- Published on
Why a decade of doomsday warnings failed to slow the AI race
- Authors
- Name
- DeepHorus
A decade of dire warnings that artificial intelligence could end humanity has failed to slow the industry's breakneck race toward ever more powerful systems. The latest tremor came when Anthropic safety researcher Jacob Coxon resigned and declared that leading AI labs are "racing straight to self-improving super intelligence and gambling with our lives." His 9 September tweet went viral and was echoed by dozens of staffers at Anthropic, OpenAI and rival firms.
The warnings are not new. In 2014, Stephen Hawking said AI development "could spell the end of the human race." At that time, generative AI was in its infancy, and the idea that machines could perform today's tasks, let alone become superintelligent, felt largely hypothetical. The public had not yet encountered ChatGPT's original generative features, which arrived nearly a decade later.
The industry has developed a recurring pattern. Every few years, researchers and tech leaders issue somber warnings about irresponsible AI development. Instead of pausing, a faction breaks away to found a new AI company. This cycle has repeated without any sustained slowdown.
Coxon's resignation, however, created unusual reverberations. Dozens of employees from competing labs echoed his concerns, and days later the CEOs of leading US AI firms responded to Anthropic CEO Dario Amodei's call to slow what he called a "reckless" approach. Amodei wrote in a blog post: "AI brings risks, and because it is such a powerful technology, these risks are serious."
The episode underscores a central contradiction: the same people building the technology are publicly warning that it may be catastrophic. Yet the competitive pressure among frontier labs continues to drive rapid deployment. No major lab has agreed to a meaningful pause.
Amodei acknowledged he has written extensively about these risks, but his call for restraint produced no halt in development. The industry appears locked into a race where warnings serve as a backdrop rather than a brake. As superintelligence grows closer, the gap between alarm and action remains as wide as ever.
