Revisiting ChatGPT…
22 minute read
November 1, 2024, 11:55 PM
Recently, a former colleague of mine shared a post on LinkedIn by Benjamin Stein (no, not that guy) that read, “Go ask ChatGPT: ‘Based on our previous interactions, what do you know about me that I may not know about myself?'” My colleague said in her post that she was using ChatGPT to help with a job search, and shared some of the things that the bot said about her. Not bad.
I responded to her post about my own experience with ChatGPT, saying, “All I know is that ChatGPT knows exactly who I am, probably because of my large online presence, and it told quite a few whoppers about me,” and then cited the Journal entry that I wrote about it last spring. My colleague responded that she loved the entry, but considering that the original entry is now more than 18 months old, as well as the rapid pace of advancement in this kind of technology, I should run it again to see what it comes up with. I liked the idea.
So I ran the entire process over again, asking the exact same questions, i.e. “What do you know about Ben Schumin?” and “Tell me about The Schumin Web,” and running each inquiry five times, using the default model, GPT-4o. My methodology for scoring each of the responses was exactly the same as before, counting the number of factual claims, and then determining the accuracy of each one. Accurate claims scored a point, inaccurate claims scored no points, and a mixture of accurate and inaccurate information scored half a point. Then take that number and divide it by the total number of factual claims made, and that’s the final score in the form of a percentage. As was the case before, I still don’t know what an expert in this sort of thing might do to rate the accuracy of these responses, but this is the best that I could come up with, plus this is the same method that I used before, giving us apples-to-apples results.
Categories: Artificial intelligence, Schumin Web meta, Social media
The future of the past…
19 minute read
June 27, 2024, 6:32 PM
Recall back in December that I spoke of the need to redesign this website in order to take advantage of various new features and functionalities that I’m not currently making good use of. Lately, unfortunately, I have not made any headway on that due to my being backlogged on new content. It’s like Jon Taffer of Bar Rescue fame once said, where an owner was too busy working in their business to be able to work on their business. However, I did recently take a look at screenshots and other materials for various concepts from past redesign efforts to serve as some level of inspiration, and while I didn’t feel particularly inspired by these old concept designs, I thought it might be interesting to share them with you, to think about what Schumin Web might have looked like had I gone further with these various ideas rather than what I ultimately opted to go with. I don’t regret not going with these various concepts because a lot of these were just explorations, but I definitely learned something about the site with each iteration.
First, recall that before the current “Modern Blue” design, which was introduced in 2012, I had been using a design that I had called “Faded Blue”, which was introduced in 2004 and was later modified into “Blue Squares” in 2008. One new thing back then was that the advertising banner, which had previously been at the bottom of the page, would now be at the top of the page. The first concept for that design was… not good. Here is the initial concept for the Journal:
Categories: Artificial intelligence, Schumin Web meta
Playing with the AI image generator…
22 minute read
October 27, 2023, 10:02 AM
Recently, a friend of mine posted some computer generated images from the Bing Image Creator, which uses the DALL-E system as its base. I enjoyed their posts, so I decided to take it for a spin myself with subjects that were more relevant to me. My first idea was to have it generate me. The way I saw it, ChatGPT kinda sorta knew who I was, so it seemed reasonable to see if Bing Image Creator could perform similarly.
The first prompt that I gave it was “Ben Schumin in Washington, DC” and this is what it produced:
Categories: Afton Mountain, Artificial intelligence, Baltimore, Fire alarms, Honda HR-V (2018), JMU, Kia Soul, Mercury Sable, Stuarts Draft, Today's Special, Toyota Previa, Washington DC, WMATA
Two stories about Woomy…
3 minute read
September 8, 2023, 11:33 PM
Recently, while we were driving home from shopping, Elyse asked ChatGPT to tell her a story about everyone’s favorite orange curmudgeon: Woomy. I don’t know what the exact parameters that she gave it were, but it came up with decent enough material to come up with something at least somewhat heartwarming. Check these out:
In the heart of Alexandria, Virginia, there lived a curmudgeonly orange octopus named Woomy. Every day, rain or shine, Woomy would climb aboard the DASH bus, his tentacles a tangle of frustration. He’d grumble and grouse about the noise, the crowds, and the whole human world around him.
Passengers would stare, amused by this unexpected commuter. Woomy’s sour expression became a familiar sight, and people began sharing stories of his antics. But despite the attention, Woomy remained steadfast in his curmudgeonly ways.
One winter’s day, as snow painted the city white, a young child with curious eyes got on the bus. Ignoring the stares, the child approached Woomy, offering a small, hand-knit scarf. Woomy’s eyes widened, his expression softening just a bit. He accepted the scarf with a tentative tentacle.
Categories: Alexandria, Artificial intelligence, DASH, Woomy
It both impresses me and amuses me…
28 minute read
March 10, 2023, 4:21 PM
You all have probably heard about the artificial intelligence tools that can write articles and such that have been taking the Internet by storm lately. One such service is ChatGPT, which is a chatbot by a company called OpenAI, which can answer your questions about various subjects. I asked the service about myself and about Schumin Web, because (A) my name is unique, and (B) Schumin Web is also unique, and (C) I’ve been around on the Internet long enough that I figure that it should know who I am. Additionally, giving it inquiries about myself and my website, I was able to do a good check of accuracy because I know me really well, and I know my own website really well.
So on March 1, I ran the inquiry five times for each, and collected five different responses for each. In evaluating what it spewed out for each one, I found that the accuracy was a bit questionable, and varied quite a bit. It got some things right, and it got some things very wrong to the point of being comical. In its discussion about Schumin Web, it was actually quite insightful, making points that even I hadn’t thought much about, doing way more than I would have otherwise expected from an AI chatbot. I was also a bit flattered, because in running other people who I feel should be far more notable than me, it didn’t know who they were, even with some additional prodding, while it knew who I was right out of the gate without any additional clarification or questioning, and it knew what Schumin Web was without even blinking.
In judging the accuracy of each output, I scored them by factual claims. A claim that was accurate got a point. A claim that was inaccurate got no points. A claim that was a mixture of accurate and inaccurate information got half a point. Divide by total number of claims to get an accuracy percentage, which would be the final score. I don’t know if experts in this sort of thing would score it this way, but it’s the best that I could come up with, and for purposes of this discussion, we’ll go with it. 
Continue reading…
Categories: Artificial intelligence, Myself, Schumin Web meta