U.S. Chief Technology Officer Michael Kratsios said Nov. 5 that more work is needed to realize “explainable” artificial intelligence (AI) technology, and said the government will offer further regulatory guidance soon on the technology’s use.
“The idea of explainable AI is extraordinarily appealing to a regulator…but we know that explainable AI doesn’t exist currently,” the CTO said at NVIDIA’s GPU Technology Conference. “There’s still incredible amounts of research that needs to be done” to reach that “explainable” state of the technology, he said.
Because of this, Kratsios said the Federal government will continue to invest in AI research and development (R&D), and his office plans to release a memo sometime in the next month that will provide some regulatory clarity regarding the technology. The coming memo, he said, will provide guidelines for the use of AI, and will seek public comment on those. Clarity on AI guidelines and standards, he said, will help knock down barriers to innovation.
Kratsios did not assign blame for the current lack of “explainable” AI, and similarly asserted that possible biases in AI algorithms – an often-stated concern of policy-makers and technologists – are not born of evil intent.
“[AI bias is] never because there’s malicious actors at the tech companies,” he said. Rather, “it’s because there are unintended consequences that we didn’t think of.” Further R&D investment in the technology, he argued, will help address unintended consequences of its use.
He encouraged continued conversation about AI use cases, and trust in American engineers working with the technology to iron out its wrinkles.
Earlier this year, the U.S. signed onto the Organization for Economic Cooperation and Development’s (OECD) Principles on Artificial Intelligence that proclaim a human-centered transition to automation.