The Mechanical Turk
Call me cynical, but it's been a long time since I last saw something in the IT or Internet world that made me think "Yay!". Something that creates a buxx for me. Lots of neat web2 sites, lots of neat new devices, 10Gbps & 40Gbps networking, etc. Nice, but no real excitement.
Tonight, though, I came upon Amazon's Mechanical Turk, and this is such a cool idea.
This is the basic idea: in a larger system or process, some tasks are better done by people than by computers. So, the idea is to have a generalized interface that allows computers to ask humans to complete a task - so the application remains in the drives seat, and the human performs "Human Intelligence Tasks" (HITs) as part of the application.
They bill this as artificial artificial intelligence. Which it spot on, and just funny.
Now, Amazon of course drives this as a WebService, and sees themselves as a global marketplace for HITs, brokering tasks between requesters and people with skills. Maybe that's the next big thing. It's a neat idea, an idea with merit, and maybe yet another idea that will change how we think of "work". I don't know. But, closer to home, the MTurk model of doing things is applicable, I think, in lots of the systems I use and build - systems for installing networks, for creating grid computation setups, as a way of having applications ask an operations centre (any operations centre, even) to complete a task. It's certainly applicable to network provisioning systems, and much neater than traditional workflow systems that keep control outside the applications.
Looking broader at all the IT systems I've built, hosted, worked with, I can see so many ways this programming model is applicable. And if it is widely applicable, maybe the marketplace idea can take off, too.
This has merit. I will have to spend some time here
Tonight, though, I came upon Amazon's Mechanical Turk, and this is such a cool idea.
This is the basic idea: in a larger system or process, some tasks are better done by people than by computers. So, the idea is to have a generalized interface that allows computers to ask humans to complete a task - so the application remains in the drives seat, and the human performs "Human Intelligence Tasks" (HITs) as part of the application.
They bill this as artificial artificial intelligence. Which it spot on, and just funny.
Now, Amazon of course drives this as a WebService, and sees themselves as a global marketplace for HITs, brokering tasks between requesters and people with skills. Maybe that's the next big thing. It's a neat idea, an idea with merit, and maybe yet another idea that will change how we think of "work". I don't know. But, closer to home, the MTurk model of doing things is applicable, I think, in lots of the systems I use and build - systems for installing networks, for creating grid computation setups, as a way of having applications ask an operations centre (any operations centre, even) to complete a task. It's certainly applicable to network provisioning systems, and much neater than traditional workflow systems that keep control outside the applications.
Looking broader at all the IT systems I've built, hosted, worked with, I can see so many ways this programming model is applicable. And if it is widely applicable, maybe the marketplace idea can take off, too.
This has merit. I will have to spend some time here