OpenAI Sets 'North Star' Goal: Building Fully Autonomous AI Researcher by 2028
Key Takeaways
- ▸OpenAI is making autonomous AI research its primary focus ('North Star') for the next several years, consolidating work on reasoning, agents, and interpretability
- ▸The company plans a September 2024 launch of an 'autonomous AI research intern' for specific tasks, leading to a full multi-agent research system by 2028
- ▸Codex, released in January, is positioned as an early prototype of the AI researcher, with widespread adoption among OpenAI's technical staff
Summary
OpenAI is refocusing its research efforts around an ambitious new goal: developing a fully autonomous AI researcher capable of independently tackling complex scientific and mathematical problems. The company plans to launch an "autonomous AI research intern" by September 2024 that can handle specific research tasks, with a more advanced multi-agent research system targeted for 2028. This initiative brings together multiple research strands including reasoning models, AI agents, and interpretability work, and represents OpenAI's strategic response to intensifying competition from rivals like Anthropic and Google DeepMind.
OpenAI Chief Scientist Jakub Pachocki, who played key roles in developing GPT-4 and reasoning models, outlined the vision in an exclusive interview, stating that the company is approaching a point where AI models can work "indefinitely in a coherent way" similar to humans. The foundation for this ambition is already being laid through Codex, OpenAI's agent-based application released in January that can autonomously execute code-based tasks. Pachocki emphasized that the goal is to create systems capable of handling research tasks that would normally require days of human effort, potentially solving problems in mathematics, physics, life sciences, and beyond.
- Chief Scientist Jakub Pachocki believes the company now has most technical components needed to create systems that can run extended research tasks with minimal human guidance
Editorial Opinion
OpenAI's commitment to building autonomous AI researchers represents a logical evolution of recent advances in reasoning and coding agents, yet the ambitions outlined here—solving complex math, physics, and life science problems autonomously—remain highly speculative and depend on breakthroughs that may not arrive on schedule. The 2028 timeline suggests confidence in the trajectory of AI capabilities, but the historical gap between AI industry promises and delivery warrants measured skepticism. This focus could cement OpenAI's competitive position if successful, but it also underscores the arms-race dynamics now driving the field, with each major lab pursuing increasingly powerful autonomous systems.
