Bill Gates: AI Crossed Bio and Cyber Thresholds This Year
Bill Gates states AI reached two concerning milestones this year, in both biological and cyber domains. During an interview with Ezra Klein for The New York Times, Gates expressed his belief that AI's capabilities now pose significant risks.
“We crossed the cyber threshold and we crossed the bio threshold early this year,” Gates emphasized.
On the cyber front, he highlighted AI models capable of identifying security vulnerabilities in code that humans have reviewed for over 20 years, leading up to Anthropic's Mythos. In biology, Gates noted the shift in power to create novel pathogens from nation-states to small groups utilizing advanced AI tools.
Challenging the prevailing view in Washington, Gates argued against relying solely on product liability and company interests for AI safety. He compared AI to drugs, aircraft, and cars, which all operate under safety regulations, not just lawsuits.
“You can’t rely on the industry to self-regulate here. I mean, it’s just insane,” Gates stated.
He proposed a layer of safeguards and monitoring for all AI models to prevent misuse, allowing open-source models while ensuring they run on regulated platforms. Gates also criticized Anthropic's filtering of its models, Mythos and Fable, which he claimed can direct users seeking cancer information to less reliable models.
Regarding job displacement, Gates noted that AI has so far created more jobs than it has destroyed, but he predicts coding, accounting, legal work, telesales, and customer support will be affected within a few years. He suggested companies should pay the same payroll tax for AI and robots as for human workers and introduced the concept of “Human Reserved” jobs, such as childcare and eldercare, exclusively for humans.
The Gates Foundation's annual strategy reviews, scheduled for October, will determine how to allocate its $10 billion budget. This year, ChatGPT, Claude, and Copilot will participate in these reviews, following Gates' recent warning on Meet the Press that AI could potentially cause a billion deaths.