OpenAI just graduated its second class of Fellows, and the results are hard to ignore. Six people walked in as machine learning beginners and walked out six months later as full-time technical staff — a conversion rate of 100%. The apprenticeship model isn’t new, but seeing it applied this effectively to AI research is notable. These weren’t traditional CS grads padding their resumes. The program deliberately pulled from other quantitative fields — classical music, statistics, mathematics — betting that deep expertise in one domain could cross-pollinate with machine learning in unexpected ways. That bet appears to have paid off. Each Fellow embedded directly within an OpenAI research team and produced a novel project, contributing to the organization’s core work rather than operating on some academic side quest.
The announcement doubles as a recruitment pitch. OpenAI is now accepting applications on a rolling basis for the Summer 2019 cohort, which kicks off in July. The message is clear: if you’re smart and rigorous in some other scientific discipline, they believe they can turn you into a world-class ML contributor in half a year. Whether that scales is an open question. Six people is a small sample size, and the selection process is undoubtedly brutal. But the fact that every single Fall Fellow got hired full-time suggests the program works — or at least, that OpenAI is very good at picking people who were already going to succeed.
There’s also an educational component worth flagging. OpenAI open-sourced part of the curriculum used to train these Fellows, releasing a tutorial called “Spinning Up in Deep RL.” It’s not a collection of passive reading material. The package includes working RL code, exercises, documentation, and tutorials designed to make someone a competent practitioner — not just someone who can nod along at conferences. For anyone curious about what it actually takes to bridge the gap from beginner to contributor, this is a concrete artifact to study. It’s also a savvy move: open-sourcing the curriculum extends the program’s influence far beyond the handful of people accepted each cycle, and it gives potential applicants a way to self-assess before throwing their hat in the ring.
