AI releases pandemics and bioweapons

Along with all the praise for the rapid advancement of artificial intelligence comes an ominous warning from some of the industry's top leaders about the potential for the technology to backfire on humanity. Some warn "AI could become powerful enough that it could create societal-scale disruptions within a few years if nothing is done to slow it down," The New York Times said, "though researchers sometimes stop short of explaining how that would happen." A group of industry experts recently warned AI technology could threaten humanity's very existence. "Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks, such as pandemics and nuclear war," reads the one-line open letter released by the Center for AI Safety, a nonprofit organization. The statement was signed by more than 350 executives, researchers, and engineers from the AI sector, including Sam Altman, chief executive of OpenAI, and Geoffrey Hinton and Yoshua Bengio, two of the Turing award-winning researchers considered the "godfathers of AI." The message is foreboding, but also vague, failing to provide any details about how an AI apocalypse could come about. What are the commentators saying? One plausible scenario is that AI falls into the hands of "malicious actors" who use it to create "novel bioweapons more lethal than natural pandemics," Dan Hendrycks, the director of the Center for AI Safety, wrote in an email to CBS MoneyWatch. Or these entities could "intentionally release rogue AI that actively attempt to harm humanity." If the rogue AI was "intelligent or capable enough," Hendrycks added, "it may pose significant risk to society as a whole." Or AI could be used to help hackers. "There are scenarios not today, but reasonably soon, where these systems will be able to find zero-day exploits in cyber issues, or discover new kinds of biology," former Google CEO Eric Schmidt said while speaking at The Wall Street Journal's CEO Council. Zero-day exploits are security vulnerabilities hackers find in software and systems. While it might seem far-fetched today, Schmidt said AI's rapid advancement makes it more likely. "And when that happens," he added, "we want to be ready to know how to make sure these things are not misused by evil people." Advertisement Advertisement - Article continues below Another worry is that AI could go rogue on its own, interpreting the task for which it was originally designed in a new and nefarious way. For example, AI built to rid the world of cancer could decide to unleash nuclear missiles to eliminate all cancer cells by blowing everyone up, science journalist and author Tom Chivers said in an interview for Ian Leslie's substack, The Ruffian. "So the fear is not that the AI becomes malicious; it's that it becomes competent," Chivers deduced. AI tools are only "maximizing the number that you put in its reward function." Unfortunately, "it turns out that what we desire as humans is hard to pin down to a reward function, which means things can go terribly wrong."