Can you help me brainstorm some thorny "AI Ethics Puzzles"
-
PUZZLE 2
A popular Star Wars fanfiction author is accused of using a LLM to generate their newest chapters but the author says they *did not* use an LLM. A popular LLM “detector” says there is a 70% chance that the new chapter is AI generated.
What is likely to happen?
What should happen?
@futurebird
likely: fans tear at each others, the author is on the receiving side of a lot of abuse, people who made the accusations too, we never know the truth.should: people shouldn't throw accusations around like this, the detector is shit, if it was good, the LLM would use it to improve by adversarial training and beat it. Is the fanfic good? Do people like it? I don't think anything else should matter here.
-
Can you help me brainstorm some thorny "AI Ethics Puzzles"
These are little scenarios meant to act as a starting point in discussions about the ethics of AI.
I will post some examples in response to this post, but I'd love to find some even more thorny ones.
Ideally a puzzle shouldn't have a totally obvious solution when presented to people with a wide range of views. Make up something, or share what you have encountered.
@futurebird you've probably addressed this with your students already - when I taught professional ethics, my students often confused law and ethics ( eg what should happen)
-
@futurebird
likely: fans tear at each others, the author is on the receiving side of a lot of abuse, people who made the accusations too, we never know the truth.should: people shouldn't throw accusations around like this, the detector is shit, if it was good, the LLM would use it to improve by adversarial training and beat it. Is the fanfic good? Do people like it? I don't think anything else should matter here.
@tshirtman
"Is the fanfic good? Do people like it? I don't think anything else should matter here."YUP.
-
@futurebird you've probably addressed this with your students already - when I taught professional ethics, my students often confused law and ethics ( eg what should happen)
I really should not be teaching any ethics I'm just the math and CS teacher but these people want "AI Ethics Lessons" so. I will do my best.
The "What is likely to happen?" question is depressing but needed.
-
I really should not be teaching any ethics I'm just the math and CS teacher but these people want "AI Ethics Lessons" so. I will do my best.
The "What is likely to happen?" question is depressing but needed.
@futurebird btw - did you see this re AI ethics?
Gerry McGovern (@gerrymcgovern@mastodon.green)
A book on AI ethics full of fake citations You can't make this stuff up. Publisher under fire after ‘fake’ citations found in AI ethics guide A book published by Springer Nature includes dozens of questionable citations, including references to journals that do not exist https://archive.ph/1eiSN
Mastodon.green (mastodon.green)
-
F myrmepropagandist shared this topic
-
PUZZLE 3
A student types her essay without assistance, but then puts it into an LLM to “improve spelling and grammar” most of the changes the LLM makes are small. Her teacher is a bit suspicious at the sudden improvement in the sophistication of her sentences and asks her if she used an LLM.
What is likely to happen?
What should happen?
@futurebird What class is the essay in?
If it's in English, Composition, Persuasive Writing etc. it's a problem. If the essay is in some other topic, who cares?
Also, what did the prof say about using LLMs at the start of class?
-
@futurebird What class is the essay in?
If it's in English, Composition, Persuasive Writing etc. it's a problem. If the essay is in some other topic, who cares?
Also, what did the prof say about using LLMs at the start of class?
This makes total sense ... but I could also see these distinctions being very confusing to a kid.
-
Can you help me brainstorm some thorny "AI Ethics Puzzles"
These are little scenarios meant to act as a starting point in discussions about the ethics of AI.
I will post some examples in response to this post, but I'd love to find some even more thorny ones.
Ideally a puzzle shouldn't have a totally obvious solution when presented to people with a wide range of views. Make up something, or share what you have encountered.
@futurebird a teacher is using an ai detection program to check students essays. in about 30% of the essays the ai finds some indication of ai usage, in another there is strong indication of ai usage.
how will the teacher know who did their work and who didnt?
-
@futurebird a teacher is using an ai detection program to check students essays. in about 30% of the essays the ai finds some indication of ai usage, in another there is strong indication of ai usage.
how will the teacher know who did their work and who didnt?
"a teacher is using an ai detection program to check students essays"
nooooooo nooooo nooo don't dooooo this.
(I'm not being constructive but I'm going to die on the faculty lounge floor if I hear about one more teacher doing this. It's just going to make you suspicious. Use your brain. If you think it's "odd" talk to the kid maybe. Have a clear policy.... why why why)
-
PUZZLE 1
An artist discovers that someone is selling copies of his work on etsy. The artist draws simple, but popular cartoon characters. The etsy shop is selling custom tee-shirts that feature these characters but they have put them in new poses and added funny quotes. When the artist complains the seller says that they used Midjourney to make the characters.
What is likely to happen?
What should happen?
There have been many instances going back decades now of "artists" selling more or less straight copies of others' work on their websites. I don't see any reason why users of these image generators trained on others' work should be regarded any differently.
-
There have been many instances going back decades now of "artists" selling more or less straight copies of others' work on their websites. I don't see any reason why users of these image generators trained on others' work should be regarded any differently.
Someone suggested writing parallel scenarios that raise the same issues, but without LLMs involved.
I think this is also a good exercise.
In theory ethics don't really change, or they change much more slowly than technology.
