OpenAI’s Ambitious Plan: From AI Research Intern to Fully Autonomous Scientist by 2028

The Dawn of AI-Driven Research

In a bold move that could redefine scientific discovery, OpenAI has announced plans to develop an “autonomous AI research intern” by September, with ambitions to scale it into a fully automated, multi-agent research system by 2028. The project, spearheaded by Chief Scientist Jakub Pachocki, aims to tackle narrowly defined research problems initially before expanding into broader scientific inquiry. If successful, the initiative could accelerate breakthroughs in fields ranging from medicine to quantum computing—while also raising profound questions about the future of human researchers.

The announcement comes as AI continues to disrupt traditional research methodologies. OpenAI’s vision is not merely to assist scientists but to create independent AI agents capable of formulating hypotheses, designing experiments, and interpreting results without human intervention. Pachocki, in an exclusive interview, described the project as a stepping stone toward a future where AI systems could autonomously push the boundaries of human knowledge.

The Road to Autonomous AI Researchers

The first phase of OpenAI’s plan involves training an AI model to function as a research assistant, focusing on well-defined problems where data is abundant and success metrics are clear. By 2028, the company hopes to evolve this into a fully autonomous system capable of handling complex, multi-disciplinary research challenges.

Experts remain divided on the feasibility of such an ambitious timeline. While AI has already demonstrated prowess in narrow tasks—such as predicting protein structures (as seen with DeepMind’s AlphaFold)—creating a system that can independently navigate the ambiguity of open-ended scientific inquiry is a vastly more difficult challenge. Skeptics argue that true scientific intuition, creativity, and serendipity may remain beyond AI’s reach for years to come.

Yet, if OpenAI succeeds, the implications could be staggering. An AI capable of generating novel insights could drastically shorten research timelines, leading to faster medical cures, more efficient energy solutions, and breakthroughs in fundamental science. However, it also raises ethical concerns: Who owns AI-generated discoveries? How do we ensure transparency in AI-driven research? And what happens to human scientists in an era of autonomous discovery?

The Psychedelic Paradox: Promising Drugs, Problematic Trials

Meanwhile, in the world of biomedical research, psychedelic compounds like psilocybin—the active ingredient in “magic mushrooms”—continue to generate both excitement and skepticism. Once relegated to counterculture, these substances are now the subject of serious clinical investigations for treating depression, PTSD, addiction, and even obesity.

However, two recent studies highlight the immense challenges of studying psychedelics in controlled settings. The drugs’ powerful subjective effects make double-blind trials difficult—if patients know they’ve received the real compound rather than a placebo, the results can be skewed by expectation bias. Additionally, the long-term risks and mechanisms of these substances remain poorly understood.

Jessica Hamzelou, reporting for MIT Technology Review, argues that while psychedelics hold genuine therapeutic potential, the hype surrounding them has often outpaced the science. Early, small-scale studies have produced dramatic headlines, but larger, more rigorous trials have yet to consistently replicate those results. The field now faces a critical juncture: Can researchers overcome methodological hurdles to prove psychedelics’ efficacy, or will the initial optimism fade under closer scrutiny?

AI and Psychedelics: A Convergence of Futures

Interestingly, AI may play a role in solving psychedelics’ research challenges. Machine learning models are being used to analyze brain scans of individuals under the influence of these compounds, helping scientists understand their neurological effects. Some researchers believe AI could identify which patients are most likely to benefit from psychedelic therapy, personalizing treatment in ways previously impossible.

As OpenAI pushes toward autonomous AI researchers, and psychedelic science grapples with its own complexities, one thing is clear: The intersection of artificial intelligence and human biology is set to shape the next decade of discovery. Whether these advances lead to transformative breakthroughs or cautionary tales remains to be seen—but the race is undeniably underway.

The future of research may not be human versus machine, but human with machine—if we can navigate the challenges ahead.