... because I have AI.
And it's ruining me.
One of my favorite scenes from The Simpsons, back when it was good, was where Lisa met Bleeding Gums Murphy, the saxophonist. He said that the reason he played was that the music was inside him and he just had to put the saxophone in his mouth and let it out. Writing and speaking is like that for many of us extroverts.
Once released, however, the need has been met and we can go about our lives. Well, until the next urge to communicate arises, that is.
When I write here, I only have an idea in my head. I don't have an essay composed, I just sit down and start typing. The words come out, formed in English composition, until I'm done. I never restructure my posts and I rarely edit them beyond grammar and spelling.
Once the blog post has been written, I move on to something else.
I've discovered that AI kills my blogging. Like singing in a soundproof room, it's fine for practicing, but it never gets shared with anyone. I've been using AI to thrash out ideas, but once I've done so, I feel like I do when I've finished a blog post. The fever is gone and I just don't have the energy to recreate it and post here.
AI has made me a hermit. It has also murdered my sense of style.
There are no elements of composition when you talk to it. It is simply Q and A. There is no wry commentary or made-up words that create a playful atmosphere. There are only brute ideas bandied about and discussed until a modicum of certainty has been achieved.
Style can't be nurtured in such an environment. It can only be stunted and warped until you become slothful and clumsy in your writing.
"If I stop practice for one day, I notice it in my playing; if I stop two days, my friends notice it; if I stop three days, the public notices it." - Pianist Hans von Bülow.
That's how I've felt using AI.
6 comments:
I've written a lot of powershell, and a lot of Jenkins' pipeline scripts, that I don't fully understand. But it gets the job done. The syntax of powershell, inside of groovy inside of pipeline, with double quotes, not single quotes, is maddening.
AI gets me across the finish line. I was stuck a couple weeks ago, and asked a coworker. He plugged the same error message into the paid chatgpt and got a different answer, that I got from the free version.
In the end, the problem was a program called insignia that exited with a -1 when it succeeded with the -ab switch verses the -ib switch. Wasted about a week (I'm part time, so that's 3 days) on that one.
"Style can't be nurtured in such an environment. It can only be stunted and warped until you become slothful and clumsy in your writing."
Yep, that about sums it up. I see people using AI, and it basically sucks out their initiative by giving a quick, easy hit of low-quality product, but in the long run knocks out their ability to write or draw anything decent.
The only people I talk to who are actually getting what they want from AI tend to be computer programmers, who are basically just using the AI as a next-higher-level programming language on top of the existing high-level languages. Everyone else seems to be getting trapped in the situation that Douglas Adams described for the robots produced by the Sirius Cybernetics Corporation: "it is very easy to be blinded to the essential uselessness of them by the sense of achievement you get from getting them to work at all."
MN, since AI draws its "intelligence" from years worth of text, the chances that it will be able to answer a recent technical problem is slim. For me, it keeps giving me PHP code that was deprecated over a year ago.
Tim, is it weed for the creative set? I've used it a ton over the last 18 months or so and it's been great in thrashing some things out, but I feel like I've reached the limits of what it can provide, particularly when it comes to fiction. The art bores me now because it all looks the same. That's not surprising because it's a computer not a creative type.
A bit similar to weed, but not quite. I understand that weed just makes it so you don't care if you accomplish anything or not. AI gives the illusion of accomplishing something with minimal human effort. But after using it for an extended periods, after the user has got out of the habit of doing things for themselves, it turns out that what it apparently "accomplished" either isn't trustworthy or isn't worth having.
Just a couple of days ago, I put together an example for the class I am teaching this fall to try to convince them to be wary of what AI gives them. I asked ChatGPT to draw me three things; a crystal structure for sodium chloride; a process flow diagram for ethanol distillation; and a schematic of a centrifugal pump. The crystal structure was OK but uninspired. The distillation diagram looked superficially OK, but on examination everything was either mislabeled, misplaced, missing critical items, and in a few cases mysterious products were being generated from nothing. The centrifugal pump that it drew was also superficially convincing, but as drawn it was totally non-functional for at least three different reasons (intake in wrong place, output both misplaced and at the wrong angle, and either the impeller was running backwards or the housing shape was inverted). When I used the same prompts as search terms in Google, they quickly give me what I wanted with no confusion. Given that the AI completely pooched simple jobs two times out of three, I am convinced that people who are depending on AI to do quality work are setting themselves up for catastrophe.
Keep on blogging!
Coincidentally, today:
WSJ: AI robs my students of the ability to think
https://www.wsj.com/opinion/ai-robs-my-students-of-the-ability-to-think-education-school-learning-880267c7
Post a Comment