From the source
Microsoft Research published a podcast episode in which partner research manager Jennifer Neville discusses how evaluating AI systems beyond standard benchmarks reveals surprising failures and what those failures teach about improving performance.
- Neville, who also holds a chair at Purdue, describes her own circuitous path into AI research, starting from math and physics through cognitive science before finding computer science.
- She explains that mandatory research in her honors program led her to statistical relational learning, which hooked her on the thrill of understanding something at the frontier.
- The episode asks what can be learned when both human and artificial paths do not unfold as expected.






