Opinion
The Doomsday Clock’s been struck by an AI plot twist
Summary
- It’s still 90 seconds to midnight, but classic threats of extinction like nukes have been joined by an AI narrative of doom. We must pay attention as experts struggle to attain clarity.
The Doomsday Clock, a symbolic countdown to human extinction, stayed at 90 seconds to ‘Midnight’ this year after its recent annual review. This is the closest it has been to that ‘doom’ mark since it was created in 1947 following World War II by Bulletin of the Atomic Scientists, which updates the clock’s time every January to highlight human-created existential threats to the world. According to the Bulletin’s 2024 announcement, larger existential challenges are becoming harder to address, given the Russia-Ukraine war, our climate crisis (with 2023 designated as the hottest year on record), increasing sophistication of genetic engineering technologies and the dramatic advancement of generative AI (or artificial intelligence).
The clock was set at 90 seconds to midnight in 2023. This year, Rachel Bronson, CEO of the Bulletin group, stated that “the risks from last year continue with unabated veracity and continue to shape this year." Those threats included nuclear war, disease and climate volatility. The clock’s reading didn’t alter this year, but AI added a startling new menace to the list of existential threats. Apocalyptic stories about AI, which has been around for a while, have gained traction.
In the latest Doomsday Clock statement, John Mecklin, science and security board editor of the Bulletin, discusses how advancing generative AI (like ChatGPT) sparked expert discussions last year on existential risks posed by AI. Its applications for intelligence, surveillance, reconnaissance, training, simulation and military purposes, including for lethal autonomous weaponry, are under scrutiny. Further, “AI has great potential to magnify disinformation and corrupt the information environment on which democracy depends," Mecklin notes. “AI-enabled disinformation efforts could be a factor that prevents the world from dealing effectively with nuclear risks, pandemics, and climate change."
Sci-fi has nourished our imagination of AI robots for years. For instance, Skynet in The Terminator movie carries out a pre-emptive strike against humanity. However, is the advent of AI quite so apocalyptic?
Deepfakes are causing havoc in society. Nina Schick stated in her 2020 book Deepfakes: The Coming Infocalypse that deepfakes would lead to the biggest communications and information catastrophe in global history—an “infocalypse," that is.
In 2021, Google had fired two ethics researchers who co-authored an AI paper ‘On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?’ that discussed the potential risks of large language models and strategies for risk mitigation. Last March, in an open letter titled ‘Pause Giant AI Experiments’ in the midst of an AI gold rush, thousands of well-known signatories, such as Elon Musk, cognitive scientist Gary Marcus, author Yuval Noah Harari and Apple co-founder Steve Wozniak, asked for a six-month moratorium on the development of systems “more powerful" than GPT-4. In a New York Times op-ed, Harari offered an enthralling critique of Terminator-style situations: “Soon we will also find ourselves living inside the hallucinations of nonhuman intelligence." However, not everyone was on board with the AI doomsday narrative. Bill Gates, for example, seems optimistic. But in 2023, the world was stunned when one of the three ‘godfathers of AI,’ Geoffrey Hinton, raised concerns about the proliferation of false information, the possibility of AI overturning the job market and the “existential risk" that true digital intelligence would engender. Once again, not everyone agreed. Another ‘AI godfather’ Yann LeCun believes AI can usher in a “renaissance," for instance. In his 2019 piece, ‘Don’t Fear the Terminator,’ he said that an AI apocalypse was improbable. LeCun even compared the ‘pause’ letter to the Catholic Church’s 1440 call for a six-month moratorium on the use of movable type and the printing press!
In a 2019 study on the resources needed for training multiple popular big AI models, researchers at the University of Massachusetts, Amherst, found that training a single AI machine can release about 284 tonnes of carbon dioxide, which is roughly five times the lifetime emissions of an average American car. That was the GPT-2 era. Since then, AI advances have increased its carbon footprint significantly.
In Steven Spielberg’s 2001 film, A.I. Artificial Intelligence, set in the 22nd century, David, a childlike humanoid robot, gets left behind in the woods with a load of scrap metal and other waste. According to the UN Global E-waste Monitor study, e-waste is expected to reach some 75 million tonnes by 2030. So, on this measure, the 22nd century is upon us already.
Turning the clock back is never simple, though. And what can we do about the newest threat: AI? Despite its intriguing appeal, the proposal for a six-month pause was found to be unworkable. Therefore, humanity’s best chance may be to attempt turning AI green and impose strict regulations on its use and development. In Mecklin’s words, “Regardless, AI is a paradigmatic disruptive technology; recent efforts at global governance of AI should be expanded."
We cannot expect the doomsday narrative of AI to fade away. Many, including LeCun, might still be in disagreement with sceptics and many experts believe that focusing on AI apocalypse scenarios is a diversion that plays down immediate risks such as the large-scale generation of misinformation. This, however, only widens the haze of doubt.
