An executive from a large technology firm has just won a Nobel. Last Thursday, the top prize for chemistry went to the head of Alphabet Inc’s AI efforts, Demis Hassabis, along with two other key scientists, for a years-long project that used artificial intelligence to predict the structure of proteins.
The day before, Geoffrey Hinton, a former executive at Google who’s been called a godfather of AI, won the Nobel prize for physics along with physicist John Hopfield, for work on machine learning.
It seems the Nobel Foundation is eager to mark AI advancements – and the notion that key scientific problems can be solved computationally – as worthy of its coveted prizes. That will be a reputational boon for firms like Google and executives like Hassabis. But there’s a risk too that such recognition obscures concerns about both the technology itself and the increasing concentration of AI power in a handful of companies.
Hassabis himself has long craved this accolade, having told staff for years that he wanted DeepMind, the AI lab he co-founded and sold to Google in 2015, to win between three and five Nobel Prizes over the next decade. At a press conference, he called the award “an unbelievable honour of a lifetime” and said he’d been hoping to win it this time around. Indeed, he initially shaped DeepMind as a research lab with utopian objectives, where many of its leading scientists worked on building AI systems to help cure diseases like cancer or solve global warming.
But that humanitarian agenda faded to the background after the sale to Google and especially after the release of OpenAI’s ChatGPT, which sparked a race among tech giants to deploy chatbot-style technology to businesses and consumers. DeepMind has since become more product-focused (information about its healthcare and climate efforts disappeared from its homepage, for example), though it has continued with health-related efforts like AlphaFold. Out of DeepMind’s roughly 1,500-strong workforce, a team of just two dozen people were running the protein-folding project when it reached a critical milestone in 2020, according to a video documentary about the effort.
The Nobel will surely give Hassabis a credibility boost at Alphabet, where he has been leading the company’s fraught efforts to keep up with OpenAI. Google’s flagship AI model Gemini has grappled with controversies over its frequent mistakes and the possibility it will choke off traffic to the rest of the web. Now perhaps a smoother path has been paved for Hassabis if he wants to become Alphabet’s next chief executive. The former chess champion is a consummate strategist and rivals Sam Altman as the world’s most successful builder of AI technology, having pushed the boundaries of fields like deep learning, reinforcement learning and games-based models such as AlphaGo, which beat world champion Go players eight years ago. Hassabis was already talking about taking on protein folding during those matches.
The glow benefits Google, too. Recent challenges from antitrust regulators over monopolistic behaviour haven’t helped its reputation as a company founded on the principle of “don’t be evil.” Now with two Nobel prizes linked to work done by its scientists, the tech giant can more easily frame itself as providing services that are ultimately good for society, as its lawyers have been arguing, and perhaps generate goodwill more broadly with the public and regulators.
But we shouldn’t forget the tension between the high-minded goals professed by Big Tech and what their businesses are really focused on. Google, which derives close to 80% of its revenue from advertising, is now putting ads into its new AI search tool. For businesses, that invites a new layer of complexity to online advertising, while consumers face the prospect of wading through AI-generated information that Google is trying to monetise and which could one day become more biased toward advertisers.
Remember also that Google’s prioritisation of human wellbeing was called into question less than three years ago when it fired two leading AI ethics experts who’d warned about the risks that its AI models could entrench bias, spread misinformation and hoard energy, issues that haven’t gone away. A September study in Nature, for instance, showed that AI tools like ChatGPT were making racist decisions about people based on their dialect.
The Nobel Prize is designed to recognise people who’ve made outstanding contributions to science, humanism and peace, so the Foundation behind it has taken a bold stance in validating the work of AI and of one company in particular. The award to Hassabis – like the peace prize given to Barack Obama one year after he was elected as president – feels a little premature. It’s still unclear what kind of broad, real-world impact DeepMind’s protein-folding project will have on the medical field and drug discovery.
Let’s hope the prize motivates well-endowed technology firms to invest much more in using AI for public service efforts like protein folding and in AI ethics research – and doesn’t muddy the debate over the very real risks that AI poses to the world too.
-Bloomberg Opinion/Tribune News Service
-TheStar