Is Google truly invested in succeeding in AI?

Today on Decoder, I’m chatting with Hayden Field, a senior AI reporter, about a hot topic that’s been buzzing in the tech world: Is Google falling behind in the AI race?

Recently, Google made headlines with a major shakeup in its AI division, Google DeepMind. Jeff Dean, the chief scientist, is leaving to start his own venture, while DeepMind co-founder and CEO Demis Hassabis is stepping back to focus on long-term research.

These changes can be interpreted in various ways, and I wanted to sit down with Hayden to explore some of the sharpest insights we’ve seen this week and what they might mean for the future of AI at Google.

At its core, this is an organizational story — and Decoder loves diving into those.

Let’s get into it: Hayden Field on the latest developments at Google DeepMind and the future of AI research.

This interview has been lightly edited for length and clarity.

Hayden Field, welcome back to Decoder.

Thanks! It’s great to be here.

It’s always a whirlwind when you’re around, Hayden.

Absolutely. There’s never a dull moment. Each week brings something wild.

This time feels particularly significant. Last week, Google reorganized its AI division, Google DeepMind. Demis Hassabis has moved up to become the chairman of DeepMind, focusing on broader research, while Jeff Dean, who helped launch Google Brain, is leaving with several others to start a new lab on Google Cloud.

This is a substantial reshuffle of Google’s AI strategy, especially as they seem to be lagging behind in the competitive landscape. I wanted to discuss the reactions to this news with you, given your extensive reporting on Google and the industry, and help put it all into perspective.

Let’s start from the beginning. Google has long been seen as the frontrunner in AI. With its resources, distribution power, and ability to integrate AI into products like Search, it seems poised to dominate.

As you noted during the Elon Musk-Sam Altman trial, many were wary of Demis Hassabis. Given all these advantages, it’s surprising to see many now suggesting that “Google’s going to lose.” What’s your take on this situation? Is it a disaster for their cutting-edge efforts?

It’s definitely concerning, but I think people are overlooking just how powerful Google remains and the vast resources it controls. They have a significant safety net, a cushion that allows them to recover from setbacks. Even if they stumble and fall behind, it’s premature to count them out entirely because of their immense capabilities. However, it is astonishing that with all that power, they’re not leading the pack right now.

Both Sundar and Demis acknowledged this in their statements last week. Sundar emphasized the company’s commitment to staying at the forefront and improving in key areas, which is CEO-speak for, “We know we need to get our act together.” Demis mentioned that the company is “entering a next chapter” and has the potential to lead, which translates to, “We’re not there yet, but we have what it takes to get there eventually.” It’s quite telling that they both recognize this.

It’s not looking great for them, but it reminds me of someone with a trust fund. They have a fallback. Don’t count them out just yet.

That trust fund is Google Search. They have a product that billions use, and they’re integrating AI into it. As we’ve discussed before, consumer AI hasn’t yet proven to be a major revenue generator, but it’s seen as the ultimate goal. Right now, the action is in enterprise AI, and companies like Anthropic are gaining traction due to their focus on that area, particularly in coding.

I want to circle back to whether Google made a misstep by prioritizing multimodality and world models, which Demis is passionate about, over core enterprise applications. They claim they want to lead, but perhaps they’re more interested in selling cloud services to companies like Anthropic, which is a growing business for them.

Let’s hold that thought for now. You mentioned Sundar’s commitment to the next chapter and the changes being made. A few months ago, I spoke with Sundar about the company’s restructuring and the creation of DeepMind. He acknowledged the challenges of pivoting the company to be AI-first. Here’s a clip:

> Sundar Pichai: It was tough to convey it outside, but I pivoted the company to be AI-first. We had all the ingredients, so in some ways I felt like the Overton window had changed. People were adopting these technologies faster than we had expected. To me it was a way to go and actually express ourselves through our products, but I realized we had to organize ourselves for it.

Sundar’s comments highlight that the company wasn’t prepared for the AI moment and needed to reorganize. Now, just a few months later, the head of Google Brain is leaving, and the head of Google DeepMind is transitioning to a new role.

Now, Google DeepMind doesn’t have a CEO anymore; it has an SVP. What do you make of that? Is this another reset, or is it just more of the same?

It’s definitely another reset, but it’s a complex situation. Demis has always been more focused on research than productization, and Jeff Dean shares that mindset. So, I’m not surprised by these changes.

