Recently I gave a presentation on AI in research to the African Doctoral Academy. Here are the slides (unfortunately, my recording of the session didn’t include my audio so I don’t have a video).
And here is an overview of what I discussed.
This presentation introduces generative AI, explaining it as a next-word predictor capable of generating coherent multimodal content across a wide range of subject domains. Generative AI competence is also rapidly improving through advances like plugins and APIs.
I encourage the perspective that generative AI as an expert system that we can leverage, rather than a source of information. It has extensive knowledge, understands complexity, communicates effectively, and is constantly improving. However, generative AI still has issues like hallucination and bias that require human guidance.
I provide an overview of several research roles for generative AI, including idea generation, literature review, data analysis, and grant writing. As the cost of training and deploying AI at scale decreases, ubiquitous access will enable new research applications and complements.
Human researchers will nonetheless retain critical roles like selecting meaningful problems, determining significance of findings, collaborating with each other, setting goals, interpreting outcomes, ensuring accountability, and providing context. The unique human contribution will remain essential but will be enhanced and extended through ubiquitous access to AI.
With responsible human guidance, generative AI represents a powerful new research tool, enabling new possibilities when complemented by human strengths. Overall, researchers should both utilise generative AI in their work and study its emerging impacts across domains.