I’ve noticed a huge push for every project lead to learn deep learning, but is it actually useful? I feel there is a growing <skill gap> between those who know the math and those who just need to manage the delivery. Is deep technical AI knowledge for leaders overhyped?
3 answers
From my experience in Silicon Valley, the expectation that a PM should understand backpropagation is definitely overblown. While foundational literacy is great, the real value lies in knowing how to deploy models, not build them from scratch. I’ve seen many leads spend months on theory only to realize they can't manage a roadmap effectively. Focus on the lifecycle and ethics instead. The market is saturated with "technical" managers who can't actually lead a cross-functional team through a deployment crisis.
Do you think the issue is the depth of the math or just a lack of understanding regarding the AI infrastructure requirements?
I agree. Soft skills and risk management in AI are much more critical than knowing how to code a neural network in a leadership role.
Exactly, Heather! Most projects fail due to poor scope definition, not because the manager didn't know the specific layers of the model.
Brian, I think it is both. Most managers get lost in the "how" of the algorithm rather than the "why" of the business case. We need people who can bridge the communication between data scientists and stakeholders. That is where the real value is in 2024, not in manual tuning.