Technology revolutions in the digital age are typically accompanied by optimism and excitement — recall Steve Jobs basking in thunderous applause as he introduced the iPhone in 2007. The major A.I. companies seem to be following a darker and weirder strategy: They like to solemnly describe the harms that their models will cause, while acting helpless to do anything about it.
Anthropic recently dropped a classic of the form: a scary-sounding report titled “When A.I. builds itself” that claims A.I. could be moving closer to the capability of “autonomously designing and developing its own successor.” The company hopes that this recursive self-improvement will bring “enormous good” to the world, but also openly worries it might lead to humans “losing control” of these systems.
The public reaction to this report focused on a section that seemed to call for a worldwide pause on A.I. development. But if you read more carefully, it becomes clear that a pause isn’t actually what Anthropic proposes. Its report says that “if it were possible” to slow down the technology, then we should, but so long as the “least cautious” actors were advancing full speed, it suggests that Anthropic will have no choice but to do the same.
Like a cat leaving a dead bird at your doorstep, Anthropic catalogs the grim future that its products might produce, shrugs its shoulders and then returns to its furious efforts to make these warnings a reality.
Anthropic isn’t alone in this nihilism. Sam Altman, the chief executive of OpenAI, frequently makes bleak claims. He compared the company’s large-language-model efforts to the development of the atomic bomb, and he posted an image on social media of the Death Star from “Star Wars” in conjunction with the release of GPT-5. He’s also argued, many times, that the only response to the economic damage that A.I. tools will inevitably produce would be to institute wide-scale financial support from the government, such as a universal basic income or a public wealth fund.
Let’s call this strategy “doom trolling.” It’s one of the defining and most arresting properties of our current A.I. moment, and I’ve come to believe that it’s morally indefensible.