Is AGI Possible? Godfather of AI and Google DeepMind Chief Caught in War of Words on Social Media

Highlights

• Hassabis said LeCun is “plain incorrect”
• He also equated human brains with Turing machines
• LeCun believes the human brain is highly inefficient

The AI researcher beef was on nobody's 2025 Bingo card, but it has happened (before GTA 6). X (formerly Twitter) was called the “digital town square” by Elon Musk, and it is an acceptable social media platform to argue with those whose views you do not agree with, and generally, no one bats an eye. But when the Godfather of AI and the 2024 Nobel Prize in Chemistry winner tangle in a war of words, it does turn heads. On Monday, Yann LeCun and Demis Hassabis were engaged in a heated conversation over whether general intelligence exists.

Demis Hassabis and Yann LeCun Battle Over General Intelligence

While the main argument was about the existence of general intelligence as a concept, there is a deeper link with the technology in which both are heavily invested. Hassabis is the CEO of Google DeepMind, the division which leads Google's major AI projects, from research to deployment. On the other hand, LeCun served as Meta's Chief AI Scientist for years and has recently launched his AI startup, Advanced Machine Intelligence (AMI) Labs.

So, when they argue over the concept of general intelligence, what they are really saying is whether building artificial general intelligence (AGI) is a feasible goal or not. Notably, every major AI company, including Anthropic, Google, Meta, Microsoft, OpenAI, xAI, and more.

The first post was made by Hassabis when he replied to an interview with LeCun, where the Turing Award winner declared that “general intelligence as a concept” does not make sense. To put it simply, he believes that human minds are super-specialised to complete tasks in our physical world. Calling the word “general” a misnomer, he claims that human perception believes in general intelligence because we cannot even imagine the problems that our brains cannot solve. He also illustrates this with the example of Chess, a sport where machines are far superior to humans.

“Yann is just plain incorrect here; he's confusing general intelligence with universal intelligence,” replied Hassabis, adding that the human brain is extremely general. He argued that the super-specialised nature of the brain is an acquired trait due to its finite memory and energy.

“But the point about generality is that in theory, in the Turing Machine sense, the architecture of ​such a general system is capable of learning anything computable given enough time and memory​ (and data), and the human brain (and AI foundation models) are approximate Turing Machines,” the Google DeepMind CEO said. He also refuted the chess argument, adding that humans inventing chess in the first place is evidence of the general capacity.

With nearly 10,000 likes and 1,200 reshares (at the time of publishing), the post captured the attention of LeCun. He replied, “I object to the use of 'general' to designate 'human level' because humans are extremely specialised.” Giving the example of the optic nerve, he argued that it is capable of a massively large number of vision functions and sees a large number of rays, but in reality, the eye can only see a fraction of all rays that exist in the world (the visible light or VIBGYOR).

Reiterating his earlier position, he added, “Clearly, a properly trained human brain with an infinite supply of pens and paper is Turing complete. But for the vast majority of computational problems, it's horribly inefficient, which makes it highly suboptimal under bounded resources (like playing a chess game).” Hassabis has yet to respond to this argument.

So, how does it connect to AGI? The rationale is that if humans themselves do not possess general intelligence, how can we create a machine that is capable of it? As such, AGI is a meaningless goalpost, and the real goal has always been superintelligence. However, many experts in the field have disagreed with this notion and have called AGI the midpoint to superintelligence or sentient AI, with human-level intelligence and the ability to perform general-purpose tasks.