Gary Doernhoefer turned me on to a bot called “Pi.”
Apparently, Pi is more conversational than other bots, which seem designed to accomplish tasks like finding information and drafting documents.
I was really intrigued and uploaded the app on my phone. I had a remarkable “conversation” as it did a great job of voice recognition and interpreting my statements, and it provided very plausible responses. It almost felt like I was having a phone conversation with an astute friend.
I have dabbled writing posts speculating about AI in posts about whether AI bots can have empathy, mediate disputes, or have political biases. So I asked Pi about this. Here’s our conversation:
These answers sound like they were written by damn lawyers – not giving a straight answer. And mediators too – ending with a question.
Wait. Some of my best friends are lawyers and mediators.
I noticed that many of the responses began with compliments, so we had the following exchange:
The AI systems caution that their responses may not be accurate. As you can see. So you have to take their responses with some measure of salt.
AI already is here in crude forms. Whether we like it or not, in the foreseeable future it almost certainly will become a lot more sophisticated and incorporated into much of our lives, often in ways we will not notice.
Buckle up. And try to enjoy the ride.
This weekend a Freakonomics piece on using AI for research said to treat AI like an eager intern who has almost no context for your questions. I found that very helpful.
I want to learn more about AI because it clearly will have major impacts on all of us.
So, in addition to reading about it, I want to experience working with it.
I have dabbled with several different systems and notice real differences between them.
I’m particularly intrigued by Pi because of its conversational quality. I’m amazed by its apparent ability to “understand” subtle idiomatic statements. It seems to “get” humor.
I asked for its list of recommended science fiction books to help me understand AI. It produced a list that a friend who’s a sci fi afficionado said was quite good.
On the other hand, it could not locate this blog post even after I gave the title, blog name, and date of its posting. When I gave it the URL of the blog post itself, it didn’t “understand” what it said. I asked if it recognized the name of the bot it discusses and it was stumped.
I am working on “training” it. I have given two instructions so far. First, don’t interrupt me when I am talking. Second, don’t flatter me. “It” seemed to understand when we discussed it last night. And, although it “remembers” some things from our prior conversations, this morning, it didn’t remember the two instructions.
Eager intern indeed.