It’s the movie of the week in Silicon Valley. A decade after the premiere of Her, many no longer see this film as a work of science fiction, as it now reflects too much of today’s reality–especially after two major AI companies unveiled new projects a few days ago (GPT-4o and Astra) that clearly bring Samantha, the virtual assistant voiced by Scarlett Johansson, to mind.
The anticipation surrounding the newest AI voice agents is so high that even Google CEO Sundar Pichai felt the need to share his thoughts on the matter. In a session with international journalists held during the Google I/O 2024 event, where Xataka On was present, Pichai was asked about the future of humanity’s relationships with AI, Google's new glasses, and the swift integration of Gemini throughout the company. Google has called it “a new era.” This is how the man at the top explains it all. This is how Pichai, the man in charge of this era, explains it all.
First, Pichai points out that Google has “definitely been on this AI journey for a long, long time, and it’s great to see us actually realize the benefits of AI across our products.” Even though the pivotal moment came with the arrival of OpenAI’s ChatGPT, Google seizes every opportunity to point out that it was a pioneer in this space long before. DeepMind’s team was indeed doing amazing work at the time, although it leaned much more towards the academic world than to the everyday Google services that people use.
The Year Where We’ll Talk to AI (And Where It Will See What We’re Doing)
“Given the scale of our products, we think we can really help users at a pretty big scale,” Pichai said. “Our Gemini models, they were built natively in multimodal... And that really starts coming to life when you see something like Project Astra.”
Xataka On also spoke with Sergey Brin, Google’s co-founder, who echoed this sentiment. Brin stated that he hadn’t expected the company’s experiments to get this far, referring to how Gemini was originally created with text in mind. “This is a vision we had while we were one of the first to, on the ground up, train a natively multimodal model,” added Pichai, emphasizing that Google isn’t following other companies, even if its presentations take place later.
Being multimodal is what paved the way for what we are witnessing this year: humans engaging in actual conversations with AI models. The New York Times was among the first to declare that the era of Her was upon us, and even though–as Spanish outlet El País points out–this comparison is mainly a marketing strategy to hype the possibilities of AI, it is fun to discuss its social implications.
“I feel like I need to say that I love my wife before I say any more,” Pichai joked before discussing Her. “With every technology, as you make progress, there are dual use cases,” he continued. “There are people who will use it to preserve memories of loved ones. You’re going to have technology that powerful. And yes, over time, you will have people getting into deep relationships with these AI assistants or agents. So I think, yes, we should prepare ourselves for all those possibilities. I think this is why we want to approach the technology in a bold and responsible way. We are committed to making progress... A great example is our watermarking [tool] Synth ID, which we have done for images. We’re bringing it to text and video as well.”
Pichai hasn’t fallen in love with his AI assistant–but he does love AI. He explains how it all started: “I saw a demo of the Brain team, what is now DeepMind. They were showing how a new technology, or a new neural network, could recognize an image of a cat. That was the first time there was an ‘aha’ moment in my brain, you know, ‘this technology is going to work,’” the CEO recalled.
At this year’s Google I/O—which Pichai jokingly described as the company’s version of Taylor Swift’s Eras Tour, but “with fewer costume changes”—there were other announcements beyond multimodality. Aside from working with voice and video, the company's AI will also be able to provide better context. This is exemplified by Gemini Pro (currently available in preview), which supports up to two million tokens. The memory and context available to AI will be significantly larger, allowing for longer conversations with more factors at play–another parallel to Her.
Pichai also defends the concept AI agents. He talks about the actions they'll be carry out for us, “be it search, doing complex queries, or Gemini trying to plan a trip for you, or over time, Chrome being able to, in the background, help you on projects... I think we’re well positioned because of our research leadership, infrastructure leadership, innovation and search, our global product footprint.”
Google Might as Well Change Its Name to Gemini
Internally, Google is integrating Gemini quickly, but it’s not an easy task. The company has approximately 180,000 employees worldwide and, in little more than a year, it's pushed all teams to make AI a core part of their work. Proof of its relevance is the prominent role played by DeepMind CEO Demis Hassabis during Google I/O. He's even rumored to be a possible successor to Pichai, even if he doesn’t seem like an obvious choice, given his more technical profile. In fact, according to Business Insider, managing dozens of teams and thousands of workers every day would actually be “Hassabis’ idea of hell.”
