8. The Role of AI in Research

The Role of AI in Research

The discussion considers how AI is changing research and how students and researchers can use it responsibly. Professor Krishna Nandiada argues that AI is moving beyond being an occasional tool: it is becoming woven into everyday work, much as high-level programming languages became standard for software development. Researchers across disciplines will need to adapt while retaining responsibility for the work they produce.

Main Ideas

• AI is relevant across disciplines. Its use is not limited to computer science; researchers in the humanities, engineering, mathematics, management, and the sciences can all use AI tools.
• Learn to use AI, but stay in control. Researchers need not understand every technical detail of a model, just as drivers need not understand every mechanism inside a car. They should, however, understand how to operate the tool, what its strengths and weaknesses are, and what consequences its use may have.
• AI can assist throughout the research process. It may help with literature searches, identifying research problems, brainstorming hypotheses, developing prototypes, implementation, and validating proposed solutions. Its suggestions are inputs to assess—not replacements for a researcher’s judgment, creativity, or a supervisor’s guidance.
• AI outputs can be plausible but wrong. AI can hallucinate, including by inventing papers or references. It generates answers from patterns in its training material, so fluent, confident wording is not proof of correctness or understanding. Even as tools improve at filtering reliable information, their outputs still need verification.
• Researchers must retain ownership. The researcher—not the AI—is accountable for the final hypothesis, methods, code, findings, and thesis or paper. Researchers should be prepared to explain and defend their work, disclose AI use where required, and never excuse an error by saying that the AI produced it.
• AI should not be avoided out of pride or fear. The guest advises researchers not to treat AI use as an ego issue. It can speed up work and enable researchers to do more, provided they use it deliberately rather than blindly.
• AI can speed up prototyping, but careless use can create extra work. AI-generated prototypes may be produced quickly. If the researcher does not understand them, debugging and repair can take longer than building the work carefully in the first place. The same principle applies beyond programming: use AI only where you can evaluate and supervise its contribution.
• AI is not a substitute for human insight. It can suggest problem statements or hypotheses, but it cannot determine by itself whether an idea is meaningful, feasible within a PhD’s limited timeframe, or genuinely worth pursuing. Researchers and their guides remain important in making those judgments.
• AI is not a “god” or a reliable predictor of the future. The guest uses the historical belief that the Sun revolved around Earth to illustrate how a system trained on prevailing claims could confidently repeat them. Researchers must question AI carefully, especially when an answer challenges or supports an existing assumption.
• Open-source AI may broaden access. The guest describes a growing range of open-source models, including models that can run on personal computers, though hardware and resource demands remain obstacles. He sees the push toward more affordable and accessible models as a way to reduce the gap between well-funded and less-resourced researchers.
• Research may shift toward harder and more creative problems. As tools take over some tasks requiring less human intervention, researchers may be able to focus on more complex, novel, or creative questions. The guest presents AI as a supervised assistant or “friend,” not a fully autonomous researcher.

Practical Guidance for Researchers

• Do not avoid AI, but do not trust it blindly.
• Learn the tools and their limitations before relying on them for research tasks.
• Use AI at appropriate stages—from early reading and literature exploration through prototyping and thesis preparation—while evaluating each contribution.
• Verify sources and claims. Check that cited papers exist and that factual answers are supported by reliable evidence.
• Assess ideas yourself. Decide whether suggested topics, problems, or hypotheses are interesting, original, feasible, and suitable for the scope and timeline of a PhD.
• Discuss research decisions with your guide. AI may offer options, but student–supervisor judgment remains important.
• Understand AI-generated work well enough to maintain it. In particular, do not rely on generated code or simulations that you cannot inspect, test, and troubleshoot.
• Keep human oversight throughout the process. Treat AI as an assistant that requires supervision, not as an autonomous authority.
• Take responsibility and disclose use when required. Follow publication or institutional rules for acknowledging AI assistance, and remain accountable for the final work.

Speakers and Sources Featured

• Professor Krishna Nandiada — Guest, described in the subtitles as a former head of department and a professor in the Department of Computer Science at IIT Madras.
• An unnamed interviewer/host — Introduces the guest and asks questions on behalf of the research methodology course.
• Sources and examples mentioned in the discussion: AI tools and open-source language models; the Hugging Face model repository; ANRF and DARPA as examples of funding contexts; books and research papers as material AI may draw on; and analogies involving programming languages, cars, calculators, and the historical Sun–Earth model.