This week I had the opportunity to speak at Women in Data Science event hosted by 6sense and the Generative AI event hosted by CFA Institute of New York. I also had the opportunity to meet the some of the new Berkeley SkyDeck startups as an advisor. Some high level reflections:
While generative AI and LLM represent the tip of iceberg in machine learning and AI, many exciting things are happening in the space that are worth watching.
Regarding the implementation of generative AI and LLM, we have to consider data, model, and deployment dependencies among other things as part of due diligence.
No doubt, it is an amazing time to build ML and AI products. An important principle in building ML/AI product is even more important now -- problem framing and finding the right use case. Making sure that early adopters are reflective of the long-tail users. This is hard and will take time to uncover and iterate.
There are many industry problems that need to be solved that do not require generative AI and LLMs. So match the problem with the right (and/or most elegant and practical) models.
When building ML and AI products, always consider tradeoffs in both technical and non-technical areas. The sum of the tradeoffs must be positive for a sustainable and reliable deployment in an enterprise. This is hard and will take time to evaluate.