What is your strategy in the midst of AI race?

[berbere]

I am a data scientist/statistician who already has a very hard time finding a decent job. The rapid pace of AI improvement deeply gives me anxiety regarding the job market. I am convinced that within a 2-3 year timeframe, the job market will narrow drastically, and some jobs will become extinct. Given all this, what is your strategy? I am not sure which topics or areas I should invest in learning. Any recommendations would be greatly appreciated.

[wryoak]

I think you have the wrong approach in fretting about what you should learn rather than taking the opportunity to pursue what you want to learn. Don’t chase the market, chase your life. The market will keep going but your life will eventually end. Don’t waste it trying to please an entity that has no consideration for you (remember you can still profit off the market even if you’re not selling high value labor to it).

[panic101 → wryoak]

Precisely. AI is hyped right now, and all its flaws are being ignored. Also over time regulators will get involved and put more restrictions on AI.

It was the same with cryptocurrencies and driverless cars.

[miguelaeh]

This is something that worries me a lot too. For context, two years ago I left my job at a big tech company to build a startup. It failed, but I still have the energy to build new stuff, so that's what I am doing. My internal conflict today is that when I left it, my worst-case scenario was to just go back to a similar job but with more experience. Now, with AI, everything is going to change, and maybe the opportunity cost of what I am doing today is extremely higher than I expected it to be when I left in case nothing goes well.

[gregjor]

My strategy is to ignore it. So-called AI is 90% hype and bullshit at this point. FOMO and companies trying to create markets so they can recoup the massive amounts invested already.

Worry if AI that can do your job cheaper and better than you can ever comes along. It shouldn’t be hard to stay ahead of LLMs — they are running out of useful training data and electricity.

[illuminant]

Creating profiles in some civilized markup (yaml?) in which human expectations may be explicitly laid out from which automated systems generate their solution.

Not everyone remembers that all this will forever relate to actual human expectations and requirements.

We will not evolve if we cannot correlate behavior and requirements, and it is obvious that LLMs who do not know make up convincing bullshit.