Why AI on its own is not a product | Marily Nika, AI PM Bootcamp

Summary

Marily Nika opens by stating that artificial intelligence on its own is not a product. Reflecting on her early career upon joining Google, she describes entering a room of scientists and researchers who demonstrated technology capable of converting spoken sound into text. While the team showed that speaking the word hi allowed the system to type the letters H and I, they were unsure of how to use the technology commercially. The team faced unanswered questions regarding what problems the technology could solve, how to prioritize those problems, how to bring the software to market, and how to find product-market fit. Nika explains that early in her career, she realized that even when surrounded by senior researchers and academics, the product manager's primary responsibility is to bridge the gap between technical capability and user value while maintaining a viable business.

To resolve confusion among students struggling with overlapping terminology such as classification, regression, prediction, control, and optimization, Nika introduces the AI, Algorithms and Applications Map. The framework arranges technical concepts systematically across distinct layers. At the foundational layer are machine learning methods, specifically supervised learning, unsupervised learning, self-supervised learning, and reinforcement learning. Nested beneath these learning paradigms are specific algorithmic classes, including classification, regression, and clustering. These algorithms in turn feed into concrete application layers, which encompass real-world tasks such as matching, medical diagnostics, financial trading, and chatbots.

Nika emphasizes that product managers must never begin their workflow from the middle of the map by picking a learning method or algorithm such as supervised learning classification. Instead, product managers must start with user problems and user goals, working backwards to formulate smart solutions and flesh out detailed product requirements. Once the requirements are defined, the product manager hands them to data scientists. The scientists then determine the technical implementation, identifying suitable algorithms and presenting trade-offs for collaborative discussion.

The presentation concludes by highlighting that underlying user use cases do not change over time, as user goals remain constant. What changes with artificial intelligence is the methodology used to solve those problems. By using machine learning, teams can address established use cases in ways that are better, smarter, faster, cheaper, and significantly more convenient.