Good news, we have successfully taught the AIs to play Mad Libs
Artwork by neural nets:
“New hot meme for webcomic artists: upload one of your comic pages to Have I Been Trained, link to the first 4 webcomics (not drawn by you) that AI registers as similar to yours!” (Can I make this a thing? I’m gonna try.)
“I read 19 comics with AI-generated artwork, all of which I downloaded or read for free online. In addition to the question of quality, I wanted to answer a few other questions: Do these comics share common characteristics? How about the creators? Are traditional comic creators getting something valuable from these tools?” A cool, thoughtful article about what people can produce using AI as a tool (and what limits they have to work around).
“Here is where it gets interesting. If I use the same prompt and add “Amazing awesome and epic”, the picture gets noticeably better. “Oh,” goes the neural net, “you wanted a GOOD picture”.” Rebloggable-on-Tumblr version of that AI art post from July 2021.
Writing by neural nets:
“I pulled a letter from the Savage Love inbox—something, low, slow, and over-the-plate—went to the ChatGPT website (www.openai.com), and asked ChatGPT to “answer this question in the style of Dan Savage’s advice column.” So, can the ChatGPT artificial intelligence chatbot really do a better job giving sex advice than I do? We’re about to find out.” [archive link]
My favorite part of that one: the AI gives basically the same answer to both questions. It registers “this is a relationship help question,” and then it outputs Generic Relationship Advice — with a few brief details of the specific question pasted in, Mad Libs style — for both.
“A close examination of the work produced by CNET‘s AI makes it seem less like a sophisticated text generator and more like an automated plagiarism machine, casually pumping out pilfered work that would get a human journalist fired.” Computers are great at taking an existing sentence and automatically replacing some of the words with synonyms. But you need a human to come up with the sentence first.
These AIs are trained with text auto-scraped from all over the internet…so, how can you keep it from presenting unsuspecting users with the find-and-replaced version of, oh let’s say, somebody’s graphic noncon Joker/Robin fanfic? Pay a bunch of humans in Kenya less than $2/hour to manually flag all the noncon Joker/Robin fanfic, obviously.
All these tech people boasting they’ve developed an AI that can replace human creatives (or are on the verge of developing one, please give them millions of dollars to finish the job) — dig a little deeper and you’ll find they’re just reinventing the Mechanical Turk with extra layers.
no subject
no subject
Honestly, as someone who's worked a lot of boring mindless jobs where at least part of the work could probably be automated...when I think about how you'd do it, "organizing all this info in a way the computer can understand, and then programming a system to process it" would still be a complicated effort that requires a lot of human attention upfront! There's just no path that's really as cheap or easy as people want it to be =(