Deutsche Bank Pours Cold Water on AI Doomsday Theories: Expert Predictions of the Future May Not Be More Reliable Than Those of Ordinary People

Beating AI News Flash: Deutsche Bank Research has released "The AI Doomsday: A Brief History of Bad Tech Predictions," pouring cold water on the recent AI extinction debate. The report argues that it is already difficult to predict how far technology can develop, let alone when it will be realized, how society will adopt it, and what impact it will ultimately have. At present, the judgments of Silicon Valley AI experts about the future of AI are not necessarily more reliable than those of ordinary people.

The report unearthed a string of prediction failures spanning nearly a century. Einstein believed in 1934 that nuclear energy was almost impossible to obtain in practice, yet within less than a decade, a sustained nuclear chain reaction was achieved. Bob Metcalfe, co-inventor of Ethernet, predicted in 1995 that the internet would "spectacularly collapse" the following year. When the prediction failed, he pulped the article containing that forecast and ate it.

The AI industry itself has also had no shortage of failures. Geoffrey Hinton judged in 2016 that deep learning would surpass radiologists within 5 to 10 years, and even suggested stopping the training of new entrants. But over the past decade, the number of radiologists has instead grown by about 10%. Deutsche Bank believes such predictions often make one mistake: treating the most easily automated part of a job as the entire job. Radiologists do not just read scans; they also provide explanations, judgments, communication, and clinical decision-making.

Predictions about autonomous driving encountered the same problem. Technological progress is only the first hurdle, followed by endless edge cases, regulation, allocation of liability, and public acceptance. Deutsche Bank even argues that how far AI can ultimately go may depend more on whether companies can truly integrate it into workflows and are willing to pay, rather than on how much the next model's capabilities improve.

The report also adds a more realistic explanation for AI doomsday rhetoric: almost all participants have incentives to make things sound bigger. Founders need faith, investors need a market narrative, consulting firms need urgency, policymakers need relevance, and the media need drama. Social platforms are also naturally biased toward negative, emotionally intense, and more extreme statements, while moderate judgments are harder to spread.