that it will, days before the AI vendor’s IPO, is a great way to make the product seem more powerful and more valuable — and make you feel like if you don’t use it, you’ll be existentially left behind.