■ SAFE RISK ■ Science & Research
Robots already do most of the actual science more cheaply, yet the astronaut corps keeps growing because human spaceflight is a political and commercial product as much as a scientific one.
“Robots explore space cheaper. But humans in space make better documentaries.”
Our AI replacement risk score — how we score jobs
The job is far more training than flying. Years of systems study, neutral buoyancy lab runs for every spacewalk task, robotics arm certification, Russian or partner-language study, survival training, and repeated simulator failures designed to break your decision-making. On orbit it is maintenance: swapping water processor components, troubleshooting a failed pump, running experiments as a technician for principal investigators on the ground, exercising two hours a day to keep bone density, and doing public affairs events that justify the programme's budget.
Automation has genuinely displaced the exploration case. Robotic missions returned samples, roved Mars for years, and mapped outer planets at a fraction of the cost and none of the life-support risk. On the crewed side, vehicles now fly themselves — docking is automated, ascent and entry are software-managed, and a modern capsule's touchscreen replaced panels of switches. Ground AI schedules crew timelines, monitors telemetry for anomalies, and increasingly plans experiment sequencing. Robotic arms and free-flying assistants handle inspection tasks that once required a spacewalk.
What keeps humans aboard is partly capability and mostly meaning. Improvisational repair of an unanticipated failure, dexterous work in an environment not designed for robots, and biomedical research that requires human bodies as subjects all still need crew. Beyond that, national prestige, treaty presence and the sheer narrative pull of people going are what unlock the budgets — no legislature funds a nine-figure programme so a rover can take a selfie. Commercial stations and lunar programmes are expanding flight opportunities, not shrinking them. Our risk score of 21 reflects a role that is more supervised and more automated but structurally in demand.
Automatability: our editorial assessment of current and near-term AI capability
Automation already took the flying. Through the 2030s expect more robotic assistants aboard, more autonomous station keeping, and AI handling anomaly triage and timeline planning. But commercial stations, lunar surface programmes and national prestige projects point to more crewed seats this decade and next, not fewer. The astronaut role is safe past 2040; what changes is the skill mix, tilting toward science, medicine and improvisation.
Purely on science per dollar, robots usually win, and that has been true for decades. Humans get sent for improvisational repair, dexterous work in environments built for people, biomedical research that needs human subjects, and — honestly — because national prestige and public imagination unlock budgets that a rover never would.
It changed the mix. Modern spacecraft fly, dock and re-enter under software control, and ground AI handles telemetry monitoring and timeline scheduling. Crew time has shifted toward maintenance, science operations and the failures nobody anticipated. Astronauts increasingly supervise autonomy rather than manually operate systems, which is a different skill set, not a smaller job.
More, on current trajectory. Commercial space stations, lunar surface programmes and expanding national corps all add seats, and private missions have widened who flies. It remains one of the most competitive selections on earth, but the number of people going to orbit each year has been rising rather than falling.
A doctorate or serious operational expertise in a technical or medical field, plus demonstrable performance under stress and in isolated teams. Test pilot backgrounds still count, but science and engineering depth carries more weight as vehicles automate. Fluency in partner languages and genuine public communication ability both matter more than candidates expect.