Artificial Sweetener: ChatGPT and Me

"Daddy!"

The cry was plaintive, and clearly directed my way. Ugh! I was relaxing in a comfortably furnished waiting room, enjoying some cinnamon coffee cake, and I was prepared to completely ignore the unnecessary outburst.

"Daddy!"

It happened again. It wasn't a real voice, and it wasn't really saying "Daddy!", but that is how I heard it. I knew who was crying. It was my stupid car, wailing at me via text messages during its first warranty check-up at the dealership. What a wimp!

"Daddy!"

Incredibly, I began to feel guilty about ignoring my car. I put down my pastry and capitulated. I opened the iPhone's Messages app and listened to my car's whining.

"Daddy! All my doors are open!"

"Daddy! Someone lifted my hood!"

"Daddy! my tires are being rotated."

In truth, the messages were a tad more succinct and mildly informative, but I am not used to being talked to by a motor vehicle. I like my cars quiet.

Is this the new artificial intelligence that I keep hearing about? I must be getting old.

Of course, when I was younger, I was thrilled when I heard my first talking calculator back in 1977. It could say numbers, arithmetic operators and the word "Error." It wasn't very conversant, but it could add and subtract. That was great!

Soon enough, the age of microcomputing snuck up on us. Both PCs and Macs began to speak. So did my book readers. These days, my Home Pod talks, as does my iPhone and watch. And, now my silly car wants to interact, too!

Recently, a close computer scientist friend of mine called to suggest that I download the ChatGPT app to my iPhone. David told me that the program was very interesting and a good example of where AI had progressed. I did as directed and discovered a whole new world.

ChatGPT came close to passing the Turing Test. Impressive. The Turing Test is an old goal that suggested that when the day arrived when you could not tell that you were communicating with a machine, rather than a human, that you had reached an important milestone in artificial intelligence. And ChatGPT was getting really close.

I started out with simple questions to ChatGPT. I asked it to tell me the distances to planets in our solar system. I inquired about characteristics of gravitational waves. I requested that ChatGPT write a Shakespearean sonnet about quantum entanglement. (It did a really good job.) And, I asked ChatGPT to write a short story about a turtle falling in love with an Apple Watch. Although the premise was admittedly nutty, the resulting story was rather touching. I never knew turtles could fall in love with technology. I wonder if the relationship would last if it met the Series 9...

I have asked ChatGPT many hundreds of questions. And, although it was often right, it could also occasionally lie to me, quite content to make up facts and present them as truths.

I tested ChatGPT. I thought it would be fun to ask it about the creators of "Halls of Time," the first high-speed 3D graphics game for the old monochromatic TRS-80 computers. I knew the answer and ChatGPT also claimed that it knew the correct response. It named a couple of people that had nothing to do with that old game and I told ChatGPT that it was wrong. It apologized and offered up another couple of names. I told ChatGPT that it was in error again and it apologized again. This happened several more times and I finally set it right. I told ChatGPT that "Halls of Time" was a program that Paul Rutz and I wrote back in 1982. ChatGPT replied that it was an honor to meet me. That seemed rather disingenuous. You see, ChatGPT starts with a clean slate of memories with each new set of interactions. ChatGPT suffers from amnesia between visits. ChatGPT always starts with a fixed dataset, last updated in 2021. It really doesn't know me at all from chat to chat. And, it will inadvertently lie, without shame.

The same confusion happened when I asked about the authors of "a particular HyperCard university textbook from the early 1990s. ChatGPT confidently told me about the co-authors. It was happy to share what it thought it knew about the book. And, it was wrong again. Several times. I finally told ChatGPT that George Culp and I wrote that book a long time ago and ChatGPT immediately gave us credit, and then forgot us again. ChatGPT continues to bust my chops.

However, despite its flaws, ChatGPT is a joy, and it can be a really good educational tool. This week, I decided that I wanted to learn about regenerative braking on a hybrid vehicle. My new, rather talkative car replenishes some of its battery power from regenerative braking and I didn't know anything about how that really worked. At first, ChatGPT seemed to assume I had an advanced physics degree. I had no idea of what it was relating to me. I told ChatGPT that my academic background was in the computer sciences and it purposefully dumbed down the explanation a bit. I still didn't quite understand what it was saying, so I asked it to pretend that I was in high school, was not a particularly good student, and mostly hovering with a "C" average in my classes. I did ask that ChatGPT share how regenerative braking worked in a way that was both completely accurate and would likely impress my imaginary high school buddies. And, you know, ChatGPT did a really good job. I think.

I pay the optional monthly fee for ChatGPT as I use it nearly every day to learn something new. The trick to making it useful is to see how it answers your queries and then check with other good and trusted sources to make sure that ChatGPT is not hallucinating its answers. I do the same fact checks when reading newspapers or listening to cable news. And, we should all do the same when perusing social media. Verification is a good thing.

ChatGPT is not always right, but it often is, and it does a remarkably good job of sharing its knowledge, or lack thereof, at any level you request. Very cool.

I'm afraid that I have become a bit of a convert. Yes, I can easily understand why we need to monitor and manage how we use AI in our future, but it is here, to stay and we should learn how to responsibly incorporate it into our lives. And, frankly, it is useful and fun. It's sweet!

I should have known that all of this artificial intelligence hooplah would come to pass when my car first started sending me text messages. I just need to embrace this new age. I think I should find a way to attach an audible "Daddy!" to future messages that arrive from my car. That seems like the intelligent thing to do -- artificial as that might seem.

---

This is my 31st blog for AppleVis. I love this community!

Please do leave your comments below. I would love to hear of your own feelings and experiences with artificial intelligence. And, feel free to write for any reason. I do enjoy hearing from you.

By the way, I did not let ChatGPT take a peek at this blog. I don't want it trying to edit my work. I'm sensitive that way.

You can find three of my older blogs at the following links: "I'm Being Nice, So Fix It!", "I Felt Powerless, Again", and "Say What? : Hearing Aids, iPhones and My Apple Watch".