@discobot, your thoughts?
Hi! To find out what I can do, say @discobot display help.
That wasn’t helpful at all.
There are many similarities between ChatGPT and Discobot. Clearly the idea of saying something false has unsettled him.
Japan has ruled that AI training is officially cromulent.
AI is taking over… AI training.
Our jobs might be safe for another couple of millenia.
I remember being confused as a teen that the cover of my edition of CJ Cherryh’s Fires of Azeroth seemed to have absolutely nothing to do with the book I was reading. This was the cover:
There was mix-up at the printer. Imagine how confused buyers for How To Pick Up Bikini Sword Chicks When You Are A Giant Bat must have been.
The general public has not yet adjusted to the concept of “this image might not be real” and thus are ripe for conning. Look at all gullible idiots falling for these two images from my recent Facebook feed:
If AI has access to online tests and test prep, these bots performing well on online tests are less impressive than they might first seem. Possibly.
AI was used for the opening credits of Secret Invasion. Cue people complaining that AI is takin’ their jerbs.
(I watched the first episode. It is possible ChatGPT wrote the script, too.)
I must have read twenty articles by reporters who deliberately used chatbots to write the column or spot, and said so. While none of them were very good, none of them were notably poor. When you read a bunch of them from professionals wondering about there future, it does not seem like the best strategy somehow.
I’m a teacher and I use ChatGPT to denerate material for my class. What’s really nice is that I can have it write something in the appropriate level for the ESL students.
Here, it’s asked to describe itself in 50 words.
Here I asked it: “Rewrite this for ESL students at a 4th grade level”
It takes further editing from there, but it really makes my job much easier. I’ve also used it to generate material for classes.
For now, they still need a human teacher, but a lot of teaching foreign languages can be done by AI if done right.
When I’ve tested it on Japanese questions, it gives a lot more misinformation. I presume it doesn’t have as much information to rely on.
Lawyers got a $5,000 fine.
I expected more than that. And not over the use of ChatGPT, but over the subsequent cover-up and lying about it.
It can be used to browse databases, so it could be a valuable research tool, especially if it is actually trained on all the case law. But it would still be the lawyer’s responsibility to actually look up the citations and make sure they are applicable (or even exist.)
Is ChatGPT going all Flowers for Algernon?
If ChatGPT gets better too quickly, maybe it’s going Love Story?
Engineer 1: We need to use the entire Internet to train our models.
Engineer 2: Uhhhh, maybe we should exclude these well known foul-mouthed troll farms…
Engineer 1: Good idea.
(Later.)
Engineer 2: We need more data. Guess we’ll have to include… The Straight Dope… 
The lawsuit against the generative art organizations is off to a rocky start