Tuesday, April 13, 2010

Labor Implications of Content-Aware Fill

So I've been checking out a lot of the the videos and commentary on Adobe's new Content-Aware Fill. I find several things about the online discussion to be fairly amusing.

Firstly, there are the people who claim that this is fake. I can sort of understand this, as it was originally posted around April Fool's. However, there's nothing all that funny about this particular technology, and nothing all the unbelievable. Of course, to some people, this is unbelievable - because Content-Aware Fill makes a lot of the menial labor parts of digital manipulation disappear.

Second, there are those people who see what it can do and get the wrong impression. They say "Finally! I'm going to have so much more time!" This belies a basic misunderstanding of market economics. If it takes less time to do something, you have fewer billable hours. While it's possible to reduce labor required and keep your prices the same, you'll quickly be undercut by your competition. This applies doubly so for a profession that's less likely to have permanent contracts.

Third, there are those people who think that this will cost people their jobs. This is the other side of the "labor saving" coin. I've often heard the argument that the only result of new technology is a shifting of labor. Basically it goes like this; I invent the cotton gin, which decreases the work required to seperate cotton fibers from cotton seeds by a factor of fifty. This makes cotton cheaper, which means more people will buy cotton, which means that I need to hire more people. Additionally, cheap cotton boosts a number of other industries, such as clothing manufacture.

I often question whether this is actually true. History has shown that increases in technology mean that labor will shift to less and less "essential" tasks, as seen by the movement over time from agricultural to industry to service. It's somewhat difficult to find the data to compare occupations over time adjusted for population increases, so I have no idea whether there are (for example) fewer farmers today than there were a hundred years ago. It actually seems likely that while the number of farmers has decreased over time, the number of people employed in secondary agricultural occupations (fertilizer, herbicide, and pesticide production, genetic engineering, tractor manufacture) has increased. Again this is just a guess - if I were in grad school, and for something other than computer science, this is probably what I would study.

What would happen first though is that prices would fall, which means more people would be able and willing to pay for graphical work. This gives a bit of a cushion. Additionally, since retraining takes both time and money, a new technology will reduce wages before it cuts any actual jobs.

Finally, there are those people who say "GIMP has been able to do this for years with the ReSynth plug-in". This is (mostly) true. But for whatever reason (I've heard it's mostly the UI) most people who do image manipulation for a living use Photoshop, and for them, if a feature isn't in Photoshop, it doesn't exist. Personally, I get most excited about technologies when they're being researched at universities. Content-Aware Fill owes a lot to PatchMatch, which was developed at Princeton by people who also work for Adobe. The problem, of course, is that it takes seemingly forever from any interesting technology to get from "cool idea" to "workable reality".

No comments: