In a recent discussion, Jonathan Burley, Director of AI at Bloomberg Industry Group, emphasized that building effective AI products requires a scientific mindset and a focus on real-world problems rather than chasing the newest large language models.
Burley, who transitioned from modeling climate systems to leading AI strategy, argued that the key to successful machine learning and data science projects is starting with the right questions. He introduced the concept of the "minimum viable experiment" as a way to test hypotheses quickly and avoid the pitfalls of flashy generative AI demos that lack practical value.
"It's not about the technology first—it's about the problem," Burley said. "A scientific approach helps you avoid wasting resources on solutions that don't address actual needs."
The conversation, hosted by Tamr, explored how adopting a rigorous, hypothesis-driven methodology can lead to more robust AI applications. Burley's insights serve as a reminder that even as AI tools evolve rapidly, the fundamental principles of inquiry and experimentation remain critical.