[Shinmon]

A full career switch will likely be hard. For everyone working on stable diffusion or GPT3 there probably 1000s of people working on rather mundane tasks in ML (nothing wrong with that).

If I were you I would leverage your existing knowledge in robotics and see how AI might apply there.

For example, computer vision will very likely be a major part of robotics. I could also image that there will be Speech2Text models where you can just tell a robot what to do. However, the main driver here is not necessarily the AI model but the people understanding how to use it in the context of robotics.

I would argue that plain AI/ML is highly scientific work and unless you have a PhD in ML or a similar field it will be extremely hard to get into it. Applying models, tweaking them for your use case and all the work around that will create the actual value in most applications.