In the early days of business computing, stupid computer tricks appeared frequently in the popular press … stories like the company that sent out dunning notices for customers who owed $0 on their accounts. (Resolution: customers mailed them checks for $0 to cover what they owed.)
Somewhere in most of these stories was an obligatory explanation, that computers weren’t really the culprits. Behind any mistake a computer made was a programmer who did something wrong to make the computer do it.
Years of bug fixes, better testing regimes, and cultural acclimatization pretty much dried up the supply of stories like these. But we’re about to experience a resurgence, the result of the increasing popularity of artificial intelligence.
This week’s missive covers two artificial-intelligence-themed tales of woe.
The first happened as I was driving to a regular destination from an unfamiliar direction. My GPS brought me close. Then it announced, “Your destination is on your right.”
Which it was, only to take advantage of that intelligence I’d have had to make a 90 degree turn that would have had me driving off the shoulder of the highway and up a steep grassy slope, at which point I could hope I’d have enough momentum to knock down the chain-link fence at the top.
Dumb GPS. Uh … oops. Dumb user, as it turned out, because I’d been too lazy to look up my client’s street address. Instead I’d entered a nearby intersection and forgotten that’s what I’d done. So AI lesson #1 is that even the smartest AI will have a hard time overcoming dumb human beings.
The more infuriating tale of AI woe leads to my making an exception to a long-standing KJR practice. Usually, I avoid naming companies guilty of whatever business infraction I’m critiquing, on the grounds that naming the perpetrator lets lots of other just-as-guilty perpetrators off the hook.
But I’m making an exception because really, how many global on-line booksellers that have authors pages as part of their web presence are there?
I was about to point a new client to my Amazon author’s page, as he’d expressed interest, when I noticed an unfamiliar title on my list of books published: The Feminist Lie by Bob Lewis.
If you’ve read much of anything I’ve written over the past 21 years you’d know, this isn’t a book I would have written. Among the many reasons, I figure men shouldn’t write books criticizing feminism, any more than feminists should write books that explain male motivations, Jews should write books critiquing Catholicism and vice versa, or Latvians should publish patronizing nastiness about Albanians.
Minnesotans about Iowans? Maybe.
But I distrust pretty much any critique of any tribe that’s written by someone who isn’t a member of that tribe and who feels aggrieved by that tribe.
But some other Bob Lewis proudly wrote a book with this title, and somehow I was being given credit for it. Well, “credit” isn’t the right word, but saying I was being given debit for it might be puzzling.
In any event, I don’t think all of us named “Bob Lewis” constitute a tribe, and I want no responsibility for the actions of all the other Bob Lewises who are making their way through the world.
And yet, somehow I was listed as the author of this little screed.
Oh, well. No problem. Amazon’s Author Central lets me add books I’ve written to my author page. Surely there’s a button to delete any I don’t want on the list.
Nope. Authors can add and they can edit, but they can’t delete.
Turns out, an author’s only recourse is to send a form-based email to the folks who run Author Central to request a deletion. A couple of tries and a week-and-a-half later, the offending title was finally removed from my list.
And, I got an answer to the question of how this happened in the first place. To quote Amazon’s explanation: “Books are added by the Artificial Intelligence system Amazon has in our catalog when the system determines it matches with the author name for the first time.”
Artificial what? Oh, right.
Which leads to one more prediction. Whereas as of this writing “artificial intelligence” has some actual, useful definitions, within two years the phrase will be about as meaningful as “cloud,” because any and all business applications will be described as AI, no matter how limited the logic.
And, as in this case, no matter how lacking in intelligence.
Bob, do they use the same AI to pay royalties?
Luckily enough, no. But thanks for asking.
My guess is that the person who responded to your report of an error in a listing had no clue how computers actually makes that happen. But instead of consulting with someone with a technical background on Amazon’s internal software, they tried to fake it, using the term “artificial intelligence” as a kind of verbal duct tape to explain to their customers all problems they, the customer service reps, don’t understand, but have to answer.
My own curmudgeon view is that I don’t think that I have yet seen true artificial intelligence in any general application. I’ve seen a lot of pattern matching, sometimes using fuzzy logic, and when used appropriately, this can be very valuable.
But, confusing pattern matching and discovery with business policies? One is a tool, the other a responsibility.
Wow! Quite a story. And quite a title to now show up when someone looks you up in Google, Goodreads, and/or Amazon.
I guess it would be interesting to see how you show up in the other author’s Author Central. This may be a difficult disconnect to make; in hind sight perhaps use of a middle initial, or formal first name, etc., would unique-ify your identity against those imposters.
Obviously, this use of “AI” flunks the Turing test, so it ain’t.
Your future prediction of “AI” being the be all/end all explanation is enlightening and probably (frighteningly) dead on. But then again, isn’t that what “automatic programming” was going to be, say 20 years ago?
What the Amazon rep called “artificial intelligence” gives AI a bad name because what they are doing is no intelligence — just matching names in a database.
I ran across this Amazon practice in another place last week. I was looking at the reviews of a chidren’s book titled Lightning. This set of reviews included ones for two fiction books with the same title — One by a major author, so there were lots of them. This made the reviews useless to me — “Greet book!” or “Lousy book” don’t mean much if you don’t know which book the author is referring to. In this case also it wouldn’t take AI for Amazon to sort this out, just checking the author’s name as well as the title. Or maybe even just actually leaving the review linked to the book that the reviewer wrote it for…
And again, I wanted reviews of an alarm clock. The reviews of all of the products of one alarm clock company were repeated under all of their products. The bad reviews of the ones that have a problem with automatic time setting made it impossible for me to tell of the one I was looking at has reliability issues.
The problem with artificial intelligence is that it’s 99% “artificial” and 1% “intelligence.” Just look at a certain major social media site that thinks it knows better than I do what is going to interest me in my feed!
I love reading the comments to your columns. In this case, Bob Harris uses a wonderful description: “artificial intelligence” as a kind of verbal duct tape to explain… I love it!!!
Then, as you point out, the use of “the cloud” to explain all sorts of things that the person doesn’t understand – more verbal duct tape.
When AI works, we won’t need to call it artificial.
Mr. Lewis, where’s the ManagementSpeak?
” I don’t trust you to put on your socks before you put on your shoes.”
As a father of four, I find this to be a valid concern.
When my daughters were young, I once made the mistake of telling them to put on their shoes and socks.
Which is exactly what they did, which was when I learned the meaning of “malicious obedience.”
Artificial Intelligence, while making great strides in learning from the information fed to it, would always suffer from being a poor imitation of Human Intelligence. Albeit a blistering fast poor imitation.
From my point of view, this is due to a couple of factors. AI is created and programmed by humans who are fallible. Secondly, most AI focuses only on the left side of brain function (logical) and not, if any, on the right side (arts/creative) – let alone a combination. Let us not even bring emotions and its influence into the equation.
How can we ever hope to fully understand intelligence and how our brains work by using the very same thing (our brains) to study it? Maybe AI should rather stand for Adequate Intelligence.
e
Well, maybe, but always is a very long time when human ingenuity is involved. My expectation is that AI will be different from human intelligence for a very long time. And AI-enhanced humanity is where things will be most interesting.