Comment by karmakurtisaani
3 years ago
I would almost say you're better off staying in your current role than going in to ML. Here's why:
* Everyone is into it these days. We were interviewing some fresh graduates for a junior quant developer role recently, and each of the candidates had some mention of AI in their CV (the role had nothing to do with AI). The ML courses out there are packed with students thinking they'll be working for OpenAI or Google (took a class out of personal interest recently).
* Related to the above, you probably need a PhD to do anything remotely interesting. The bulk of the AI work is setting up build pipelines and finding/cleaning data.
* The field has a major winner takes all factor to it. The biggest companies will develop their own systems (or rather, a handful of top school PhDs in these companies will), and the rest will use tools made by others. The trend is towards bigger models with more training data, small companies will have a hard time competing in this space.
* Available jobs: check LinkedIn job postings for Java developer vs. ML researcher. I haven't done this, but have a strong feeling how it will look like. You fear the dev jobs will be obsolete soon, but why would the ML jobs not be?
To balance this out, I do have a few counterpoints:
* ML is a helluva lot more interesting than plain old java coding.
* To contradict myself a bit, I was recently almost offered a position with a strong focus on ML. Unfortunately the project ultimately fell through.
* Maybe the LLM success will cause VCs to pour money on AI start-ups, and there will be more opportunities in the space - for a while. However, to become hireable for ML roles, you'll need a year or two of intense study.
I began my dive into ML a while back (remember when Data Science was the new hotness?) and quickly realized that a lot of the most interesting work really is "winner take all" in the sense that if you can't compete at the cutting edge of your field, you're really going to struggle to show value to your employers. As someone learning ML on the side with no plans to go all in, I decided just to learn how to use ML models as a tool and continue to focus on my main job as a software engineer.
Good point, competing with the younglings steeped in ML is hard. Better to bring strengths that come with experience in adjacent skills.
OP might do best to leverage both their existing Java/full-stack skills and their newly developing ML skills. Not sure what’s a good way to achieve that though.