It's basically masters/PhD. A lot of ML involves research work and those degrees are strong filtering mechanisms for selecting that skill. You caaaan do ML adjacent work but it's a lot of the grunt tasks and you'll grow frustrated in about a year.

Source: PhD student. Collaborate in NLP/CompLing lab. The recurring refrain mong the masters students in that lab is they wanted to kill themselves at Conversational AI jobs.

Piggybacking on this. There are a lot of data science jobs that hire out of undergrad. A lot of the work is actually more data analysis and simple business analytics, but you can get started with a broad data science role, learn on the job, and specialize in ML

Yeah, I may have come off a bit too negative because of second hand responses. You'll still have good opportunities to acquire experience and build a starting resume. The glass ceiling just comes down hard for some people if interested in mainline ML.

Yes, but it's hard and you have to be good and get lucky. Internships are the way.

>A lot of ML involves research work and those degrees are strong filtering mechanisms for selecting that skill.

Why is that filtering mechanism for that skill?

Because thesis and dissertation work require some level of independent research to simply receive the degrees.

Ok, so that proves that someone has the skill. But when someone doesn't have master/phd, that doesn't prove that someone doesn't have that skill. In other words, if someone has no master/phd degree, but has published a research paper, then they have also proved to have that skill, so it should be possible for someone with bachelor / no bachelor to get a job in ML. Is that correct?

No, many hiring managers will use arbitrary criteria to reduce the number of applicants they need to evaluate for a job. Degree requirements are one of those. While yes, there probably are a few people who are in ML jobs without meeting degree requirements, in general, you're going to struggle without them.