I agree with Martin on this. Saying a computer can recognize some form of communication doesn’t make me believe that a computer will soon be able to initiate that form of communication.
Terry is incrementally rightish about computers "designing." More and more tasks that had been assumed to be part of the task of designing have been relegated to machines; that will, no doubt, continue. Does that turn into some Zeno-ish paradox where designers keep getting halfway replaced?
The book cover example strikes me as a counter-intuitive place to start demonstrating that a computer might take over the job of a graphic designer. If you’re talking about the relatively straightforward arranging of type automatically in a space, why not? But if you’re talking about the sorts of covers of the quality that Martin does and writes about, it seems unlikely that software will do the task start to finish. (Newspaper layout seems like low hanging fruit if newspapers survive long enough.)
The article strikes me as an example of the problem of declaring statistical significance = actual significance. Remember that the question in the study is whether the software can merely recognize the general subject matter of a book. In the first place, the computer’s task seems like a real "So what?" Recognizing that the stuff in my mouth is dessert doesn’t make me likely to become a pastry chef.
But let’s assume that the task of recognition meant the task of creation were doable. Imagine if you hired me to design your book cover and I declared my solution to be sufficient because someone can tell what sort of book it is. Would you pay my bill?
Now imagine I said my solution was sufficient because *someone* can tell what sort of book it is but a large majority of people won’t even be right if given two more guesses after failure. The article tells us that "[t]he algorithm listed the correct genre in its top 3 choices over 40 percent of the time and found the exact genre more than 20 percent of the time. That’s significantly better than chance."
Yes. That’s significantly better than chance. It’s also useless if recognizing categories is the sole measure of success for cover designs. I suspect, however, that there would be widespread agreement that it’s not the sole measure.
Gunnar
Gunnar Swanson
East Carolina University
graphic design program
http://www.ecu.edu/cs-cfac/soad/graphic/index.cfm
[log in to unmask]
Gunnar Swanson Design Office
1901 East 6th Street
Greenville NC 27858
USA
http://www.gunnarswanson.com
[log in to unmask]
+1 252 258-7006
-----------------------------------------------------------------
PhD-Design mailing list <[log in to unmask]>
Discussion of PhD studies and related research in Design
Subscribe or Unsubscribe at https://www.jiscmail.ac.uk/phd-design
-----------------------------------------------------------------
|