OpenAI scientist claims AI nearing level of human research intern

OpenAI is making steady progress toward developing artificial intelligence (AI) systems capable of performing at the level of a human research intern, according to its chief scientist Jakub Pachocki.

Speaking on a recent podcast, Pachocki said advances in coding, mathematics and physics indicate that AI systems are becoming increasingly capable of handling complex, multi-step technical tasks with reduced human supervision.

He noted that a key measure of progress is how long an AI model can work autonomously on a task.

“The distinction between a research intern and a fully autonomous researcher is largely about how long the system can operate independently,” he explained.

OpenAI has previously outlined internal goals to build an “AI research intern” by 2026, with ambitions to develop a fully autonomous AI researcher by 2028.

CEO Sam Altman has acknowledged the targets are uncertain but emphasised the importance of transparency.

Pachocki highlighted rapid improvements in coding tools, including systems that now handle a significant share of programming work.

He also pointed to mathematical benchmarks as a reliable way to measure reasoning capabilities.

Despite the progress, he cautioned that AI is not yet capable of operating independently at a full researcher level.

For now, the focus remains on extending the duration and complexity of tasks AI systems can manage with minimal oversight.