Classically, what you're looking for in a stable career is to be part of a generational cohort that is well-positioned to transition from something emerging, risky, and low-value, into a more established senior role. And then stay there forever, holding the expertise captive until retirement, at which point the economy now has to figure out what to do without your indispensable knowledge.
This is actually a process that has been a big part of recent news: the Boomer generation, being oversized and long-lived, sucked up a lot of the jobs. By the time their Millenial children arrived, a lot of career paths had no emerging prospects because the parents were still there, still in the same roles, and the world had become fantastically more competitive through globalization. But now they've finally begun to age out altogether, and that transition plays into the dynamic of high unrest, financial turmoil and anxiety around tech that we're experiencing.
The thing is, tech is something societies decide to invent pragmatically, based on what available science allows. We invent "automobile tech" because policy and available resources supported its widespread use in the richest parts of the world. In the ones destroyed by WWII, auto tech still existed but was complemented with reinvestment in rail tech.
So with ML AI, the tech is something we're currently looking for models of application for. The science is cool, and there are certain things it's great at, but contra the "learn AI" replies, that doesn't mean the "AI industry" is something you'll gain the most leverage from by approaching it at its most fundamental level, getting a degree in, and then simply signing up for a research job. That is one of the most competitive routes you could take, as the "easy part" of the field is already behind us, and now everything is going to be about little nuances of improving the tech and integrating it better.
What we currently see in terms of AI users is relatively unsophisticated application: people who log on to one of the apps, send a few basic prompts, and then stop there, satisfied with the result. But I believe the way to think about this is rather to become sufficiently AI-literate to use it to storm the gates of some other field, as a combination threat; a layered synthesis of traditional know-how and new methods. This is why some artists are unconcerned, while others are panicked; one group sees a way to enhance what they're doing with another layer of tech, another sees a threat to their routine illustration and asset creation gigs.
But in "storming the gates" you have to expect to arrive at a surprisingly empty field, where nobody knows what you are doing or whether it's valuable. It might take several years to build up visibility, just trying things and publishing results, before you find a path to monetize on it.
If the basic idea you're building from is coherent - it doesn't contain contradictory elements, and you can use it as a philosophical framework to assess the value of your output - you will stay motivated and eventually succeed. Study enough philosophy to get what it means to be coherent and build such a framework; it'll pay off.