I've been talking to jr developers and co-ops. lately, they are all interested in specializing in AI. Part of this seems to be a reaction to the feeling that "AI will swallow up all software development jobs".
But the reality in field is very different at the moment. Right now, many ML companies are losing money. ML systems are often flaky and churn is common if the value proposition isn't followed up on. They raised money during a period of low finance charge rates and are now struggling to acquire patrons and meet boards new expectations around profitability.
Other companies that are making money in the ML space tend to be selling ML tools to ML devs or ML companies, like hugging face, wandb, or those companies that provide inference as a service. Those companies are more software companies than they are ML companies, and if they have ML features, they are usually as "accessory features" rather than part of the core product.
However, lots of traditional companies are adding ML features to their product as well. That's a good thing, right?
Well, yes and no. Rarely, some of these companies may end up adding ML features which become a core part of the product.
However, many ML features may never leave alpha or beta and will end up shelved because they are too expensive to maintain and operate compared to the value they provide users. ML products and features are even more challenging than regular software products to build, meaning their value proposition needs to be higher to justify their creation. company leaders don't understand the total cost of ownership of ML products very well yet, and many are getting burned.
All that said, there are plenty of companies building products around ML and making money successfully. However, it seems like this is more the exception at the moment, rather than the rule. Even in these cases, there tends to be a higher concentration of SWE as compared to MLE.
Specializing in AI/ML is currently risky, akin to specializing in web3. its not clear how much AI work there will be in the long run. People specializing in this technology now may end up fighting over a small pool of quality jobs.
If you are genuinely very concerned about AI taking over all software jobs sometime during your career, it is likely better to attempt to specialize in a particular field, where you can be focused on solving the problems in that field, including AI. Being a domain expert in developing software for healthcare, money management, agriculture, etc, will provide more job security than specializing in a technology which hasn't yet proven itself, in my opinion.