If you’re feeling optimistic, you could see this as a ripple effect from your conversation with Sundar. However, I don’t think this bodes well for the company. As we’ve seen with OpenAI, executive shakeups often lead to instability. People get anxious, and productivity can suffer. Many of those working on AI at Google were drawn there because of Jeff Dean or Demis. What happens now?

They’re likely thinking about logistics and who can best accelerate productization. That might be the right call, but considering the talent leaving and the potential for more departures in the coming months, it could be a rough road ahead.

There’s a lot of buzz around Gemini 4, which hasn’t been released yet. They announced it at Google I/O, but their focus has been on the 3.5 class models, which aren’t leading the pack anymore.

Do you think the executives believe Gemini 4 will propel them back to the forefront, or are they leaving because Google isn’t as committed to being a leader?

As noted by SemiAnalysis, “These are not the actions of people excited about Gemini 4 Pro.” Google is still committed to being a frontier lab, but the question is whether they’ll achieve that goal. A few months ago, I would have said they were on track, but now it feels more uncertain. SemiAnalysis has suggested that DeepMind is no longer a frontier lab. While I still see it as one, that could change quickly, and I didn’t expect it to shift so fast. There’s a risk they could follow in Meta’s footsteps, falling behind while making promises without delivering.

I appreciate you bringing up SemiAnalysis. They do solid work analyzing AI and tech. Their insights resonate with our ongoing coverage of Google, but some aspects feel like new challenges for the company.

To quote them, “For all intents and purposes, we believe DeepMind is no longer a frontier lab. Google will continue meandering on and releasing models, but their odds of reaching the state of the art again have dropped to zero.”

That’s a bold statement. The AI landscape is constantly evolving, and to claim that Google will never again reach the cutting edge seems extreme. Do you think there’s merit to that assertion?

That’s a bit premature. It’s been a few months since we reported on Gemini 3’s initial hype, which had them leading the model race until something else came along. I don’t believe their chances of reaching the state of the art have dropped to zero. They’re shifting focus towards productization, which could help them regain their edge.

However, this approach may lead to them constantly playing catch-up with other frontier labs that invest more in long-term research. It’s akin to a writer who’s always a step behind someone who’s innovating. That’s the trajectory I see for them. If they don’t invest in long-term research and niche areas, they might reach the state of the art again, but they won’t consistently lead.

That’s a new perspective for me. Evaluating that claim is tricky, especially with the leadership changes in Google’s AI division. I can’t read Sundar’s mind, and it’s challenging to gauge whether he’s committed to winning the race or if he believes Gemini is sufficient for generating revenue within Google’s products.

What I can assess is the second part of the SemiAnalysis claim, which resonates with anyone familiar with Google. They state, “The issue with Google was not Jeff Dean or Noam Shazeer, but rather their extremely bureaucratic, painfully slow, and strategically timid culture.”

That’s quintessential Google. I often ask Sundar about the company’s culture and bureaucracy. His statement about making DeepMind the center of the company, followed by the departures and SemiAnalysis’s assertion that they’ll never be at the frontier again due to their bureaucratic culture, feels very familiar.

Have you heard from people at Google about whether the culture needs to change further to compete?

There’s a reason many are flocking to Anthropic right now. While Anthropic isn’t without its issues, I’ve seen talent from OpenAI, Google, and Meta gravitating there. Many in this field are beyond just monetary incentives; they want to work for companies that align with their values. Google has alienated many in that regard.

Both Jeff Dean and Demis signed a public letter in 2018 opposing the use of AI for lethal autonomous weapons. Jeff Dean has expressed concerns about mass surveillance in AI. He even supported Anthropic in a legal brief regarding the Department of Defense, while Google has been more permissive with military applications of AI.

I foresee a potential exodus in the coming months, especially since Jeff Dean was seen as a key figure keeping many at Google.

This raises a significant question: Who will uphold the morals and values of the team in a Google that seems more willing to collaborate with the Trump administration and the military on AI applications that some find troubling?

There’s been extensive reporting on how Google’s acquisition of DeepMind came with a pledge to never use its technology for military purposes, which has since been diluted. Jeff Dean recently reiterated that AI shouldn’t be used for such purposes. There are employees at Google expressing their desire not to commercialize AI in the current manner.

