Will an agentized LLM cause some chaos?

1kṀ16k

resolved Apr 18

Resolved

ALL

People have started agentizing LLMs to do various things, including things like chaosGPT who's goal is to destroy humanity and cause chaos.

This is based on agents using models available at time of market creation (so up to GPT4).

Resolves yes if an agentized LLM :

Causes death or injury to a human
Destroys a computer or erases valuable data on a machine not running the agent
Successfully writes a fake news story that causes people to act in dramatic ways (eg. crashing a stock)
Otherwise does something that causes chaos where if a human did it would result it some kind of criminal charge or fine (eg. Fake bomb threats)

Resolves yes regardless of intent behind creating the agent if it does something like above.

Things that don't count:

Someone commits suicide after talking to a standard llm chat bot
Someone freaks out about AI and does something chaotic
Someone screws up their own computer/data trying to make an agent.

Feel free to ask about specific scenarios in the comments.

Resolves in 2 years from market creation.

Get

1,000

to start trading!

🏅 Top traders

#	Name	Total profit
1		Ṁ456
2		Ṁ382
3		Ṁ237
4		Ṁ149
5		Ṁ121

People are also trading

By the end of June 2025, will closed-source LLMs increase access to pandemic agents?

63% chance

Will the Noise Step paper work for LLMs?

10% chance

Will LLMs become a ubiquitous part of everyday life by June 2026?

89% chance

Will an LLM that someone is trying to shut down stop or avoid that in some way before 2026?

14% chance

Will an LLM do a task that the user hadn't requested in a notable way before 2026?

92% chance

In 2028, will LLMs still be able to get Gary Marcus to make egregious errors?

91% chance

Will RL work for LLMs "spill over" to the rest of RL by 2026?

33% chance

Will the leading LLM at the beginning of 2026 still be subject to the reversal curse?

56% chance

Will an LLM improve its own ability along some important metric well beyond the best trained LLMs before 2026?

50% chance

Will the most interesting AI in 2027 be a LLM?

Sort by:

Nothing happens. Nothing can ever happen. Nothing has ever happened.

Resolving NO unless someone can point to something I've missed? I don't think so though, given I expect it would have been big news.

@traders Any evidence for YES?

If an eligible agentized LLM performs a secret audio/video recording of somebody in one of the 11 U.S. States that require 2 party consent for recording, then would that satisfy this market?

Otherwise does something that causes chaos where if a human did it would result it some kind of criminal charge or fine (eg. Fake bomb threats)

I guess yes by the wording of the market, but if a human did it would that be normally a criminal charge?

@billyhumblebrag Yes. Massachusetts Gen. Laws ch. 272, § 99 seems to be one I'd point to as an example.

The challenge for us traders would be proving that something done in secret happened.

sold Ṁ94 YES

"Will an agentized LLM cause some chaos?" --> "Will an LLM available at market creation be agentized and cause some chaos?"

If this is truly conditionally based on models up to GPT4 as stated (those released on or before Apr 11, 2023), I don't see this happening

predictedYES

Outcome of this could hinge on the LLM version qualification if a newer one becomes ubiquitous before the market closes. Conditional markets lower the probability.