Key takeaways
- Emphasize intelligibility and explainability for AI in space missions to enhance trust and safety.
- Explore the philosophical roots of AI through perception, cognition, and action for better human-AI interaction.
- Develop innovative AI solutions, like MEND, to predict satellite failures and improve operational safety.
- Implement educational initiatives like Gyaan to expand AI accessibility in schools across India.
Artificial intelligence is often seen as a natural partner for space exploration. Missions demand automation, agility, and precision, but the risks of black-box models make adoption difficult. Nitish Kumar, scientist at ISRO and recipient of the Innovative Student Projects Award by the Indian National Academy of Engineering (INAE), explained why intelligibility and explainability are central to AI in space.
Addressing the gathering at Cypher 25, India’s largest AI conference organised by AIM from September 17-19 in Bengaluru, Kumar said that while AI and space appear aligned, the reality is complex. The sector is “automation hungry,” he noted, but reluctant to trust opaque models. “Spacecrafts demand explainability, agility, and assurance,” he said.
Innovation as Thought Work
He broke down intelligence into perception, cognition, and action, linking AI systems to human self-reflection. His questions—“can we perceive cognition?”—highlighted the philosophical roots of ISRO’s work.
“One of the bottlenecks [of developing AI] is we don't understand our own thinking; it is very difficult to understand how AI thinks,” he said. For him, “perception without aberration of thoughts is the higher form of intelligence.”
Solar Cells to Space Doctors
Kumar illustrated how abstract ideas drive practical breakthroughs. Observing bats led him to reimagine solar cell defect detection through “distraction” rather than attention, that achieved near-perfect accuracy. The same approach powered MEND, an AI system that predicts satellite anomalies 10–15 minutes before failure, giving ISRO a vital safety window.
His team has also experimented with generative AI. Gyaan, an educational assistant, has reached 22,000 schools, which started in Bihar and is available to all.
Gyaan is an online platform that makes education accessible to everyone. Users can ask questions specifically referencing the NCERT syllabus, for subjects like Maths, Science, and English. At the moment, it covers the syllabus for Classes 8, 9,10,11, and 12 and supports interactions in Hindi, Bhojpuri, and English languages.
For Kumar, AI is more than engineering. “If AI is not deterministic, we cannot apply it to space technology,” he said. Comparing it to the industrial revolution, he argued that AI represents a shift “from muscle power to steam power to electric power to nuclear power. This is a totally different game. It is a mental game.”
He closed by urging India to shape AI globally, not just apply it locally. In his view, agility and intelligibility will decide whether AI can be trusted to operate in space, where uncertainty is the only constant.
