Plenty of Work for Humans still left in the ML World
We are living in an age of unprecedented growth and power of advanced AI models. The incredible achievements that the best of these modes are capable of impresses even the most hardened technoskeptic. The pundits are increasingly lamenting not the overhyping of AI, but rather its seeming untrammeled growth. In all of this its good to remember that there are still plenty of problems for which the Large Language Models and their brethren are no match. Most of these are niche applications of ML, datasets and use cases for which bigger is not always better, and the power comes from grasping the context, not brute compute. You may think of these as “niche” applications of ML, but oftentimes these niches are entire sectors of the economy, such as insurance, finance, retail, etc. Each of them requires a careful bespoke approach, with a lot of room still left for human expertise, insight, and judgement. None of it will be made redundant any time soon.