OpenAI: Jakub Pachocki Bets on AI for the Future but Remains Cautious

The Evolution of Jakub Pachocki's Work with AI

Jakub Pachocki, the Chief Scientist at OpenAI, has long favored a traditional approach to coding, writing every line of code manually. However, recent advancements in artificial intelligence have changed the way he works. Now, experiments that previously required a week of work can be completed by AI in just one weekend. Pachocki acknowledges the efficiency of these tools, although he remains cautious about their ability to handle complex design tasks.

A year ago, Pachocki didn't even use autocompletion, preferring to code in Vim, a minimalist text editor. His meticulous approach to coding, which he described to MIT Technology Review as "very particular," made him skeptical of AI tools. However, the performance of the latest models has convinced him of their utility for certain tasks. "It's hard to argue against that," he admits, while emphasizing that AI cannot yet replace human expertise in designing complex systems.

Towards Greater Autonomy in Research

OpenAI, under Pachocki's leadership, aims to develop an autonomous AI research intern. Announced last fall, this system is designed to handle tasks that would typically take a human several days. Its launch is scheduled for September. By March 2028, OpenAI hopes to introduce a complete "AI Researcher," a multi-agent system capable of solving complex problems across various scientific fields such as mathematics, physics, biology, chemistry, economics, and politics.

Pachocki envisions a future where these systems could operate as autonomous research laboratories, reducing the need for large human teams. "Of course, you still want people in charge and setting the goals," he clarifies, but he anticipates a radical shift where data centers could manage the work of large organizations like OpenAI or Google. This concentration of technological power could be unprecedented, according to him.