I've been using ChatGPT and its friends to help with homeschooling. It can be useful for basic regurgitation of information (that you then have to fact check of course).
But AI developers are already running up against the Ourobouros problem, where there's not enough data to train off and there's no way to distinguish human content from AI content on the web, so AI is going to be training on its own output, which will just reinforce its biases and create crazier and crazier hallucinations. That's bound to lead a large moment of disillusionment.
I'm more hopeful on the benefits for specific, narrow application of AI trained on more concrete data, for example in developing new materials or engineering structures.
There is a place for it. AI and even ChatGPT are a lot bigger than what I'm talking about here. I work with a company that develops AI as a manufacturing and engineering tool, and it does well in certain applications. Chatbots have their place.
The problem is, it really can't replace a human where a human connection is needed. Unfortunately, people are trying to take that shortcut. If you can't write well beyond the basics or need only formulaic content, sure. It'll help and save you a few minutes.
But I'm so sick and tired of the enthusiasm that AI could "human" better than humans can.
(Unless you’re in a charming old European building, in which case you will be tempted by a pull cord with a lovely brass knob dangling at the end, and will briefly consider the possibility of hanging yourself with it. At least you’d have a lovely view.)
Just read this post again and love it even more. And still laughed out loud at this paragraph.
This Drake AI post is really good. Though I'm going to have to ask you to send me licensing fees for the data you scraped off me to train it.
Heehee! :) Drake AI it is.
I hear you , though.
I've been using ChatGPT and its friends to help with homeschooling. It can be useful for basic regurgitation of information (that you then have to fact check of course).
But AI developers are already running up against the Ourobouros problem, where there's not enough data to train off and there's no way to distinguish human content from AI content on the web, so AI is going to be training on its own output, which will just reinforce its biases and create crazier and crazier hallucinations. That's bound to lead a large moment of disillusionment.
I'm more hopeful on the benefits for specific, narrow application of AI trained on more concrete data, for example in developing new materials or engineering structures.
There is a place for it. AI and even ChatGPT are a lot bigger than what I'm talking about here. I work with a company that develops AI as a manufacturing and engineering tool, and it does well in certain applications. Chatbots have their place.
The problem is, it really can't replace a human where a human connection is needed. Unfortunately, people are trying to take that shortcut. If you can't write well beyond the basics or need only formulaic content, sure. It'll help and save you a few minutes.
But I'm so sick and tired of the enthusiasm that AI could "human" better than humans can.
Time and time again, while reading this, I burst out laughing with tears running down my cheeks. Bravo for this fabulous creative production!
(Unless you’re in a charming old European building, in which case you will be tempted by a pull cord with a lovely brass knob dangling at the end, and will briefly consider the possibility of hanging yourself with it. At least you’d have a lovely view.)
Just read this post again and love it even more. And still laughed out loud at this paragraph.
Interesting, thought provoking writing.
I've just read this a second time. It's actually brilliant. And so much fun.
You are my biggest fan, mom ;)