The challenge is that governments and militaries are the primary funders right now, and Google excels at generating revenue. They want to finance AI development through actual profits rather than relying on debt like many startups. Can Sundar navigate this conflict, or is it simply the reality of being a massive corporation focused on revenue?

It’s not surprising for a large U.S. corporation to prioritize shareholder value. It seems they’re subsidizing their AI efforts. For instance, I recently used Gemini for hours without signing in, as it was the only tool available for free. They’re trying to attract developers and engineers to their platform, which requires funding.

While they could approach this differently, given their size and revenue from other areas, it doesn’t mean they won’t continue down this path. However, I believe this could lead to significant talent loss. One current Google employee mentioned that losing Jeff Dean was a major blow to the company’s moral compass, and it could result in mediocrity. As you pointed out, they may find themselves perpetually catching up rather than leading.

This brings me back to the framework I mentioned earlier. Does Google genuinely want to win at the frontier, or are they content to sell Google Cloud services at high rates to competitors like Anthropic and OpenAI while maintaining functional models for their consumer products?

There’s a scenario where Gemini serves to enhance Gmail and provide AI insights, but the real profit lies in selling cloud services to Anthropic, which is securing government contracts. Google seems fine with this arrangement, which poses challenges when your competitors are also your customers.

How do you see this unfolding? Will Google prioritize winning at the frontier, or will they settle for being a step behind while profiting from their competitors’ ambitions?

Currently, they’re likely to keep pursuing the frontier because their investors expect it. It boosts their stock to claim they’re competitive with OpenAI and Anthropic. They don’t need to follow the path of Amazon or, in my view, Meta.

In a year or two, if the brain drain continues and they fall further behind, they might shift strategies. But for now, there’s no reason for them to stop chasing, especially since it’s a compelling narrative for earnings calls and keeps investors satisfied.

The AGI aspect is particularly intriguing in the context of Google. Demis Hassabis was the final speaker at Google I/O, discussing protein folding and drug discovery, which are his passions. He concluded by confidently stating, “We are at the foothills of the singularity.” Google avoids the term “AGI,” opting for “the foothills of the singularity,” which is a bold assertion.

I thought that was a striking way to wrap up Google I/O. Later, I asked Sundar if he agreed with that sentiment. Here’s what he said:

> Sundar Pichai: Demis and I have had long, deep conversations on this topic.

> Nilay Patel: I figured.

> In this context, for him, the advent of AGI is what he thinks of as the singularity and I think—

> Do you have a definition of AGI? Have you debated? Do you have an agreement?

> We debated a lot. Both Demis and I are very close to how we think about AGI … There is a harder definition of AGI, which is that it has to more comprehensively do the wide range of tasks, including cognitive tasks, in a way that’s comparable. We’ll at some point actually put it out as a company, and we are working on that. But that’s what he’s talking about in this context.

I’ve been pondering Sundar’s response since then. Do you have insight into his definition of AGI or the singularity based on that?

It aligns with the broader industry perspective. In the OpenAI lawsuit, we saw their definition of AGI, which describes it as “a highly autonomous system that outperforms humans at most economically valuable work.” This seems to echo Sundar’s thoughts on cognitive tasks.

That’s the definition I’ve been using for years, and it appears to match what he’s saying. He mentioned that Google would eventually release a definition of AGI or the singularity, but as far as I know, they haven’t done so yet. I suspect it aligns with the industry consensus that it will surpass human capabilities in economically viable tasks.

They’ve stopped emphasizing that aspect, likely because it implies job displacement for many.

I bring this up because Demis is deeply invested in scientific pursuits, particularly in health outcomes. With his transition to chief scientist, Sundar reiterated, “We’re at the foothills of the singularity.” It seems Demis is eager to pursue AGI, world models, and protein folding, while Google’s immediate focus is on monetizing AI through software development and automating business processes for large corporations. These priorities seem increasingly divergent.

This highlights a cultural clash within Google: the tension between pursuing profit and adhering to ideals. The AI industry is rife with uncertainty, especially regarding how companies will monetize their innovations.

How do you perceive this? Did Demis lead them astray in pursuit of accolades, or is there validity in allowing him to chase the singularity while the company focuses on practical applications?

There’s a distinction between letting him pursue the singularity and allowing him to conduct research in biology and drug discovery, which seems to be his new focus. It appears he and Jeff Dean will be exploring similar areas but in different contexts.

