A direct link to John Markoff’s NYT piece here…
It seems like deja vu all over again to me.
I remember in 2000 as the bubble was starting to burst it seemed everyone had a “semantic search” or “contextual keywording” solution in the works. Brilliant “artificial intelligence” applications that were going to change the way the web worked, that would do it better, faster and more accurately than any lowly subeditor. There were the truly amazing feats of powerpoint and the equally evocative product evangelists pitching truly smart apps and 3d semantic search modelers, that now seem more like early tag clouds than anything else. Companies like Autonomy, Metatagger and a slew of others I’ve forgotten offered promises that would surely transform the net with their utterly brilliant algorithms written by their CEO, who they always mentioned in hushed tones *might* be the smartest person they’ve ever met.
They all seemed to gravitate toward towards discussions of how search and keywording had trouble contextually dealing with terms like “apple and washington” – does the user really want a computer or a fruit? Adding the word Macintosh to the fray doesn’t help much either. Great discussions, that one could noodle on forever, and easily repeated in meetings later to impress the boss, as the concept of contextualization was, er, neat.
The problem was the computers just weren’t that smart. The scheme always relied on developing a huge overarching taxonomy for the way you do business, then devoting resources (as in those subeditors they didn’t think so much of before) to “train the system.” The rub is they weren’t really teaching the system as much as trying to catch obvious errors. Like when it failed to identify a fruit vs. computer, every single stinking time.
So now we’re back to the point where our smart@ss computers are going to make things happen better, faster and more accurately than our subeditors can. The problem I see is this: while my computer today is much faster than the one I ran in 2000, I don’t see it as particularly smarter. In fact, it’s no better at discerning canned spiced ham product in my in box than it’s predeccessors were. So these are the machines that are going to fix things for us?