My experience (as a non-CS person) has been that aside from where there is a very large maths component which might block people without further academic-style study (and I wouldn't necessarily even count ML in that, since the maths needed for much of ML is relatively low level, it's certainly not graduate school level understanding maths), there are relatively few areas of software which have high barriers to entry in actually doing stuff - where the barriers are are people willing to take a risk on letting you have a go. That's usually much much easier once you're in a company than if you're applying for a role from the outside.
Every time I've felt like I didn't understand something and felt overwhelmed at the scale at a task, 3-6 months down the line of throwing myself at the problem and trying to understand it, I've realised it's not as hard and part of the barrier was just the unfamiliar terminology and unfamiliar tools. Sure, there is a degree of needing to learn new stuff - which is true in any job and in life - to do new things. But those barriers are not normally insurmountable. That's been true for me in basically every area. It is also why I'm fairly willing to give people a chance, so long as they are able to demonstrate some knowledge which would be able to transfer.
Every time I've felt like I didn't understand something and felt overwhelmed at the scale at a task, 3-6 months down the line of throwing myself at the problem and trying to understand it, I've realised it's not as hard and part of the barrier was just the unfamiliar terminology and unfamiliar tools. Sure, there is a degree of needing to learn new stuff - which is true in any job and in life - to do new things. But those barriers are not normally insurmountable. That's been true for me in basically every area. It is also why I'm fairly willing to give people a chance, so long as they are able to demonstrate some knowledge which would be able to transfer.