This is a long-standing narrative in tech: the clash between product development and research. Companies often prioritize immediate results over long-term investments that could yield future products. They want to know what can be delivered quickly, which creates tension, especially as competitors like OpenAI and Anthropic prepare for public offerings. Google has many investors to satisfy, and this tension has been building for years.

I suspect the situation with Demis was mutual. There are rumors he may have wanted to leave entirely, but it would have been too dramatic for both him and Jeff Dean to exit simultaneously. So, he opted for a different role, but that’s speculative. Regardless, it seems he wasn’t entirely pushed out; rather, he’s more interested in long-term research, which aligns with the mindset of many in the field.

It’s a classic case of being careful what you wish for. When Sundar established Google DeepMind, he had to choose a leader, and he selected Demis over Jeff Dean. Now, both are leaving, highlighting the challenges of operating within Google’s structure compared to pure research.

One question I have is about DeepMind itself, which has historically operated independently from Google. With the leadership changes, it’s now being led by an SVP who reports directly to Sundar. There are indications that authority may shift back to Mountain View. How do you see this evolving within Google?

I can see the power dynamics shifting back to the Bay Area. DeepMind has enjoyed a degree of autonomy for a long time, and I’ve visited their office multiple times. It has a collaborative atmosphere, focused on long-term research, which contrasts with the urgency I feel in other AI labs.

Perhaps Sundar has decided that he needs to increase pressure for quicker product development. I can envision the center of gravity moving back to the Bay Area.

I’m particularly interested in how the culture at Google will change without these leaders. Jeff Dean and Demis were not only technical innovators but also moral leaders who influenced the company’s direction. With their departures, there are concerns about what values will guide the company moving forward.

Koray Kavukcuoglu is taking over DeepMind, and I doubt he’ll be as vocal on moral issues. He’s more focused on product development and speeding up processes.

In recent months, I’ve spoken with AI employees across various companies, including Google, and many feel the messaging is to “keep your head down and focus on the mission.” It seems this approach will intensify now that Jeff Dean and Demis are gone.

One source mentioned that Jeff Dean was one of the last voices willing to speak out on moral issues. With Koray stepping in, it’s likely that the culture will shift further away from addressing these concerns.

To wrap up, there’s a lot of chatter suggesting that Google has lost the AI race. However, as we’ve discussed, Google has significant advantages.

It’s generating revenue, which many competitors aren’t. It’s selling Google Cloud infrastructure to its rivals and has substantial distribution channels through Search and Gmail. Google has structural advantages and is essentially playing with house money.

Is it feasible for them to lose the AI race given this context? Or is it more about whether Anthropic achieves recursive self-improvement and claims AGI, which would force everyone else to lose by default?

They can’t truly fail, but they might alter their strategy. As we discussed, they could become more like Apple, relying on Google AI while focusing on different competitors. Similarly, Amazon seems to have stepped back.

I don’t believe Google can fail outright, but they might eventually realize it’s not worth their time. If the brain drain continues and they find it impossible to attract top talent back, they may decide to cut their losses. However, that’s likely a ways off.

It’s crucial for executives to recognize that many key players in developing this technology are deeply committed to their values, not just motivated by money. Reselling that vision will be challenging if they lose talent. OpenAI is facing similar issues, so it will be interesting to see how this unfolds.

It’s worth noting that Demis is still with the company as chief scientist, and he’ll likely continue to engage with the London office. However, the real questions are about what they’ll deliver, the roadmaps, and who gets access to resources. Those dynamics are likely to change, with significant downstream effects.

If you had to look for indicators of Google’s success or failure, what would you focus on? Is the culture a problem, or are they just retrenching to become more aggressive?

I typically assess their major releases and how impactful they are. Are they introducing something that their competitors launched six months ago? Are they executing well? That’s how I gauge their progress.

With Gemini 4, I’ll be watching to see if it can create the same buzz that Gemini 3 did, or if it falls flat and only serves enterprise clients without generating excitement.

I’ll also be monitoring talent retention. If a significant number of DeepMind engineers leave or if they struggle to hire and offer increasingly desperate compensation packages, that’s a red flag. Whenever Demis eventually departs, that will be a major signal.

We’ll keep an eye on all of this. I’m confident the chaos in the AI industry will persist, and I look forward to having you back on Decoder, Hayden. Thanks for joining me today.

Thanks for having me!