AI is a lot like pi
But I am not sure most people recognize that…
Full disclosure, each of the last few years someone has announced “hey, its pi day!” and I had to do some math. I think it is awesome that an irrational number and iconic element of mathematics and geometry now has its own day. But I am a very practical guy who learned this stuff when calculators were still contraband in the classroom (even my college adopted their use the year after I finished my math & chemistry electives). To me — pi is 22/7ths. Just roll with me on this.
On pi day, I found myself having coffee (no pie) with a young entrepreneur in the Artificial Intelligence space. I should note that he would say Machine Learning. He was eagerly spouting statistics that he found astounding. The one that came with the most energy was -
“More than 80% of companies in that artificial intelligence space have yet to produce working AI!”
I didn’t find that astounding, amazing, or even overly accurate. I suspect that a more hard core analysis would yield a more accurate number than 80+ and likely much higher! Given where I am about to take this article — more accuracy would have been both fitting and ironic. Coming back around.
Artificial Intelligence is a lot like pi — it is immensely useful, has a good deal of buzz, and has impractical levels of accuracy.
Let’s spin that around. AI is quickly spreading everywhere, at least in theory. Although practitioners like my entrepreneur friend are quick to note that much of AI either isn’t or is better labeled something else (i.e. Machine Learning). Put simply — chatbots, self-driving cars, and predictive algorithms don’t have nearly as much in common as their AI umbrella would lead you to believe.
Umbrellas are pi enabled technology… only no one thinks that (not even Mr 22/7ths here). Note — this one actually excludes the chatbots and the other far flung outliers. Kudos to Michael Watson at Opex Analytics.
Actually, much like pi, the umbrella of AI is really all about accuracy and optimization. AI can cycle faster and deliver far more accurate results than humans. That is also the whole point of knowing pi to a thousand places. Only, how important is that, really?
One day, some AI will calculate pi beyond its current 2,000,000,000,000,000th digit. Yes, that is right. Yahoo may have gone belly-up, but at least they improved our knowledge of pi along the way. Prior to that, we only knew pi to the 2.7 trillionth place. If this seems a touch ludicrous, it is probably more white and nerdy. All jokes aside… where is the value in this degree of accuracy? And one more time around the circle.
At the end of the day, when the Earth has used its own pi technology to complete yet another revolution, businesses find themselves in a very precarious place with AI. They are looking at a technology that keeps lauding additional accuracy and speed. Only…
Businesses were automating long before AI had real buzz. They were automating when accuracy was an issue — but computers long ago crushed that. Today, some jobs are beyond human ability — forget cost. They were automating when adaptation became the issue and AI has helped with that — but Industry 4 0 is built on more than just machine learning. It is only one element under that umbrella.
The problem then is two-fold. Some businesses are adopting AI because everyone else is — not knowing if additional optimization is even meaningful (actually, not even realizing that is the question!). Others have recognized the value of AI but are quickly learning that optimization was never their limiting factor. It all boils down to scale.
If you want to build nano-spheres or measure the circumference of an electron, I can see where knowing pi to a trillion places might help. Business tends to work on an opposite scale. If you are moving or monitoring billions of transactions a day or week or looking for one fraudster in ten million, Artificial Intelligence is pure gold. Otherwise, it may be more of a party trick… maybe even a bit irrational.
Final disclosure — I am a 22/7ths man. I have been labeled a contrarian. I have often noted that companies should hold off on their love affair with big data — until they can get small data right. To me — AI often feels like the technology no one is ready for, yet.
But I am also a student of history. Sometimes you need to stretch the boundaries of imagination and let others chase shiny objects, so the rest of us can focus on developing the disciplines and foundations that really matter. And history also says that one day, when the buzz is about quantum neural trans-dimensional intelligence — AI will probably have gotten there.
Thanks for reading!
