Why the odds are against AI doomers

In 1968, biologist Paul Ehrlich warned in his book The Population Bomb that feeding humanity would be impossible, predicting millions would starve despite efforts. However, the Green Revolution doubled major cereal crop yields between 1960 and 2005, while real food prices plummeted by over half. When asked about his 1968 predictions in 2015, Ehrlich said his language would be even more apocalyptic today, comparing the current threat to AI.

Surprisingly, AI experts are far less concerned than the general public. Pew Research data from 2024 revealed 51% of Americans felt more threatened by AI than excited, while only 15% of AI specialists agreed. Over half of the specialists anticipate AI having a positive impact on the US within 20 years, versus just 17% of the public.

Despite these concerns, a survey of 2,778 AI researchers found a median 5% chance of catastrophic outcomes, with 68.3% believing positive outcomes were more likely. Examples of AI success include Google DeepMind's AlphaFold mapping 200 million protein shapes, earning a Nobel Prize; AI-assisted breast cancer screenings catching 29% more cancers with 44% less radiologist work; a UC San Diego AI reducing hospital sepsis deaths by 17%; and an AI assistant boosting customer support productivity by 14% and novice workers by 34%.

These advancements, such as Microsoft's AI correctly diagnosing 83% of complex medical cases against 36% for generalists, highlight AI's potential benefits. AI is also being used in public health, with a Malaysian AI developed by the Ministry of Health screening diabetics for retinopathy in about 30 seconds. While risks exist, researchers and authorities are working to mitigate them. Fear should guide prudent actions, but hope should drive progress.