can crowdsourcing create the jobs of tomorrow?
Crowdsourcing is the hot new thing in the sciences. Social sciences look at how people solve problems in an extremely large group. Biologists, astronomers, and physicists are designing games to harness the creativity and energy of an army of dedicated players to find exoplanets, find how proteins fold, and solve cosmological mysteries no supercomputer seems to crack with any degree of efficiency. Intelligence agencies want to get a vast network of informants to get real-time data about every global hotspot from those on the ground. And a lot of tech executives are thinking of ways to monetize the wisdom and power of the networked crowds. That’s not an easy task since crowdsourced services must be reliable, instantly available, and standardized. How would a company get predictable, reliable, quality work from a loose association of people on the web? The military, intelligence agencies, and scientists with addictive games can lure in participants by the tantalizing natures of their tasks. But who wants to sit there and sort business addresses all day for peanuts? Well, a team at MIT decided to focus on at least one facet of monetizing crowdsourcing: making the crowd respond quickly.
Speed was something they certainly got. Workers responded to task alerts within 500 milliseconds when the alerts came within about five seconds of their pre-recruitment for a task. Even 20 seconds out, the responses still came in under three seconds. Not bad at all for a crowdsourced real-time system but then again, it’s not a startling result because the crowd was being prepaid for its services, limited to 50 people, and tasks came at them quickly and furiously. So we know that the crows replied quickly but we don’t know how well it did or if its reliable enough to be prepaid for projects more complex than making a simple classification or clicking some buttons on a form. A fun fact mentioned in the paper was that spam is actually a really massive crowdsourced endeavor paying about 0.5¢ per word. You see how well that works, right? Or rather you don’t because you live in the 21st century and have spam filters for your e-mail and IM. But really, do take a trip to your spam folder to see how many ways one can misspell words you wouldn’t ordinarily think would be that easy to misspell and the sheer amount of hours and effort wasted on something that won’t be seen by some 99.999% of its targets and if it is, promptly deleted and blocked. Bottom line: crowdsourcing menial tasks isn’t all that great.
Even worse, with enough computing grunt and a big enough server farm, you could do the work of a massive crowd in performing said menial tasks, and probably with higher accuracy because your program will not drift off or get bored. It’ll be cheaper and more reliable too. Crowdsourcing shines when users are given complex and challenging things to do. Scientific research currently tapping dedicated gamers and amateur astronomy buffs are seeing their data analyzed faster and faster and veteran users are tackling more and more complex problems which yield possible answers to questions that stymied small labs unable to throw adequate brain power at the problems. And that may well be where the future is, crowds being paid to tackle complex issues, design new products, and come up with practical applications for new ideas to monetize them, earning cash rewards in the process to keep them engaged and excited. Maybe the future of work isn’t sitting in a drab gray cubicle from 8 am to 5 pm, but lounging in a coffee shop, playing a game that could help molecular biologists unlock new pathways for more effective drugs, or helping programmers fine-tune their designs for oh, say, I’m just going to guess an aspiring AI framework or something like it by experimenting with it, and getting paid by labs and corporations for all their efforts and the mountains of useful data they’ll generate?
See: Bernstein, M. et al. (2012). Analytic methods for optimizing realtime crowdsourcing arXiv: 1204.2995v1
