Hello again and welcome back to Fast Company’s Plugged In.
Once a year, Google holds a conference on its campus called Research@. Scientific researchers from inside and outside the company pack an on-campus auditorium for a highly compressed look at what its 26-year-old Google Research lab has been up to. This year’s edition emphasized AI—surprise!—and covered a vast amount of ground, from predicting natural disasters to fact-checking technical documents to helping doctors better serve patients.
After the keynote, I sat down with James Manyika, the member of Google’s senior leadership team who oversees Google Research. (His title—senior VP of Research, Labs, Technology & Society—conveys the breadth of his responsibilities.) He told me about Google Research’s mandate and approach to its work, and why he believes AI has already resulted in more scientific progress than most people realize. Our conversation has been edited for length and clarity.
For the sake of people who didn’t get to attend this conference, talk a bit about what Google Research is all about. You’re part of Google and Alphabet, and there’s research going on all over both companies beyond your group.
Google has had Google Research since almost the very beginning. The founders initially liked the idea of having a research organization, almost like Bell Labs. That’s what Google Research has been.
The way we think about the mandate and scope of Google Research is both foundational and applied research in a wide range of areas: things to do with fundamental computer science, AI, and machine learning, but also science itself. A lot of the work informs and helps our product areas. Things like how we think about knowledge graphs and search and retrieval of information. The fundamental computer science that goes into designing our infrastructure and hardware. The team that invented the transformer was in Google Research, so all of that too.
The parts that people aren’t as aware of are the scientific work. My favorite example is the fact that Google Research has had a genomics program for the last 12 years. We’ve had a connectomics program, to understand the wiring of the brain, for 10 years. And we’ve had the moonshot to build a quantum computer for 13 years.
One other thing that Google Research has as its mandate is a fair amount of translational work. The science and the research is fantastic, but a lot of work in crisis response, community resources, and resilience happens in Google Research.
Some of the stuff on the stage this morning involved issues like firefighting and medical AI. To what degree are you focused on these kinds of big important jobs for the world that are not always directly related to Google’s core as a business?
To me, all of those are fundamentally related to Google’s mission sense, which is organizing the world’s information and making it universally accessible and useful.
Things like flood forecasting are actually part of a moonshot to say, “What if no human or computer on Earth was surprised by a natural event?” Then that translates into, “Okay, so what are the natural events?” It’s a list that includes everything from earthquakes to floods to wildfires to monsoons to cyclones. We’re trying to work toward all of those. We’ve made more progress in some than others.
It’s the same thing with health. What if all diseases could be diagnosed, detected, understood, and eventually treated and cared for? What would that look like? That drives the fundamental research, the work on language and accessibility. Again, that goes back to the mission, which is, if you’re going to organize all this information and make it useful, boy, it better be in all the languages. And it better [overcome] every kind of access barrier, which we think of not just as language, but also things like [helping] people with disabilities, for example, speech challenges.
One moonshot right now is supporting 1,000 languages. A couple weeks ago, we hit our most recent milestone, which is 300 languages. We’re lucky in the fact that this is a company that’s wired to think about how we do these ambitious things.
A lot of the news onstage involved jobs that AI is already good at. But there was also discussion of it not always being as accurate as we’d like, and that it may disrupt jobs and the economy in ways that are going to cause issues. What percentage of your work and thinking relates to AI potentially causing problems and trying to avoid them?
A lot. In fact, we spend probably as much time as we do on the exciting possibilities. We think of them in a few different categories—the performance challenges and risks that could cause harm.
If you think about hallucinations, that could be problematic. But we also think about other categories of challenges and risks that have to do with misuse of this technology. That’s why you see us invest across Google Research and Google DeepMind on things like privacy and SynthID, which we’ve expanded beyond our own systems.
Are there technical solutions and innovations related to misapplication and misuse? Of course. But you need more than just that. You need society to come to collective views about how we deal with those things. We try to advance the ambitious, useful things, but also work on the downsides and the complexities that come with this.
How do you engage with researchers outside of the company?
I’ll give some concrete examples. I mentioned connectomics, the wiring diagrams of brains. We’ve had a deep collaboration with Howard Hughes Medical Institute and also the Lichtman Lab at Harvard for over a decade. When we published recently the work on Drosophila, the fruit fly, it’s the result of 10 years of deep collaboration. The work that Quantum team does has been a decades-long collaboration with academics at Caltech and University of Chicago.
Michael Brenner, who was on stage, is a full professor in applied mathematics at Harvard. He’s been there for 25 years. He’s also been a fellow here for the last 10.
There are a lot of people out there who are worried about AI, or skeptical, or just don’t see value in it. If they understood some of these scientific accomplishments, would they feel better?
It will help, but but it won’t be enough. I just think these scientific things need to be more understood. But the indirect nature of that will always make it arm’s length, not like a chatbot you interact with every day.
I think the bigger challenge and opportunity that we have is with regards to things that people engage with tangibly and directly. When it comes to work, the labor market effects, and education, people will experience that quite directly. That’s where we have to make a lot more progress.
I think it doesn’t help if some in the field itself make pronouncements that aren’t being borne out. The one that I find worrisome is the labor market impact, because most of the apocalyptic pronouncements on that front haven’t happened yet, and it’s unclear whether they will happen.
That’s not to suggest that we should sit back and relax. I think we should think about those changes in the labor markets and plan for them. Training is part of it. We should think about the role of safety nets.
Those direct things are probably a big part of the anxiety and concerns that people have. We have to make progress on those things and show that in fact there are paths through this.
You’ve been reading Plugged In, Fast Company’s weekly tech newsletter from me, global technology editor Harry McCracken. If a friend or colleague forwarded this edition to you—or if you’re reading it on fastcompany.com—you can check out previous issues and sign up to get it yourself every Friday morning. I love hearing from you: Ping me at hmccracken@fastcompany.com with your feedback and ideas for future newsletters. I’m also on Bluesky, Mastodon, and Threads, and you can follow Plugged In on Flipboard.
More top tech stories from Fast Company
Building an AI data center costs billions. Defending it is another story
From Russia to the Persian Gulf, drone attacks are forcing governments and technology companies to rethink how they protect their most valuable computing infrastructure. Read More →
Tin Can’s next product brings a relic of phone culture to Gen Alpha
The brand that made landlines cool again is introducing the Rolodex to a new generation. Read More →
Nintendo just made an unusual decision—and it’s going to be good for business
As Sony moves to phase out physical game discs, Nintendo is reassuring fans that cartridges aren’t going anywhere. But its commitment comes with some caveats. Read More →
What happens when your employees’ AI agents start coming to work? (Spoiler: It’s bad)
With 82% of organizations running AI agents they don’t know about, personal agents like Muse and Dots are about to make a hidden problem much bigger. Here’s what to do now. Read More →
He helped build Meta’s smart-glasses technology. Now he wants to take on Big Tech’s grip on personal AI
With Awear AI, Ash Saulsbury is building a platform that would let smart-glasses users take their AI assistants’ memories across devices and brands. Read More →
Trump’s new ‘superintelligence’ website is a rebrand fail
The newly branded SI.gov website has a problem: Its messaging is out of sync with public opinion. Read More →