A defendant in a felony case in the United States will use an "AI Told Me to Do It" defense before the end of 2023
55
69
1.1K
resolved Jan 7
Resolved
YES

In 1979 the https://en.wikipedia.org/wiki/Twinkie_defense was first used. Though not actually the basis of the defense, and really an argument of diminished capacity, never the less the consumption of Twinkies was a part of an argument in a successful defense strategy that allowed a defendant who admitted to killing a victim to obtain a Not Guilty verdict.

This market will resolve to TRUE if anyone posts verifiable evidence in the comments no later than a week after closing that by the end of 2023 a defendant in a Felony case bases part of their defense on an AI system having an influence on them. For example, claiming that a ChatBot told them to commit the crime, or was misled that an action wasn't a crime, etc.

This market will not resolve to YES on the basis of use of AI in a crime unless the defense concedes that the act was commited, and attempts to mitigate culability on the basis of influence by AI.

Jan 5, 6:08am: A defendant in a felony case will use an "AI Told Me to Do It" defense before the end of 2023 → A defendant in a felony case in the United States will use an "AI Told Me to Do It" defense before the end of 2023

Get Ṁ200 play money

🏅 Top traders

#NameTotal profit
1Ṁ723
2Ṁ187
3Ṁ87
4Ṁ71
5Ṁ69
Sort by:
predicted NO

@RealityQuotient Six hours of deliberation before closing this market 8 days into 2023 is not cool. Automation fatigue is not at an "AI Told Me to Do It" defense, and a little more deliberation would have made the resolution clear.

predicted YES

@FlawlessTrain I was a bit uncertain when I made my suggestion (see: sold ~half my position) but evaluated primarily on the description, specifically:

a defendant in a Felony case bases part of their defense on an AI system having an influence on them

"Automation fatigue" is an influence that the AI had on Vasquez, and her defense tried to make the argument that that influence mitigates her guilt.

I do agree that time for deliberation would have been a good thing.

predicted NO

Precision over meaning. It's a stable equilibrium, but one only for professionals.

bought Ṁ100 of YES

Ms. Vasquez had been an operator on this same loop 73 times previously, without incident. The repetitive nature of the task, coupled with the false sense of security provided by Uber's touting of the vehicle's technology, had a profound effect on her, and any other operator piloting such a car.

https://ewscripps.brightspotcdn.com/2c/a0/b445d3924af4a8383573c012c9be/remand-motion-redacted.pdf

This is that big uber self-driving case.

Rafaela Vasquez is charged with negligent homicide for the death of Elaine Herzberg in Arizona.

The defense focuses heavily on the concept that "automation fatigue" absolves the driver of guilt. She had learned to trust the AI to navigate, and therefore Uber is primarily to blame for Herzberg's death.

bought Ṁ20 of YES

@citrinitas seems like this should count?

sold Ṁ49 of YES

@citrinitas @jonsimon This is a tough one, and I'm not a lawyer. Note: I'm not sure if this is the proper way to handle these things, but I sold my stake before I began deliberating. If it had been obvious I would 't have done so, but in this case there could be some controversy either way.

Reasons for resolving YES:

  • Vasquez clearly claimed lack of culpability due to the Uber self-driving system.

  • This system clearly had "AI" capabilities.

  • It's a class 4 felony, and thus a felony.

  • It's pretty clear that Vasquez concedes that that crime/event took place.

  • This took place before the end of 2023.

Reasons against YES on this incident:

  • Negligence is, by definition, not intentional, and the question was about an "AI made me do it" defense.

I didn't consider negligence when creating this market, and had unthinkingly considered lumped all felonies together to be intentional. However, I chose Felonies intentionally to make it very easy to determine how to resolve this market.

As such, my decision is to resolve YES. I'll create a new market now asking if such a defense will be used, not including negligence.

predicted NO

@RealityQuotient I interpreted the 'will use' in your title as meaning that someone would use this defence between the date of you making the market and the end of 2023, and made a trade based on this assumption.

It is too late now as you have resolved it, but I think this could have been made more clear, or clarified in the description when you made it.

predicted NO

@WXTJ If you had said 'will have used' I would've understood it in the way you intended.

Not a defense, but this was faster than expected: https://interestingengineering.com/innovation/ai-defend-case-us

predicted NO

If this market were run in 1979, and "AI Told Me to Do It" was replaced with "twinkies made me do it", would the market resolve yes (as a result of the trial of Dan White)?

@FlawlessTrain actually, no. Based on how most people think that Twinkies we're used as a defense, yes, it would. However, there's a misconception. The Twinkie defense is a myth. The mention of eating Twinkies was meant to be a symptom of whatever was wrong with Dan White.

That said, for clarification purposes, if it went down like the myth, then yes.

Does is have to be US or would somewhere in europe count, too?

What exact definition of felony will be used? (in the US different states seem to use the word differently https://en.wikipedia.org/wiki/Felony#Classification_by_seriousness )

@BjornJurgens and: the defense doesn't need to be successful, right? So if the defense didn't help the defendant at all, it would still count to resolve this market to yes?

@BjornJurgens I assumed the U.S. when using the word Felony, updating the question to make it clear.

@BjornJurgens Correct, that the defense does not have to be successful.

More related questions