Working Up

by Dwayne Phillips

I realize that my plea won’t do any good, but could we please use the right terms in describing “artificial intelligence?”

Everybody is doing artificial intelligence (AI) these days. Really? Consider…

1. Artificial intelligence
2. Deep Learning
3. Machine Learning
4. Neural Networks
5. Pattern Recognition

The vast majority of what people call AI today is really #5—pattern recognition. People have studied and applied pattern recognition for decades. There is nothing new here. Feed a million photos into a system of equations, adjust the coefficients, and viola, you can recognize a flower and a cat. Of course it’s a bit more difficult than that, but really, that is what we are doing.

Tell me when deep learning can solve,

“It is May 2, 2017, what should and could I do?”

Answer: Mothers Day is coming (that’s the first answer). Now create something wonderful, i.e., consider your mother’s life, as you know it, and do something delightful.

That answers considers the entirety of your mother’s life—as you know it, the entirety of your life—as you remember it, and what you and your mother consider delightful. You can’t adjust the coefficients to create something delightful.

That delight is intelligent.

Of course this little blog post won’t change anything. The marketers will continue to holler AI-this and AI-that and all that wonderful stuff so please buy our processors and software APIs and our people (yes, we even sell people here). And the coefficients will adjust and we’ll expand the order of magnitude of the coefficients again and recognize patterns in super-duper-ultra-really-big-high definition video and allow a novice to back an 18 wheeler into a tight spot. Oh, wait, someone solved that problem decades ago, but everyone forgot to look it up (true statement, not fake news).

And please call me when an algorithm can create delight. I really do want to see that in my lifetime.