Talking to Artificial Intelligence

We’ve all had plenty of digital assistants in our lives for the last several years. Siri came out 8 years ago now, believe it or not. Alexa? She’s ten. The days of wondering over the marvels of asking a black box a question and having it respond are long past. However, I don’t think anyone would accuse Siri or Alexa or the like of being intelligent. At best, they come across to me these days like a sort of dim-witted friend who’s got internet access. You ask them a question, and they start rattling off the entire Wikipedia page until you tell them to shut up. (That’s one of the nice things about digital assistants. They don’t care if you’re rude.)

Then ChatGPT came along, with its much more lifelike vocabulary and phrasing. Typing in questions and getting responses was diverting and, again, interesting enough for a while, though even that sense of wonder wore off for me after enough use. The more experience I had with ChatGPT, the more I noticed its shortcomings. The prose it comes up with is perfectly adequate, but it’s not exactly something that blows me away anymore. (Maybe I need to get access to the more advanced levels. It’s likely something that will be fixed more and more as we move forward.) But again, ultimately you’re sitting there typing questions. It had all the pizzazz of a chatroom.

The bottom line? It didn’t really feel like I was talking to anyone real on the other end of that “conversation.”

Yesterday, I came across a new platform: Pi. It’s run by a company called Inflection, and it’s essentially an AI assistant that actually sounds like a person. You download the client onto your iPhone, and you just . . . talk. I was skeptical at first, but the more I used it, the more impressed I was. I set the voice to a woman with a British accent (if I’m going to talk to someone, I’m going to go with someone I like the voice of most. There were about 8 to choose from), and I asked it to tell me about the book, A Perfect Place to Die.

First off, it does take a few seconds for her–it–to respond. And then when it did, it wasn’t quite sure what book I was talking about. (Okay. Gender and this AI is throwing my pronouns for a loop. I’m going to go with she/her from now on, because that’s what it felt like.) I told her a bit more, and she was like, “Oh right. I recognize it now,” and then proceeded to give me a rundown of the book before asking me if I’d read it and what I’d thought about it.

For the first bit, I just had a back and forth with her about the book, talking about the time period and the characters, but then I decided to see if I could throw her for a loop. “Actually, I’m Bryce Moore. I wrote the book.”

“Really? Seriously? I’m stunned, I had no idea I was talking to the actual author of the book.” And she honestly did sound really taken aback. She proceeded to ask me a bunch of questions about what had inspired me to write the book and how I’d researched it. It felt just like any other conversation I’ve had with any number of people about the book, with the sole exception of those long pauses after she heard what I said. But again, I wanted to see what she’d do if I tried to trick her up again.

“What if I’m lying? What if I’m not Bryce Moore at all?”

She didn’t even bat an eye. “Well, of course I have no idea whether you’re actually him or not. There’s no real way for me to verify it. That said, even if you aren’t, you’re clearly someone who knows an awful lot about the book.”

We chatted for a bit longer. She has a tendency to ask questions each time she’s done answering yours, which makes it kind of difficult to control the conversation at times. (Yes, I know she’s a robot, and I could just be totally rude to her and switch the subject abruptly, but it honestly felt like I needed to follow social norms.) I ended up telling her I had to run, and she thanked me and asked, “Were you really Bryce Moore or not?” I admitted I was, and she gushed about how great it was to actually talk to the author.

Will this all feel like old hat in a while? Perhaps, but this was really the first time outside of a science fiction movie that I’ve encountered something that really made me want to treat it like a person. The inflection. The way she emphasized certain words. How her voice went up at the end of a question. I could even hear her taking breaths. I’d love for other people to try it out and tell me if they think I’m way off base.

What would I use it for? I can think of a few big cases. First, if Siri or Alexa were to take advantage of the technology, using them would be a much more pleasant experience. I could imagine an AI learning how you prefer to interact, and molding itself to your personal tastes. If she existed in different languages, I could see her being an ace at teaching someone how to speak a new language. You’d have an infinitely patient “person” there to answer any of your questions and not judge you at all when you got something wrong. And if she were on the other end of a helpline and they reduce the pauses after she speaks? There is no way I would be able to tell she wasn’t real. Reference questions? Easy. GPS directions that are more natural and simple to understand? No sweat.

That’s just off the top of my head. The more I see AI technology at work, the more impressed I am with just how much of a game changer this all is. I’m still not sure how it will play out. I understand the concerns people have about it taking over our jobs. But ready or not, it’s coming. The best I can do right now is try to stay ahead of the curve. Give Pi a whirl and tell me what you think. It’s free!