AI Generated NPCs

[BaHalus]

I tried to use chat GPT this way, not very good at that.

But then I used it the other way around. I generate a *.txt GCS file and asked for a background for a character with those stats.

The result was pretty good. I gave it three tries (the last one was horrible, as it forgot completely the original character sheet). I use most of the elements of the first background, a few of the second, and minor adjustments.

[RyanW]

Originally Posted by BaHalus

I tried to use chat GPT this way, not very good at that.

But then I used it the other way around. I generate a *.txt GCS file and asked for a background for a character with those stats.

The result was pretty good. I gave it three tries (the last one was horrible, as it forgot completely the original character sheet). I use most of the elements of the first background, a few of the second, and minor adjustments.

I was generating some NPC names and backgrounds (mostly so everyone in the background of a campaign isn't named Dave) and it came up with Caroline Chang, soon to be sworn in as the first female US President of Asian descent (not intended to be a political game, so the President is a background character). Born in 2036 and elected in 2060.

I pointed out that she was too young to be President, and ChatGPT assured me that while most to hold the office were significantly older, the age requirement was only 35. So then I asked how old she was when elected.

"Caroline Chang was born in 2036 and the hypothetical election year is 2060. Therefore is Caroline Chang were to win the 2060 Presidential election, she would be 24 years old in 2036, plus 24 years between 2036 and 2060, making her 48 years old at the time of her election." (Emphasis added)

I've heard that girls mature faster than boys, but being 24 when she was born is a bit much.

[BaHalus]

Yes! GPT is really awful with mathematics!
Never trust its results.

[RyanW]

Originally Posted by BaHalus

Yes! GPT is really awful with mathematics!
Never trust its results.

It is, though, pretty good at creating Fate character sheets.

Name: Avatar Kyoshi
High Concept: Powerful Earthbending Avatar

Trouble: The Burden of Legacy

Aspects:

• Unwavering in the face of adversity
• Protector of the Earth Kingdom
• Long shadow of past Avatars
• Unconventional problem solver
• Struggles with personal relationships

Approaches:

• Careful +2
• Clever +3
• Flashy +1
• Forceful +4
• Quick +1
• Sneaky +0

Stunts:

• The Avatar State: Once per session, can tap into the full power of the Avatar State to unleash devastating elemental attacks or overcome seemingly impossible obstacles.
• Earthbending Mastery: +2 to carefully overcome obstacles related to earthbending.
• Kyoshi Islander: +2 to cleverly create advantages related to her connection to the people and culture of Kyoshi Island.

Refresh: 3

Physical stress: [ ] [ ] [ ]
Mental stress: [ ] [ ] [ ]

Consequences:
Mild:
Moderate:
Severe:

The only real errors here are that it always picks two extra aspects (should be five including High Concept and Trouble, but it always picks five in addition to High Concept and Trouble) and it doesn't seem to know the difference in Fate Core and Fate Accelerated. Here I had asked for Fate Core, and it instead generated a Fate Accelerated character, but included the separate physical and mental stress boxes which are a feature in Fate Core.

It is, overall, a very good representation of Avatar Kyoshi. It made her a little more powerful than a standard PC (usually, approaches are one +3, two +2, two +1, and one 0).

Edit: And of course as soon as I say it "always" picks two extra aspects, it does the next one I correctly.

[Opellulo]

Like all ML shenanigans the more (precise) data it has the better results it provides: character sheets for D&D are mostly fine, the ones for Pathfinder are often hilariously mixed between 1st and 2nd edition and with Vampires it get dizzy often mixing versions and WoD sources.

Another funny test is asking for sheets of famous people; it's mostly there just don't ask for nuance:

Name: Kayne West Total points: 150 Attributes: ST 10 [0]; DX 11 [20]; IQ 13 [60]; HT 11 [10]. Advantages: Charisma +2 [10]; Fame +3 [15]; Musical Ability +4 [20]; Wealth (Very Wealthy) [30]. Disadvantages: Egomania [-15]; Fanaticism (Christianity) [-15]; Impulsiveness [-10]; Enemies (Media) [-10]. Skills: Acting-12 [1]; Singing-16* [4]; Business-14** [4]; Musical Composition-17* [8]; Art History-14** [4]; Politics-12** [2]; Religious Knowledge-14** [4]; Diplomacy-13*** [+1/2]; Leadership-14*** [+1/2]; Musical Aesthetics-17* [+8]; Woodworking-9**** [-1]; Marketing-13** [+1]; Public Speaking-15*** [+1]; Applied Psychology-12** [+1]; Socializing-13*** [+1].
Equipment: Designer clothes; Luxury car; Luxury house; Laptop; Microphone; Wristwatch; Smartphone; Recording studio.

Name: Karl Marx Total points: 150 Attributes: ST 9 [-10]; DX 10 [0]; IQ 15 [100]; HT 10 [0]. Advantages: Fame +2 [10]; Mathematical Genius [15]; Literary Talent +2 [10]. Disadvantages: Allergy (Tobacco) [-5]; Enemies (Bourgeoisie) [-20]; Poor [-15]; Partial Deafness [-5]; Chronic Illness (Bronchitis) [-20]. Skills: Economics-18* [16]; Philosophy-18* [16]; Journalism-17** [4]; English-14*** [+1]; Native Language (German)-15*** [+0]; Mathematics-18* [+8]; Politics-17** [+4]; Comparative Religion-14*** [+1]; History-17** [+4].
Equipment: Worn clothes; Beard; Hat; Books; Glasses; Pen and paper.

The thing i find disappointing is that often the chat can't follow my changes in languages or the weird cross language slang... All the tech in the world has the same exact problem of the T9.