the path of technological advancement is rarely smooth, and artificial intelligence is no different.While researchers and tech giants push towards artificial general intelligence (AGI), the hypothetical AI that matches human intellect, they face significant hurdles. Current AI models frequently enough struggle with hallucinations, inaccuracies, and basic mistakes.

Google Ceo sundar Pichai has dubbed this stage “artificial jagged intelligence,” or AJI,during a recent appearance on Lex Fridman’s podcast.

“I don’t know who used it first, maybe Karpathy did,” Pichai said, referencing Andrej Karpathy, a specialist in deep learning and computer vision who previously cofounded OpenAI.

AJI serves as a fitting metaphor for AI’s current state: a trajectory marked by both remarkable breakthroughs and fundamental errors.

In a 2024 X post titled “Jagged Intelligence,” Karpathy defined the term as “a word I came up with to describe the (strange, unintuitive) fact that state of the art LLMs can both perform extremely impressive tasks (e.g. solve complex math problems) while together struggle with some very dumb problems.” He illustrated this with examples of advanced large language models failing to recognize that 9.9 is larger than 9.11,making “non-sensical decisions” in tic-tac-toe,and having difficulty with counting.

Unlike humans, “where a lot of knowledge and problem-solving capabilities are all highly correlated and improve linearly all together, from birth to adulthood,” AI’s inconsistencies are often unpredictable, according to Karpathy.

Pichai concurred, noting, “You see what they can do and then you can trivially find they make numerical errors or counting R’s in strawberry or something, which seems to trip up most models. I feel like we are in the AJI phase where dramatic progress, some things don’t work well, but you’re seeing lots of progress.”

The Future of AGI

“I feel like we are in the AJI phase where dramatic progress, some things don’t work well, but you’re seeing lots of progress.”

According to Pichai, when Google DeepMind was established in 2010, the team estimated AGI would be achieved in 20 years. Google acquired DeepMind in 2014. Now,Pichai anticipates it will take longer,but he believes that “by 2030,I would stress it doesn’t matter what that definition is as you will have mind-blowing progress on many dimensions.”

He also emphasized the need for a clear labeling system for AI-generated content to “distinguish reality.”

Pichai has previously discussed the potential benefits of AI development, outlining four