Google’s Chief Scientist Jeff Dean Says He Does Not Like to Talk About AGI

Google DeepMind's chief scientist, Jeff Dean, has revealed a specific topic related to artificial intelligence (AI) that he does not like to discuss. In a recent podcast episode, the researcher, who is working on AI technology under the company's DeepMind division, spoke about a range of topics that include machine learning, Google Brain, AI, Gemini, and the future of the technology; however, he added that he does not prefer to talk about artificial general intelligence (AGI). Notably, Dean believes the subjective nature of its definition acted as a deterrent for him.

Google's Chief Scientist Does Not Like to Talk About AGI

In a recent episode of The Moonshot Podcast, Dean discussed several AI-related topics with host Eric Teller (Astro Teller). During the nearly hour-long conversation, the Google DeepMind researcher shed light on neural networks, safety, and security in AI, and the breakthroughs made by the company that he contributed towards. However, when Teller brought up the topic of AGI, the scientist immediately declined to talk about the topic.

Highlighting his aversion towards the AI-related topic, he said, “The reason I tend to steer away from AGI conversations is lots of people have very different definitions of it, and the difficulty of the problem varies by factors of a trillion.”

Google CEO Sundar Pichai, recently in an interview, admitted that achieving AGI with the current technology might be an impossible task. Making his point, he reportedly highlighted that while AI has made significant progress towards self-driving technology, it still struggles in learning to drive, which takes a human about 20 hours on average.

Dean, however, differs from Pichai in the last point. He said that AI models today are probably already better than the average person at most “non-physical tasks.” Justifying his statement, the scientist said that most people are not good at a random task they have not done before, but an AI model can perform it reasonably well.

However, he did admit that at the same time, AI “will fail a lot at a lot of things, and they are not human expert level in some things.” Dean added that this conflicting nature of AI is the reason he does not like to talk about AGI.

While there is no consensus, AGI is commonly understood as a type of AI system that possesses human-like cognitive abilities such as reasoning, problem-solving, planning, using strategy, and making judgments. It is said to be able to understand, learn, and perform any intellectual task that a human being can.