AI could lead to apocalypse and human extinction by 2100, new study says

A new study by the Lancet Commission on 21st Century Global Threats to Health warns that the malicious use of artificial intelligence (AI) could potentially lead to human extinction by 2100. The researchers say governments need to take these risks seriously and act early to prevent AI from being misused.

In Short

• Commission identified 17 interlinked threats across health, climate, conflict and technology
• Researchers used models across 204 countries to compare best and worst scenarios
• Preventing nuclear conflict could avert 133 million deaths and 9.35 billion DALYs

Artificial intelligence (AI) could make humans extinct by 2100. Researchers from the Lancet Commission on 21st Century Global Threats to Health warn that the malicious use of AI could pose a serious threat to humanity. This could include using AI to develop biological weapons or cause other forms of catastrophic harm.

The findings come from the Lancet Commission on 21st Century Global Threats to Health, which examined 17 major threats that could affect human health and survival between 2025 and 2100. The commission says the future will be shaped not by a single crisis, but by a combination of risks involving technology, conflict, climate change, disease, and social and economic conditions.

While researchers say the chances of this happening remain low, the consequences could be devastating if such a scenario becomes reality.

How AI could threaten human survival

As part of the assessment, the researchers identified two major threats that could potentially wipe out humanity by 2100. First is the malicious use of AI and second is a nuclear conflict. An accompanying editorial in The Lancet, titled Rethinking threats to human survival, warns that if either scenario becomes reality, “the scale of harm ... would inflict an apocalypse on humanity”.

The researchers say both threats could have devastating consequences, although the chances of either leading to human extinction remain low. They warn that these risks should not be ignored and that steps need to be taken to prevent them.

However, the researchers note that it is difficult to estimate how likely these threats are, as there is limited historical data to draw on.

The assessment also looks at what could happen if governments take steps to reduce these risks. It estimates that preventing the malicious use of AI could save the equivalent of 10.4 billion years of healthy life by 2100. The report uses a measure called disability-adjusted life years (DALYs), which accounts for healthy years lost because of illness, disability or premature death. Preventing a nuclear conflict, meanwhile, could avert an estimated 133 million deaths and 9.35 billion DALYs over the same period.

17 threats could shape humanity's future

AI and nuclear conflict are not the only risks highlighted in the assessment. The researchers examined 17 threats across 204 countries and territories, including climate change, pandemics, air pollution, drug-resistant infections, conventional wars, childhood malnutrition, obesity, poor mental health and income inequality. The idea is to understand how these problems could affect people's health and lives over the coming decades.

The commission, co-chaired by Professor Christopher JL Murray of the University of Washington and Dr Natalia Kanem, says the threats are interconnected and need to be addressed together.

“The striking finding is not only the scale of these threats, but that humanity still has the agency to act,” Murray said.

The commission recommends building health systems that can respond to emerging threats, investing in scientific innovation and improving the monitoring of global risks.

- Ends