Hacker News Reader

Comment by codingdave | Hacker News Reader

And most product managers would confirm that most companies do not need a good machine learning pipeline.

As an older engineer currently having fun digging into ML: yep, a well thought-out set of good-ol business rules can do a lot of what teams are applying ML to these days.

Nothing wrong with learning new things, though. I can get behind that. It's important to stay intellectually honest, however.

On the other hand, research efforts on vision, language and autonomous systems are genuinely pushing boundaries.

Supervised learning using a variety of different neural network pipelines is doing some genuinely impressive things from a business perspective. But not much beyond that at this point (again, from a business perspective).

Business rules are concrete and have people and names attached to them. ML allows the robot to take all the blame.

Really? Because most product managers I meet are the ones going “how can we put AI and Machine Learning into <some task that doesn’t need it>?”

I 100% agree that most companies don’t need nearly as much ML as think though.

full disclosure: I'm a PM in FANG

I think that's what the bad PMs do. Good PMs think 'what is the smallest possible thing we could do to start truly adding value against our problem space?'

A good heuristic is that good pms are problem oriented - bad pms are often focused on solutions first

Reply on news.ycombinator.com