re: Memo to Apple: Siri's Charms Fading Fast
I just wanted to write a quick note to tell you how much I enjoyed your
article. I think you hit the nail on the head. I agree with your points, and your writing
style is highly entertaining as well. Good job.
Of course, I also think there was one minor misunderstanding in your
article. You said that:
"It's possible that Apple thinks that integrated natural language processing
isn't a core business. But I'm betting that it's going to be someone's very
profitable business, or at least a key weapon in the battle for smartphone
Well, in 1990 I got my doctorate in Educational Technology and my
dissertation was on artificial intelligence, knowledge engineering, expert
systems and natural language processing. For the past twenty years I've
been keeping up with that literature - almost as a hobby and habit - so I am
still up to date on the progress the industry has made.
Here's the dirty little secret that few will admit - the amount of progress
has been dismal. We are no closer to cracking the "natural language
processing" barrier now than we were in 1990. My professors at Lehigh
University in the now-defunct artificial intelligence lab explained (back
then) why we wouldn't ever be able to move forward, but because of the
complexity of the reason, many people don't understand. I've tried (in my
own book on Entrepreneur's Guide to Managing Information) to explain why,
but it is still difficult for people to grasp the concept. Basically, what seems to humans easy as pie is, to computers, absolutely impossible. For example, have you ever played
"Which of these things is not like the others?" - the famous Sesame Street game for toddlers?
A human toddler is great at it. A computer cannot do it. At all. Not even a little bit. Because a computer has no context. None. At all. Ever. A computer must still be explicitly told every single thing - it cannot generalize information from previous information.
The bottom line is that SIRI, and any (all) other natural language
processing programs are destined to fail. They always work "for a while",
for "simple" things, but eventually, they fail the Turing test. Eventually,
the holes in their thinking process (or lack of thinking process as the case
may be) appear. In the end, they will not be able to do what we need them
to do - no matter how much we think they should be able to, no matter how
much we wish they could. We are still many many years away from a computer
(no matter how powerful) that can understand natural language, or "think" in
a way that makes sense to human beings.
So if you think Siri will get better in a few years, I hate to disappoint you. It's not the connection, or the bandwidth, or the power or memory of the computer that's the problem. It's the fact that we don't - yet - have natural language processing that works often enough to get beyond the "beta" stage.
Just my opinion