Google DeepMind’s Demis Hassabis says AI isn’t a ‘silver bullet’ – but within five to ten years its benefits will be undeniable

The founder of Google DeepMind has hailed the arrival of agentic AI as a major step forward for the technology and struck an optimistic tone about its potential in years to come, while tempering expectations in the near term.
Demis Hassabis, CEO at Google DeepMind and one of the UK’s most prominent voices on AI, told assembled media at Google Cloud’s Gemini for the United Kingdom event that AI will bring exciting developments in the coming year.
As evidence for the potential AI holds, Hassabis pointed to developments such as Alphafold 3, which can accurately predict the structure of molecules such as DNA, RNA, and specific proteins and is the successor to the model that last year won him the Nobel Prize in Chemistry.
He also stated that Google Cloud’s climate, weather, and multimodal models will play an increasingly important role in providing scientists and organizations with the data they need
Seeking to clarify comments he made at the Paris AI Action Summit, Hassabis explained his outlook for AI in the coming years.
“I also think that just an earlier comment to clarify about AI being overhyped in the near term, what I was referring to there is really this idea that it’s a ‘silver bullet’ to everything in the next couple of years,” he said.
“I don’t see that happening just yet, I think we’re still quite a few years away from something like AGI happening.”
Artificial general intelligence (AGI) is generally understood as any AI system possessing intelligence equivalent to that of a human. Hassabis contrasted this with artificial super intelligence (ASI), an AI system that greatly exceeds human intelligence and may be able to improve itself, which he said is still theoretical.
At the same time, Hassabis said that the extent to which AI can bring about change in the next decade is underappreciated.
“I think today’s systems, they’re very impressive, but there’s still a lot of things they can’t do,” Hassabis said. “But I think over the next five to ten years, a lot of those capabilities will start coming to the fore and we’ll start moving towards what we call artificial general intelligence, so a system that’s able to exhibit all the cognitive capabilities that humans can.”
Focusing on Google DeepMind’s current operations, Hassabis pointed to the productivity improvements presented by AI agents and noted that it was working hard on creating ‘world models’ that can accurately simulate our world and be used to make detailed predictions about it.
Key to making good on this promise will be reducing the error rate present in current AI models, he explained.
“If your world model even has, say, a one percent error rate in it, if you plan over 50-100 steps that one percent compounds. It’s like compound interest and by the time you’ve done 50-100 steps you’re at potentially 50/50, it’s random where you’re going to be.”
AI, big business, and the UK
Hassabis was joined by Thomas Kurian, CEO at Google Cloud, alongside Mark Read, CEO at WPP and Allison Kirkby, CEO at BT Group, to make new commitments to UK AI and discuss the technology’s impact on the sector.
Concerning its current use cases, Read stated that WPP is already finding its partnership with Google Cloud to be highly beneficial in terms of improving workplace productivity.
“We’re using AI today to drive insights really quickly, come up with ideas for brainstorming creative sessions with clients,” he said. “Sometimes clients write very long briefs and we can shorten them extremely quickly into things that people understand.
“We can use AI to produce millions of variants of work, much more cheaply than we could before, or to create media plans and optimize media plans in real time. So there’s probably not any part of our business that won’t in time be touched by AI in one way or another.”
Read added that WPP has 41,000 workers using AI already, with training provided online and in-person. Regarding the potential to overhaul the UK’s public and private sector with AI, Kirkby said BT Group sees a unique opportunity to reinvent its operations to incorporate AI as it upgrades the country’s infrastructure.
“Where we’ve seen some of the biggest productivity aims are in our software developers,” Kirkby said, adding that some of its close work with Google Cloud includes blocking vulnerable customers from falling prey to AI-generated voice scams. BT Group produces as many AI patents per year as Google DeepMind.
Risks and customer worries
Hassabis was prompted to address the sustainability concerns of AI by The Economist, with Google having increased its emissions 48% compared to 2019 amid widespread AI adoption. Hassabis dismissed the idea that data centers would place an outsized strain on energy demand and by extension exacerbate emissions.
“I think that there’s quite a big misnomer here in my view, the amount of compute to train a model can be quite large – but you do that for once only and then it’s deployable relatively cheaply from an energy and cost perspective and very widely.
“I think the models that we’re going to build for example for things like climate modelling but also material design, all of these, I think this technology and AI is going to be one of the main drivers of solutions to fight the climate situation. So I think the amount of energy they use is going to be very small compared to the amount of savings they’re going to make.”
Hassabis backed his claims up with Google Cloud’s findings on using AI models to reduce its own consumption, which resulted in 30% savings on overall power use for data center cooling.
Kurian noted that Google is investing heavily in clean energy sources such as renewables and nuclear, and that its AI models are being used to support the research of new potential energy sources. Moreover, he insisted the natural cycle of model optimization will drive down inference costs as AI continues to develop.
“We’re very proud also here in the UK that we’re over 90 percent sustainable already in our data centers here in the UK, as we are in many other locations around the world,” he added.
MORE FROM ITPRO
Source link