Well now, it’s been five weeks since the last post. If you’re looking for volubility or frequent posting, this is not it.
Anyway, I’ve been reading and thinking about AI, and as I have been for a while, using my subscription to ChatGPT mostly to help with technical questions, which it can be useful for.
I’m seeing increased sharing and posting of AI-generated crap on Facebook. It’s almost instantly recognizable, and tends to have a vapid, not quite all there vibe to it. The images present as photographs, but are entirely too stiff and perfect. They don’t show wrinkles or photographic defects and artifacts, just bland, smooth lines and colors. And I’m seeing storylines of “celebrity does nice thing” over and over.
I rarely tell other people what to do, particularly with such low-importance things as sharing memes. So I’ve refrained from saying “This is AI bullshit”, I just scroll on by and make a mental note about the person’s gullibility. I also vow as usual not to do it myself. I’ve posted false things, but have corrected or deleted them when informed they’re not true.
The thing is, there’s no replacement for actual people speaking, writing and creating art. At least not yet, and I believe not for a very long time to come, if it happens at all.
One valid interpretation of AI, or LLM, or AGI, is that it’s not even close to being a person, it remains a mathematical equation, or rather a set of equations or algorithms. A computer program, in the long line of programs first envisioned by Alan Turing in the late 1930s.
It’s not smart to anthropomorphize an equation, even if it does a fairly good job faking a human presence.
My personal experience with ChatGPT, looking for help with technical questions:
- It’s not smart, it’s usually well-informed about things it can read, such as software manuals or hardware user guides. Well informed in the sense of “I’ve read all the manuals, what do you want to know that’s in the manual?” It still gets basic things, such as what’s in a menu, completely wrong.
- It can be stubbornly persistent in wrong answers, even when admonished.
- After admonishment, it continues to do the wrong thing, and denies it’s doing the wrong thing. A bit childish, actually. This is worse when I try graphical/image creation using ChatGPT. “Two branches!” Because I mentioned branches, AI adds more branches to the four already there.
- It will propose technical fixes, then when the fix creates further problems, propose further technical fixes to fix the problems it just created, and on and on. I’ve learned to nip this unproductive activity in the bud, but it persists in this simulation of an inadequate human technical support entity. It’s like a well-informed but not savvy technical support person. If I had endless time to fix fixes of fixes, I’d go along with it, but I don’t want fixes of fixes, I want one fix that gets me to the desired result. Evidently that’s too hard for my ChatGPT friend to understand. I’ve also done enough computer technical support to know when it’s going off the rails.
So, at this point I’ll continue using AI for a very narrow purpose, kind of like a more conversational search engine. I remain a skeptic about what it produces for me, but then that’s the way I was raised by my scientist professor father and intelligent, well read mother. Don’t just trust sources, gather as much information as you can, ask questions, come up with one or more working positions but don’t invest too much in any one of them. Science.