AI: Artificial. Yes. Intelligent? – Insights by Guy Higgins
I recently read an article from the New York Times, Artificial Intelligence Hits the Barrier of Meaning. The article, by a computer scientist with decades of experience with artificial intelligence (AI), highlights the problems with creating artificial intelligence code that can actually understand. She points out that her smart phone French-translation-AI app translated, “I put the pig into the pen” into a French phrase that placed said porker into a Mont Blanc-type pen. The “AI” didn’t understand – didn’t grasp the difference in meaning between a writing implement and a confined area. Almost immediately after reading this article, I came across a 20-minute video, Out of our minds. This video is a self-described as a “…comparative cognition road trip across the US in search of a map of the mind.” I found that description to be a tad bit self-aggrandizing since the video was centered on psychologists bemoaning their collective inability to understand human intelligence and whether or not what they’re doing is actual science or floundering around in some kind of neural alchemy.
What struck me though, was that one of the psychologists observed that, “…we don’t even know what intelligence is!” (emphasis added).
If we don’t know what intelligence is, how can we create an artificial one? The developers of AI algorithms seem to be focusing on a set of resources to create computer programs that can “behave” in a manner analogous to intelligent behavior. Those resources seem, to me, to be:
- The use of computer “learning” – code that is designed to do it’s job better and better as the code performs in more and different circumstances and takes logic paths that result in either better or worse performance. The code is design to increase the likelihood that it will use successful logic paths and decrease the likelihood that it will use unsuccessful logic paths in the future. Basically, it learns not to touch a hot stove, but unlike Mark Twain’s cat, it also learns that it can touch a cold stove because the algorithm doesn’t zero out the likelihood of touching any stove just because it got burned on one hot one.
- The use of program approaches such a neural nets that support computer learning without any initial subject matter code. They permit learning from a zero baseline.
- The use of the massive amounts of data that are becoming available via the aggressive collection of data enabled by cheap sensors and the Internet. The idea here is that there must be ponies somewhere in this gargantuan pile of horse manure.
As I think about it, it seems to me that there are some significant problems that have to be overcome before we come to rely on AI as some kind of panacea:
- The problem of contextual understanding – It is impossible to create code that can foresee every conceivable and inconceivable situation. Therefore, any true AI must be able to understand circumstances in context and respond appropriately in context. This seems to be a problem when it comes to stuffing said 600-pound porker into that Mont Blanc.
- The problem of scope – A useful AI needs to be capable of responding across a wide range of circumstances and endeavors. I’ll boldly assert that an algorithm that can play Go better than any human being but can’t recognize a raptor against a reddish morning sky isn’t intelligent.
None of this means that humans will never develop computer code that is intelligent, and it certainly doesn’t mean that we don’t already have a great deal of code that performs tasks for us that we once deemed doable only by humans. What it does mean is that, as leaders, we need to recognize that a great deal of the hype surrounding AI is still just hype and that we need to understand that AI, like any tool, works best when used for the tasks for which it was designed.
Guy Higgins, Firestorm Principal – Firestorm Colorado
Guy Higgins and his business partner, Jennifer Freedman are Firestorm Principals. committed to the importance of planned resilience in the face of disruption or crisis.
With more than sixty years of combined business experience, they are committed to the importance of planned resilience in the face of disruption or crisis. They assist companies and organizations plan for and achieve breakthrough performance, even in the face of disruptions and disasters that can severely cripple their ability to do business or put them out of business entirely. Learn more about Guy and Jennifer.
Interested in working with Firestorm? If you are looking for a professional consulting and training business opportunity that will put your experience and contacts to work, check out our Franchising opportunities. Firestorm can show you how preparedness can create opportunities for you and help others.
Additional Insights by Guy Higgins:
There’s a time and a place for checklists – Insights by Guy Higgins
Small Failures Result in Big Crises – Insights by Guy Higgins
When Everything is Going Well – The Duty of a Leader by Guy Higgins
How Corporate Culture Contributes to Corporate Success
Debiasing – How Leaders Can Avoid Biased Decision-Making
“What should we do now?” “What should we say?”