There have been times when satire has been mistaken for reality. A list of Onion stories that were mistaken for reality (or was it a mistake?) is here. When I say mistaken for reality I mean that a large set of people were fooled.
My own Ramsey-History-Hoax (blog here, latest version of the paper here) has fooled some people; however the number of people is small since the number of people who know the underlying math is small.
In my last blog (see here) I said that the Pope Leo XIV majored in math (that is true) and then I gave a false title for his thesis (I HAVE SINCE REMOVED THE ENTIRE PASSAGE).
Later in the post I said that his ugrad thesis was not on that topic, but gave another false title (I HAVE RMEOVED THAT AS WELL.)
I thought the reader would know that it was false, but one comment inquired about it so I left a comment admitting it was false.
This is all very minor: Not that many people read this blog and very few non-math people would care about what the topic of the Pope's undergraduate thesis.
The last part of the last sentence is false. Its the POPE! People Do care about his background.
But surely my blog post isn't so well read so as to make the fictional title of his thesis a hoax that fools a lot of people.
Even so, I left a comment wondering if LLM's might learn the incorrect title of the Pope's ugrad thesis.
A reader named E posted the following:
It might be too late. I did this search this evening:
E: Did Pope Leo XIV study Ramsey Theory?
Gemini: Pope Leo XIV, whose given name is Robert Francis Prevost,
earned a Bachelor of Science degree in mathematics from Villanova
University in 1977. His undergraduate thesis focused on Rado's Theorem
for Nonlinear Equations.
0) This may not be too bad- one would have to ask about The Pope and Ramsey Theory to get that answer. But in the future this answer might pop up on the question`What did the Pope Study as an Undergraduate' or similar questions.
1) Might future satires or April Fool's Day jokes be mistaken for reality in the future by AI and hence reach a much larger audience than this blog does?
2) If so, should we be careful with what we post (not sure how to do that)?
3) What about people who have a much larger following than complexityblog (yes, there are such people)?
4) In the past one had to be a celebrity or similar to change peoples perception of reality (see Stephen Colbert and Wikipedia here). Now a complexity blogger may be able to change people's perception of reality. Hence I ask
Is Satire Dangerous in the AI-Age?
I think the incorrect statements you posted would have to be funnier and more interesting before it makes sense to call them satire.
ReplyDeleteAGREE for the example of what I posted. The question of if posting false statements of any sort is dangerous is real point.
DeleteWow, you wrote "This statement is false", and the "AI" belived it. This is not intelligent.
ReplyDeleteWhy should we change the way we write for something this stupid. If people belive this stupidity, it is not your fault. Also false statements everyone belives is not a new thing. Historical Philosophy and Pre-Sience are full of them.
That wasn't me, but it should have been.
DeleteMore specifically, LLM's don't do "truth", they just do random text generation.
So lying to an LLM is going to cause problems.
The bottom line is that the whole LLM idea should have been laughed out of the room from the start. It's exactly and only automated stupidity on steroids.
For Bill: I know you collect TCS and math related songs, but have you thought about collecting TCS related crossword puzzle clues? Here's a good recent one from The Times crossword:
ReplyDelete"Computer hypothetically retaining much when reconfigured"
I already collect to much- but I AM curious whatthe answer is to your clue.
DeleteThe clue was too short because I forgot to add the usual letter counts so it should be "Computer hypothetically retaining much when reconfigured (6,7)". This should make it doable.
DeleteThis is a cryptic clue: The meaning is covered by the first two words and the "when reconfigured" is a hint that it is an anagram of the other two preceding words.
DeleteI solved it! I cheated! I typed in the clue to google and got the answer. I then verified it since the answer was indeed an anagram of `retaining much' clever! In an earlier era I would have had to try to do it myself- PRO- exerices the mind. CON-I might not have been able to get it.
DeleteIn my humble opinion, I believe it is dangerous regardless of the recent surge in AI's capabilities. But, should one abandon satire? Absolutely not. But, perhaps a tag or a label distinguishing said articles would be appreciated (especially if truth is the frequent attribute of the publisher's posts). It is already hard to navigate AI-fabricated posts--let alone appending human-made ones as well.
ReplyDeletethere is a dilemma here. IF you label it, it takes away some of the fun. For example, satire article that said Trump wanted to annex the Vatican is only effective if while reading it you believe it (I did). The LAST paragraph reveals that its a satire, but I can imagine someone not getting that far. But I AGREE with you- this was already a problem in the past (my Ramsey Satire has fooled one HS student and one College student who put the results in their thesis) but it will get worse with AI.
Delete