AI is advancing at a seemingly exponential pace, but the regulatory environment is lagging, creating an “urgent” need for leadership, according to Google DeepMind’s Demis Hassabis.
The Nobel Laureate wrote in a LinkedIn essay this week that the U.S. is primed to lead worldwide efforts to mitigate the technical risks of AI as it proliferates across workforces and in the public sphere.
“Currently, as a field and as a wider society, we aren’t doing that,” Hassabis writes.
In its current incarnation, AI is already delivering significant returns, but the next phase of the technology—dubbed Artificial General Intelligence, hypothetically thought to exceed human intelligence—could create medical breakthroughs, new energy sources and more. But those outcomes need robust safeguards in place today.
“We have to navigate this critical period of development thoughtfully and carefully,” Hassabis writes. “Urgent action is needed to address risks that might arise as we get closer to AGI.”
See also: Why good governance is the key to keeping transformation alive
A new ‘Frontier’ for AI governance
Hassabis proposes a U.S.-led standards body, either as a self-regulatory entity or derived from a public-private partnership.
The framework Hassabis describes involves the designation of “Frontier Model” organizations to test AI in relation to issues like national security. Their “Frontier Labs” would develop and share best practices with an emphasis on cybersecurity, talent vetting and more.
“Specific agentic AI tests could look for attempts to bypass safety guardrails or signs of deception,” he writes, “and ensure best practices, such as digitally watermarking AI-generated images and generating human-readable output tokens to understand model reasoning.”
Their models would then be assessed by the standards body.
It’s an approach, Hassabis says, that would be technically focused, yet designed to drive innovation and incentivize “responsible behavior.”
“Being designated a Frontier Lab would carry significant prestige and be open to any organization by building models that meet the benchmark criteria,” he writes.
The emphasis on collaboration among stakeholders mimics other recently announced efforts calling for partnerships to drive responsible AI innovation.
For instance, RAISE US will bring together state governments, AI firms and major employers, including Amazon, Anthropic and Microsoft to build the “workforce response to AI.” The initiative involves reskilling and redeployment efforts to support talent in an AI-powered economy.
“What we collectively do now will determine how the next phase of civilization unfolds,” Hassabis says. “By safely stewarding AGI into the world, we can enter a new golden age of scientific discovery and progress, and usher in a bright future of incredible human flourishing.”
Credit: Source link