“We are at this stage where all the different products and all the different projects around Google are actually internalizing Gemini and rethinking how we do things... And I think that’s really energizing, because we are looking for the next step for the road ahead and the right way to work with users,” Pichai explained. “We’re talking about one of the most profound technologies humanity will ever work on,” he added. As in, ever. In Pichai's view, once AI goes fully autonomous, there will be no more tasks where humans are the better option.
“As a company, we’ve been investing in it for a long time. We are developing state-of-the-art models, and working to deploy it to billions of people in a way that they can make a difference in their lives,” he continued. “Personally, I love seeing innovation... It was a great week to be in technology.”
From Android to Glasses
In the Project Astra demo, a user can be seen interacting with a virtual assistant using voice and video through a pair of glasses. Pichai confirms Google’s plans to launch its own glasses with Astra, paving the way to creating a new ecosystem around them: “Project Astra is part of our technology in phones, but it really comes to life in a form factor like glasses. We’ve always had a vision for that... We are investing with our ecosystem in AR glasses, and over time, there will be interesting products to talk about.”
In the past, Google I/O has focused entirely on Android. But these are different times, and Andoid announcements were moved to the second day. Among the new features of Android 15 there is a plan to put an end to spam calls using AI, based on Gemini Nano.
“What an exciting moment we are having, with how operating systems are going to evolve... Each year, we’re going to be able to pack more of it on demand. What is Gemini Flash today will be Gemini Nano,” Pichai explained. “And I think the ability to bring so much intelligence on device really impacts a lot of capabilities. You see this with features like [the one Google just announced where] we can detect whether a call is a fraud call, privately on device, if you choose to opt-in. But we are scratching the surface of what’s possible... I think the possibilities are endless. I feel like in my head there is a decade of roadmap of all the stuff I want the teams to build. We’ll be busy at work.”
Google and the World’s Big Problems
This is an election year. More than 3 billion people will vote this year. Google is aware of the role AI can play in this context, but Pichai is confident that the risk can be mitigated.
“It’s one of the most important questions,” he said. “As Google, we have invested in elections integrity. That’s one of our highest priorities as a company, particularly in products like Search and YouTube. We deployed SynthID or AI-assisted Red Teaming to stay ahead of these problems. We undertake a lot of research through Project Jigsaw as well. So we first understand patterns in the world, and then we report on them. We share information that is appropriate with the right people, the right governments.”
“I think we’ve made a lot of progress,” Pichai stated. “I think we are in a moment where as a society we are able to more easily adjudicate what is real, what is not... I’m cautiously optimistic we’ll be able to do our part in handling all of this well. I think the stakes get higher in the future, but for this year, I am cautiously optimistic.”
As one of the world’s most powerful executives, Pichai has diplomatic responses for even the most controversial questions. Regarding Project Nimbus, which is the code name for the company's contract with the Israeli government, and the firing of nearly thirty employees for protesting it, Google’s CEO says those dismissals had nothing to do with voicing concerns. Pichai assured journalists in the session that Google gives its employees “more voice than pretty much any other company." The Google CEO justified the company’s firing of the employees by stating that it was establishing “a code of conduct” that is not disruptive to the workplace.
“We don’t want [our AI systems] deployed in the emergency systems for weapons and so forth. That’s always been our view, always. And so Project Nimbus is a company that’s been in a commercial arrangement with a government, as many other commercial cloud arrangements with many governments. In those arrangements there are terms of use, and they are always in place,” James Manyika, the vice president of research, technology, and society at Google, explained.
When it comes to governments, Pichai believes it's important to discuss the importance of AI regulation. “It makes sense to me that countries are thinking about an extraordinarily important topic. I think as a government, when you’re thinking about the impact AI will have on society, it seems right to me that they are debating these topics. It’s important to strike a balance. AI is going to drive a lot of economic opportunities. It’s going to affect industries across the board,” he said.
“I think, over time, we’re going to need more global frameworks... What makes the internet the force it is today is that it’s a global good. We all agree on common standards and a way of working on it together, and hopefully the same can apply to AI as well.”
Related: Google I/O 2024: Everything the Company Announced About Gemini, Android 15, and Its AI Advancements