Wednesday, March 11, 2026

Google chief scientist Jeff Dean: AI needs “algorithmic breakthroughs” and AI isn’t answerable for a lot of the increase in emissions in data centers

Google chief scientist Jeff Dean: AI needs “algorithmic breakthroughs” and AI isn’t answerable for a lot of the increase in emissions in data centers

Google sent a jolt of Discomfort entered the climate change debate when the corporate announced that emissions from its data centers will increase by 13% in 2023, citing the “AI transition.” in its annual environmental reportBut in response to Jeff Dean, chief scientist at Google, the report doesn’t tell the entire story and puts an excessive amount of of the blame on AI.

Dean, who’s chief scientist at Google DeepMind and Google Research, said Google won’t back down from its commitment to run on one hundred pc clean energy by the tip of 2030. However, that progress is “not necessarily linear” because a few of Google’s collaborations with clean energy providers won’t come online for several years.

“These things will significantly increase the proportion of carbon-free energy in our energy, but we also want to focus on making our systems as efficient as possible,” Dean said Tuesday in an onstage interview with Fortune’s AI editor Jeremy Kahn at Fortune’s Brainstorm Tech conference.

Dean went on to say that AI isn’t as answerable for the increasing use of information centers and thus for CO2 emissions as critics claim.

“The increasing energy consumption of AI is the focus of attention, and that consumption is definitely increasing from a very small level,” Dean said. “But I think people often confuse that with overall data center usage – of which AI is currently a very small portion, but which is growing rapidly – and then attribute the growth rate of AI-based computing to overall data center usage.”

Dean said it was vital to take a look at “all the data” and the “real trends underlying it,” but didn’t elaborate on what those trends were.

Dean is one in all Google’s first employees. He joined the corporate in 1999 and is is taken into account one in all the important thing figures who transformed the early web search engine into a strong system that would index the online and reliably serve billions of users. Dean co-founded the Google Brain project in 2011 and led the corporate’s efforts to change into a frontrunner in AI. Last 12 months, Alphabet merged Google Brain with DeepMind, the AI ​​company Google acquired in 2014, and made Dean chief scientist reporting on to CEO Sundar Pichai.

By combining the 2 teams, Dean said, the corporate has “a better set of ideas to build on” and might “pool computing power so we can focus on training a large-scale project like Gemini, rather than multiple fragmented projects.”

Algorithmic breakthroughs needed

Dean also answered an issue in regards to the status of Google’s Project Astra – a research project that DeepMind lead Demis Hassabis presented in May at Google I/O, the corporate’s annual developer conference. Hassabis described Astra as a “universal AI agent” that may understand the context of a user’s environment. A video demonstration of Astra showed how users can point their phone’s camera at nearby objects and ask the AI ​​agent relevant questions, resembling “What area am I in?” or “Did you see where I left my glasses?”

At the time, the corporate said Astra technology could be integrated into the Gemini app later this 12 months. Dean, nevertheless, put it more conservatively: “We hope to have something available to test users by the end of the year,” he said.

“The ability to combine Gemini models with models that actually have agency and can perceive the world around you multimodally is going to be very powerful,” Dean said. “We’re obviously using this responsibly, so we want to make sure the technology is ready and doesn’t have any unforeseen consequences, which is why we’re going to introduce it to a smaller group of test users first.”

As for further development of AI models, Dean noted that additional data and computing power alone won’t be enough. A couple of more generations of scaling will get us significantly further, Dean said, but ultimately it’ll take “some additional algorithmic breakthroughs.”

Dean said his team has long focused on ways to mix scaling with algorithmic approaches to enhance objectivity and reasoning skills in order that “the model can imagine plausible outcomes and reason through which one makes the most sense.”

Advances like this are vital, Dean said, “to really make these models more robust and reliable than they already are.”

Read more coverage from Brainstorm Tech 2024:

Wiz CEO says ‘consolidation in security market is really a necessity’ as reports of $23 billion Google acquisition emerge

Why Grindr’s CEO believes ‘synthetic employees’ will spark a brutal battle for one of the best talent at tech startups

Experts fear that the Cold War between the US and China could turn right into a hot war: “Everyone is waiting for the hammer to fall in Asia”

Recommended newsletter:

CEO Daily provides crucial context to the news that leaders across the business world have to know. Every weekday morning, greater than 125,000 readers trust CEO Daily for insights into the C-suite and its surroundings. Subscribe now.

Latest news
Related news