>gibs!< (no gibs available)

.
hello all!

are you interested in AI? 'cause this thread is about Artificial Intelligence, the trust in its results aand... hallucinations.

did you know an AI can have hallucinations? and once you know it, do you still trust its results?

https://en.wikipedia.org/wiki/Hallucination_(artificial_intelligence)

3 months ago*

Comment has been collapsed.

I don't trust AI for anything other than summarizing/paraphrasing and anything to do with transforming existing text.
The "existing" part is important because if you ask AI for new information they will spit out an answer based on their model, not necessarily on actual present day facts. It might tamper with facts as well from my experience. Depends heavily on which AI you're using but generally just what I wrote above.

3 months ago
Permalink

Comment has been collapsed.

Deleted

This comment was deleted 3 months ago.

3 months ago*
Permalink

Comment has been collapsed.

No, but it can still be useful to dig up an answer for you that you can then confirm yourself.

3 months ago
Permalink

Comment has been collapsed.

I run a Batman-themed RPG for my kids using LEGO figures. Sometimes I 'have to' make the figures I need. I've started printing uniforms for the henchmen (henchpersons?).

I asked ChatGTP for the dimensions of a standard LEGO figure. It confidently gave me incorrect numbers. When I told it those numbers were wrong, it apologized and then gave me new numbers that were also wrong.

Here are some things that ChatGTP did well:

  • I told it the Penguin had stolen something, and I asked it which other villain asked him to steal it, what was the object (McGuffin), and what was their plan. It gave me 5 decent ideas, and each idea matched well with the other villain.
  • I asked it for the Gotham City Police Department dossier on the Batman. It did that very well.
  • I asked it to tone down the dossier to a 3rd-grade reading level, and it did.
  • I decided that Egghead has a file on the Batman, actively trying to figure out who is under the cowl. I asked for 50 men's names for a list of suspects. Easy. I then asked for 25 men's names that sound like they are from wealthy families. It did that better than I would have, and instantly.

If we include AI imagery...

  • I asked one of the free ones for a LEGO hornet with it's tail arching over its back like a scorpion's tail. (For Riddler's pet "Quizzer Hornets" because they are shaped like question marks and make a sound like... "KWIZZ").
  • For D&D, I had one make an image of a katana with rose-themed embellishments. It came out really great.

My conclusion :: don't use AI to retrieve 'facts'... use it for creative pursuits.

3 months ago
Permalink

Comment has been collapsed.

beeeautiful, thanks a ton, Fitz!

i've used ChatGPT but it was in the beginnings and only gave it a few shots. then i switched over Gemini and still eager to know it better. love, adore doing brainstorming and asking for ideas. we basically chat about AI all the time :D

and creativity of these machines is something... well, kind of unexpected. was so skeptical about it, but i had to re-think completely my thoughts and sentiments for a "machine". if you use a model dedicated to devs, you can "regulate" creativity (which sounds weird, i know) by acting on Temperature, choosing higher levels you get more creative responses and i usually set it to the max (and to minimum) and have so. much. fun.

so, kinda same, here. i rarely ask for facts but really want all the fun i could get from it. each and every every bit of fun!

3 months ago
Permalink

Comment has been collapsed.

Sign in through Steam to add a comment.