Headlines, subject lines, and business publications all say that AI is embedded in every product or service, and will be found in more as time goes on (yes, these statements are contradictory). Prove it!
I’m going on a quest to see where I can find AI enabled, embedded or otherwise utilized in products. At least, this is how this post started. But, I really like making fun of AI, so that’s how it ended. Observing the goofy things AI does.
There’s probably a psychologist that would denounce my preoccupation with the insufficiences of AI as some sort of displaced feeling of powerless and kicking the robot-dog. Fine. Except, I would never kick a dog. Or other creature. I have problems pruning plants or pulling up weeds, because I don’t want to hurt living things.
Anyways, here are a few recent AI experiences:
Call centre attendants. Had a nice chat with an AI call answering thing when I needed a tow for my dysfunctional car. The AI was pleasant enough to talk to and understood me better than these sorts of things have done in the past. However, it failed when it came to recognizing my desired destination. I gave the name of the business. Google would have known where it was by the name. A human would have been able to figure out where I wanted to go. The AI solution was to send me a map, which required turning on location services. We both already knew where I was, so location services wouldn’t have told it anything more. It needed to know where I wanted to go. AI fail. Hint for AI: I don’t want my car towed 8,000 km. You can look for the desired destination within 10 km of where I am.
Shopping algorithms. I’ve found book buying on line to be much better than in person, which I ascribed to the algorithm’s ability to make suggestions based on my preferences rather than the in-person criteria of liking the cover. However, today’s shopping experience was different. After selecting a book with the world ‘Turtle’ in it, which was about the Indigenous experience in colonial policy, AI suggested children’s books about turtles. Major [political] AI fail.
Autofill. My phone made an event in my calendar, based on a voicemail I’d just listened to. Too many platforms/modalities to connect to be anything other than AI. I wish it asked first if that was what I wanted to do.
Information source fail. I asked Google maps to find me a transit route to an event, even though I had my own ideas about the quickest one. This wasn’t a simple query because it transversed several transit systems and 75 km. Google’s answer was funny, because it was so inefficient. My route took 1.5 hours, one transfer and was cheaper. Google maps included multiple transfers, 2.5 hrs., and the passenger train system that covers all of Southern Ontario (Via) rather than the commuter train for the area (Go). This is kinda like telling someone to rent a car instead of taking a taxi/Uber to a nearby restaurant in a large NA city.
And just plain overall AI fail. While doing some research online about Netflix, the top match in my Google search was an article that sounded authoritative, at first. As the article progressed, there were lines interspersed about random information about autonomous vehicles. Then back to media streaming, and autonomous vehicles, then Netflix, then autonomous vehicles. Epic nonsense.
Enough examples. How about a few generalizations?
Generative AI could take the place of information gathering. Just when Wikipedia has matured to a credible business model, along comes AI, which writes prettily. Being on the receiving end of the information gatherers/presenters, my main comment is “people will believe anything if it sounds authoritative”. AI does well, but not necessarily by doing good.
Thinking through a day in my life, taking a leisurely Saturday as an example: I paid my bills online, ripped up the floor covering in the hall, planted seeds for the summer vegetable garden, did a little online shopping, watched a few YouTube videos and made a home cooked meal. I don’t see much current or future need for AI involvement in any of these activities.
A tiny read about how banking will involve AI suggests improvements in the usual services, such as answering customer questions, making customers aware of products and fraud detection. One question: why when I pay a bill on Saturday, it can’t be processed until Monday. Is there really a human we have to wait for to come to work on Monday? Maybe AI could be trusted to take $62 out of my account and give it to my cell phone carrier. Also, the fraud detection I’ve experienced so far is fraudulent – as in: texts and voicemails claiming my accounts have been breached and I need to give them all my information and first born to rectify the problem. An AI filter on my in-coming communications to better remove the phishy-wishes would be great.
I’m not entirely negative on the potential for AI. There are useful things imaginable for the technology. .
And there are some great advances I’ve happened across, for example:
- an AI-based practice interview thing – sound like a great tool for those relatively new to the job market to develop some confidence and skills
- AI’s to read X rays, pathology slides, credit card transaction patterns, and other things where a certain amount of knowledge is required but the volume is mind-numbing, which causes humans to error due to attention drift.
Now and in the future, I suspect AI will make many things better for a lot of people, and some worse, hopefully for fewer people. The silliness is a sign of an industry in an introductory phase. The bumps and sidetracks into products that don’t work or no one wants are milestones on the way to finding really useful applications. After all, early social media had MySpace, early kitchen appliances included electric carving knives, and early cars were started by cranking a lever by the front bumper.