Occasionally I pitch in as part of the crowd in the crowdsourced mapping site Tomnod. The site has periodic campaigns in which one contributes to a humanitarian task such as looking for hurricane damage in Vanuatu or signs of poachers in Africa. Recently, a new campaign type has appeared: volunteers training the site’s algorithms to spot certain features. In the campaign pictured, we are reporting whether or not there are buildings in the pink area, differentiating them from ponds, crop piles (I think), etc. The algorithm will take our human skills and learn how to do this on its own, ultimately aiding in mapping rural Asia for health delivery.
This is a leading example of what we are finding in our research on the future or work: automated systems are learning how to do ever more tasks, at an accelerating pace. They may need human help to get going, but a system only needs to learn something once. After that, it is likely to do the task better or faster than the humans that trained it, and the algorithm can be multiplied as needed.
This could lead to one of the dystopic versions of the post-job future. The analyst John Robb puts it this way:
As you can imagine, training bots to do everything a human mind can do is going to be a HUGE industry. An industry so big it is going to create some of the biggest companies in the world (Turk companies could employ hundreds of millions of people all over the world, making them 100x larger than the largest employers in the world today). You can guess what this dynamic will look like. Micro-loan offered at extortionate interest rate financing training for turking job. Turking job lasts a couple of months. Earnings are garnished to pay loan. Bot eats job. New loan required for more training. Cycle repeats.
(“Turk companies” is a reference to Amazon’s Mechanical Turk, the first market for crowdsourcing micro-tasks to human workers.)
There are other, better futures, but much of what we have learned in the future of work project suggests this one cannot be dismissed lightly.