DECEMBER 2, 2023 – In my search for new direction (see yesterday’s post), this morning I checked Bloomberg.com—my go-to news source—glanced at a few headlines, then clicked on “AI.” Whenever I see “AI,” I still mistake the “I” for an “l,” as in “Hal,” without the “H.” Hal, of course, was the computer-gone-rogue aboard the space station in the movie, 2001.
In any event, a number of headlines filled the “Hal” page on Bloomberg. At the top of the list: “Meta says There’s Been No Downside to Sharing AI Technology.” In this age of spin, I read the message between the words: There has been a downside, and Meta knows it. I lowered the cursor to the next headline: “Anduril Builds a Tiny, Reusable Fighter Jet that Blows up Drones.” I clicked and read the article. Yikes! Science fiction has definitely gone non-fiction—and yes, there is a downside to AI, shared, open, or closed.
But I wasn’t finished with reports about “Hal.” Next up: “The Year ChatGPT Changed Almost Everything.” The article was a rehash of information I’d already read or heard about elsewhere.
We’ve all heard about ChatGPT, friends of mine have had face-to-face encounters with it. Their descriptions of ChatGPT’s extraordinary literary prowess and infinite access to knowledge are mind-bending. But until this morning, I’d never experimented with this powerful form of “Hal.” I decided that now was the time—given my need for a map and compass for this blog.
My query: “What’s a good new subject for a post on the blog site writemakesmight.net?” My expectations were high. Surely this version of “Hal” would cull through my posts to date, a dozen years’ worth of Facebook posts, and every accessible online source bearing my name and everything and everyone having an algorithmic relationship to anything and anyone associated with me. I imagined that in 20 seconds flat, “Hal” would spit out a substantively riveting and perfectly crafted essay for me to “copy and post” straight to my blog site.
In the event, I got garbage; a generic response designed for rank beginners looking to post their very first blog essay about . . . nothing except their desire to make money by posting blather on their homespun blogsite. Specifically, my query elicited the following:
Choosing a good subject for a blog post depends on the target audience and the overall theme of the blog. However, considering the general theme of the blog site “writemakesmight.net,” which suggests a focus on the power of writing, creativity, and perhaps empowerment through words, here are a few suggestions: 1. “The Impact of Words: How Writing Can Shape Your Reality”; 2. “Unleashing Creativity: Writing Prompts for Every Genre”; 3. “Crafting Compelling Characters: A Guide for Writers; 4. The Art of Storytelling: Techniques to Captivate Your Audience.
And so on. I scoffed out loud. “Sorry, Hal,” I said. “You failed. I’ll figure out my blog posts . . . on my own . . . just as I have from the outset. Meanwhile, now that you’re cranking up production of those ‘tiny reusable fighter jets,’ do the world a favor and keep them out the hands of the really bad guys.”
Subscribe to this blog and receive notifications of new posts by email.
© 2023 by Eric Nilsson
1 Comment
The claim made in the media is false:
“…ChatGPT’s extraordinary literary prowess and infinite access to knowledge…”
Of course, the source is non-trivially large, but not close to infinitude.
But more important, AI specifically *excludes* peer-reviewed scientific papers that are behind a paywall (almost all the most carefully vetted) and has *no* code that can wisely assess the validity of claimed results. (Often the peer reviewers themselves debate results’ validity.)
Furthermore, AI makes no attempt to sort false from true in the vast manure pile that is the Internet. Its only technique is that all true information is essentially the same, while fiction is randomly wrong (except for propaganda), hoping that correlative statements are more reliable.
My own experience is that AI essays are at the sophomore level and shallow, and sometimes contain errors.
The quality of AI responses is strongly dependent on the quality of the questions it is given.