Hacker News Reader

Human in the loop machine learning. All the successful big players use it, but pretend it is all ML. Merely making the human explicit and finding the best trade-off between human and algorithm can do wonders for machine learning.

Plus it has better optics than "AI WILL EAT YER JORBZ". In this case human in the loop ML inherently creates jobs as an essential element.

View on HN

Not a bad suggestion, but are the optics really that much better? Definitely gives me a dystopian vibe of "paying humans pennies to help AI automate away what few other jobs they have left to them".

Depends where the humans live. Some areas of the world this is great work: good pay and safe work conditions.

Also depends on whether there is a limit to this work. It seems there is plenty to go around given the enormous amount of training data ML needs.

And as ML becomes more widespread the value of this work will increase and pay will go up.

Also, the current paradigm is a simple active learning approach. There may be a large variety of ways humans can contribute at various skill and hence pay levels. E.g. compare the work of original 'computers' to today's enormously well paid software developers. As one example consider the FoldIt protein folding game, which was very successful.

Reply on news.ycombinator.com