Google CEO Sundar Pichai has struck a cautionary tone in an interview with the Financial Times about the future regulation of artificial intelligence.
Speaking in Helsinki, Pichai warned against broad regulation of AI, arguing instead that existing laws could be repurposed sector by sector "rather than assuming that everything you have to do is new."
"It is such a broad cross-cutting technology, so it's important to look at [regulation] more in certain vertical situations," Pichai said.
Read more: A former Google engineer warned that robot weapons could cause accidental mass killings
"Rather than rushing into it in a way that prevents innovation and research, you actually need to solve some of the difficult problems," he added, citing known issues with the technology such as algorithmic bias and accountability.
Pichai's comments are set against the backdrop of Google's own rocky relationship with artificial-intelligence technology and a backlash to some of its projects.
The company canned a drone contract with the Department of Defense — codenamed Project Maven — after fierce employee backlash, with thousands signing a petition and multiple resignations. Engineers felt the project was an unethical use of AI, and Google allowed its contract to lapse in March.
But artificial intelligence remains a core future growth plank for Google more generally.
Pichai said in a first-quarter earnings call this year that almost three-quarters of Google's advertising clients were using automated ad technologies, which use machine learning and AI. The firm makes the bulk of its revenue from advertising, reporting $33 billion in revenue for the second quarter. Google's fast-growing cloud business also uses AI. And its new health division is part of the firm's efforts to commercialize research produced by its AI lab, DeepMind.
Laura Nolan, a former Google engineer who was recruited to Project Maven and subsequently resigned over ethical concerns, told Business Insider via email that while Pichai might be speaking out of self-interest, she agreed that blanket AI regulation might be too broad an approach.
"Laws shouldn't be about specific technological implementations, and AI is a really fuzzy term anyway," she said.
Sandra Wachter, a professor at the Oxford Internet Institute, also told Business Insider that the contexts in which AI was used changed how it should be governed.
"I think it makes a lot of sense to look at sectorial regulation rather than just an overarching AI framework," Wachter said. "Risks and benefits of AI will differ depending on the sector, the application, and the context. For example, AI used in criminal justice will pose different challenges than AI used in the health sector. It is crucial to assess if all the laws are fit for purpose." She added that in some cases existing laws could be tweaked, such as data protection and nondiscrimination laws.
Both Nolan and Wachter, however, cited specific areas that might require new laws — such as facial recognition.
"Algorithmic decision-making is definitely an area that I believe the US needs to regulate," Nolan said. "Another major area seems to be automated biometric-based identification (i.e., facial recognition but also other biometrics like gait recognition and so on). These are areas where we see clear problems emerging due to new technological capabilities, and not to regulate would be to knowingly allow harms to occur."
"The use of AI in warfare absolutely needs to be regulated — it is brittle and unpredictable, and it would be very dangerous to build weapons that use AI to perform the critical function of target selection," she added.
