AI Must Be Made Reliable And Understandable, Say Experts
Hyderabad: At the India Machine Learning 2025 Symposium held at the BITS Pilani Hyderabad campus, researchers said AI systems can no longer be treated as black boxes and must be able to reason, explain their decisions and operate safely as they are increasingly used in real-world environments.
Many speakers said the next phase of AI research is less about making models larger and more about making them understandable and reliable. Saurabh Prasad, a professor at the University of Houston, said the work at GeoAI shows how combining data across sensors, space and time improves decision-making only when systems can explain how conclusions were reached.
Abhilasha Ravichander, a research scientist at the Max Planck Institute for Software Systems, said studying how large language models store and organize information is necessary to reduce opaque behavior and unexpected output.
The need for clarification has become even more acute in multimodal artificial intelligence discussions. Mistral AI research scientist Khyathi Chandu said harmonizing voice and language helps systems respond more naturally and consistently.
Hisham Cholakkal, an assistant professor at the Mohamed bin Zayed University of Artificial Intelligence, said multimodal language models increase capability but also raise questions about accountability if decisions cannot be monitored.
Trust and evaluation stand out in the sessions. Sunayana Sitaram, Microsoft’s chief applied scientist, said AI systems still perform unevenly across languages and cultures, making clear evaluation standards critical.
Soujanya Poria, an associate professor at Nanyang Technological University, said existing language and multimodal models have reached their limits and new research approaches are needed rather than incremental scaling.
The sessions also reflected the same concern. Mercy Ranjit, senior researcher at Microsoft Research, demonstrated how healthcare AI agents are designed with security controls built into workflows. Mrinmaya Sachan, assistant professor at ETH Zurich, talked about AI tools in education that prioritize transparency about how results are produced.
A panel discussion on India’s AI future saw researchers calling for stronger academic research, better support for PhD academics and closer university-industry collaboration to create reliable AI systems. The program also included a datathon and early career talks focusing on security, privacy and interpretability.
