Nobody cares about "AI" (Chatbot: I disagree.)

Pancake

Fluffer
Joined
Sep 20, 2018
Messages
664
Location
Canada
SL Rez
2007
Joined SLU
2011
SLU Posts
8642
I wonder how long before people start including instructions like "make a few credible typographical errors, misspelled words, and inappropriate punctuation."
I always include “Do not use — in your response to separate thoughts” since I find they all use this and a sure give away of AI written cover letters and proposals I receive.
 

Innula Zenovka

Nasty Brit
VVO Supporter 🍦🎈👾❤
Joined
Sep 20, 2018
Messages
23,984
SLU Posts
18459
I'm not sure I get the appeal.
I'm sorry -- I should have expanded on that. As I keep saying, I find AI very helpful for coding, but it's taken me a couple of years to learn how best to use it, and a lot of that has been learning the kind of prompts and settings it requires to make it do what I want it to. I think a lot of people are dissatisfied with AI, or use it to produce results everyone else finds unsatisfactory because, like any other tool, it takes time and practice to understand how to make it work for you.

So if I don't want AI to produce text for me that's obviously written by AI, it seems reasonable that it'll take some effort to make it write the way I want it to, and since AI is, as we are so often told, a stochastic parrot (though I don't think that adequately explains how it can look at my code and tell me why it's not working as expected), trying to get it to write like me rather than a stereotypical AI is probably a good start.

I'm not sure I'll ever use it for anything. The few times I've asked it to write boilerplate emails for me it's produced perfectly serviceable results that have led to useful replies, so I don't really feel the need to improve on that, and I don't want to use it to write posts here, except perhaps when I want to summarise something, but it just seems an interesting, and possibly useful, experiment.
 
  • 1Like
Reactions: Noodles

Free

*censored*
VVO Supporter 🍦🎈👾❤
Joined
Sep 22, 2018
Messages
42,554
Location
Moonbase Caligula
SL Rez
2008
Joined SLU
2009
SLU Posts
55565
Still not seeing it. Especially in regards to the "Check out this chat" material, which was about having it write social media posts (for "you", I assume).

Setting up an A.I. agent's parroting skills to use only ones previous writing to train on may mean it could possibly replicate ones style or word choice. But social media content tends to be confused, messy, opinionated and often dealing in current novelty. How could anyone expect it to respond to some topic or event presently in the zeitgeist you've never written on or engaged with previously? As we know it can't think your thoughts, so unless the goal is to have it post only on topics you're already discussed, or stick to simplistic platitudes, it seems like the effort would be pretty limited.
 
Last edited:

Noodles

The sequel will probably be better.
Joined
Sep 20, 2018
Messages
6,084
Location
Illinois
SL Rez
2006
Joined SLU
04-28-2010
SLU Posts
6947
I am so tempted ...

I actually would love to see what came up from a local AI trained on all my writing ever. I have converted essentislly everything I have ever written, including paper joirnals, since the 90s, into Markdown files. I have no idea the word count but its easily tens of thousands of little 500-10,000 word snippets. Probably millions of words.

Honestly, Claude already feels like its picked up a lot of my natural patterns when responding to me. And it already remebers everything about most of the computers I have hat it help me do things on. Often just by computer name, which shows up when I paste some terminal commands in.
 
  • 1Like
Reactions: Innula Zenovka