Google bullish on AI benefits in the face of fears over unchecked growth

William Malcolm, senior director of international privacy legal and consumer protection at Google, cut an isolated figure speaking at the AI Regulation Summit in London hosted by City & Financial Global today (October 5).

Flanked by other lawyers and finance professionals, he was the sole member of the panel representing a Big Tech firm, the likes of which have been leading the rapid growth of generative AI in finance and beyond. Without the technologist, the panel could have easily devolved from listing the rational concerns of human input in AI systems and potential ethics violations into discussing various science fiction plotlines concerning AI.

Indeed, there was a point in the cavernous Cavendish Conference Centre where both Skynet and Terminator were name-checked as features of potential realities if unregulated AI usage were to continue. Both the audience and other members of the panel had even speculated on a potential “extinction risk” to humans once AI begins to think independently.

That these examples were touted seriously made the decision to have a delegation from Google on the panel interesting. Google’s usage of AI has been wide-ranging for many years, operating across most of its major applications, such as Translate, Maps and Lens. The Google AI division has research facilities in three continents, and Google’s 2017 innovation Transformer forms the basis for ChatGPT, and with its chatbot Bard, Google recently entered the generative AI space.

When asked about his stance on AI regulation, Malcolm was bullish on emphasizing the positives AI brings to a range of industries, while respecting the need for regulation.

“At Google, we want to be bold and responsible when it comes to using AI,” he said. “We want to do it together with regulators, civil society, academics, and partners.”

Malcolm pointed to the positive impact of AlphaFold, a research initiative conducted by Google DeepMind, which is able to accurately predict 3D models of protein structures and is accelerating research in human biology at previously untold speed.

“[AlphaFold] saved the equivalent of 400 million years of research time in just a few weeks. That’s the scale of the opportunity,” Malcolm said.

He also addressed the issue of hallucinations, an issue with chatbots in which the AI can malfunction and ‘hallucinate’ content stated as fact but is actually gibberish. Sometimes these answers appear plausible in the context of an otherwise factual answer, which can confuse users and cause headaches for companies that seek to make generative AI a key part of their business model.

Google Bard’s website lists hallucinations as a limitation of its product, but Malcolm says the problem is somewhat unavoidable due to Bard’s role in generating new information rather than repeating old information.

“It’s a feature, not a bug, right? Generative AI is just that. It’s not a knowledge-based product, it’s generating information. I thought that was clear,” he said. “Human oversight cannot be in every aspect of the assessment, so you have to get a balance.”

Earlier this year, Google urged financial regulators and market participants to re-evaluate their established risk management guidance as it pertains to AI. Malcolm believed that while there is indeed much to be concerned about in terms of risk management, he says that humanity runs risks to its benefit and progress by not deploying these emerging technologies, just as much as deploying them could be a threat.

Malcolm says that a strategy taking into account these risks is the best solution at the moment for companies looking to use AI.

“Everybody wants to get the benefits of AI and manage the risks,” he says, “but everyone accepts that those risks can’t be reduced to zero.”