Ask HN: What angle of startups or AI isn't covered by YouTube creators?

(This is my first comment on hacker news) but I would like to see more content targeted to people who know how to program but don't know the slang and words in this specific field. For example Sebastian Lague did a great tutorial on machine learning https://www.youtube.com/watch?v=bVQUSndDllU I would love to see something like this but from the angle lets pick a random programming language and lets program this together from scratch.

In general, most content on YouTube is a bit pedestrian. I would like to see more serious academic resources referenced, longer and more in-depth videos, and better production values. Something like Every Frame a Painting but for other topics.

Everything. The state of YouTube content on ML is similar to what the state of software engineering would be if there were only videos on "Hello, world!" or conference talks about awesome products demoed on "Hello, world!" problems.

Practically none of the content creators seem to have ever executed a project that used ML, or worked in an ML team, and those who are doing that professionally don't do YouTube videos, except for talks in conferences that they did something, not really how they did it.

I think you are in a great position and can do a lot of good by showing how you and your team execute a project from start to finish.

Seasons and episodes. One season per project. One episode every week or two that addresses the problems solved, and how. I want to look at your screen and listen to conversations and brain-storming. How do you do ML, in long form videos from the field, not some bullshit 5 minutes videos with sentences like "then you only have to deploy your model". How do you collaborate on a project, how do you choose metrics that translate to the real world, how do you retrain your models, how do you do workload management, how do is your private cloud, do your ML practitioners have self service infra for training, how do you do self service deployment, manage data and access, how do you test data is not tainted, deploy and expose models, embed models.. How do you collaborate. How do you track knowledge, decisions, and assumptions and hypotheses.

For example, if you could take papers like "The ML Test Score: A Rubric for ML Production Readiness and Technical Debt Reduction"[^1] or "Hidden Technical Debt in Machine Learning Systems"[^2] and show how your team is dealing with each point, it would be great content.

All the thing that anyone who has spent a day in the real world knows isn't covered by 99% of the content online because the people doing it are either too busy, consider it competitive advantage, or don't know it even exists (had people working at FAANG try our internal platform, and they don't even deal with these problems because they're in "research" and they have teams of ops they can offload their notebook to).

[^1]: https://static.googleusercontent.com/media/research.google.c...

[^2]: https://papers.nips.cc/paper/5656-hidden-technical-debt-in-m...