Singularity starts now? OpenAI Cofounder builds AI system that gets better on its own

AI may soon become so smart that it can improve itself. While this might sound like science fiction or something that could happen a decade from now, an experimental project by an OpenAI co-founder has developed an AI system which is capable enough to autonomously run experiments to improve its own models.

At the start of this year, Tesla and xAI boss Elon Musk predicted that 2026 will be the “year of the Singularity.” Singularity is described as the moment when artificial intelligence becomes so advanced that it begins to surpass human intelligence and starts improving itself. Now, just three months into the year, it seems Musk’s prediction may already be showing early signs. A new experimental project from Andrej Karpathy, one of the early members of OpenAI and former Director of AI at Tesla, shows how AI systems could autonomously run experiments to improve other models.

Karpathy recently introduced a project called “autoresearch.” He built a system in which an AI agent repeatedly modifies and tests the training code of a language model in an automated loop. Instead of human researchers manually adjusting code, testing parametres and evaluating results, this AI agent itself proposes changes to the model, runs experiments, measures performance and keeps the versions that perform better.

In simple terms, Karpathy’s experiment shows how an AI system can try to improve itself without constant human intervention. Sharing results from the project online, Andrej Karpathy revealed how the AI agent ran hundreds of experiments autonomously and gradually improved the model step by step.

“Who knew early singularity could be this fun? :) I just confirmed that the improvements autoresearch found over the last two days of (~650) experiments on a depth-12 model transfer well to depth-24, so nanochat is about to get a new leaderboard entry for ‘time to GPT-2’ too. Works,” he wrote in one of his posts on X.

Karpathy’s project quickly became a topic of discussion across the AI community, highlighting how the project could show early indicators of what technologists call the technological singularity. “The singularity has begun. so many signs,” wrote Tobi Lutke in response to the development. Musk also agreed with Lutke's comments.

The term singularity refers to a point where artificial intelligence is expected to surpass human intelligence and begins improving itself rapidly without human intervention. Once machines become capable of designing better versions of themselves, often linked with the idea of Artificial General Intelligence, researchers suggest it could accelerate progress rapidly, potentially leading to breakthroughs far beyond its current capabilities.