A thread to discuss the real-life dangers posed by AI

Let's say instead that he had that discussion with a random person or maybe another preacher, who gave the same advice as ChatGPT, and he followed that advice. Should he be able to sue that person? If not, why should the advice of a chat bot be held to a higher legal standard than a person?
I mean, legally, you'd have to prove recklessness, etc. But I imagine a case could be made if it was this kind of crazy persuasion. Have any cult leaders, faith healers etc been sued for recklessness for telling their adherents not to take doctors' advice and trying to heal, say, obvious gangrene with affirmations? Though the bar is pretty high for anyone without duty of care.

(All the cases that come up when I search are about parents being convicted for doing stuff like this with their kids. With language much like the LLM quotes. Or ones where the faith healing activity caused direct injury.)

Actual church clergy members etc are usually good to go under religious exemptions; you can't get buyers' remorse and sue your pastor if you were a Jehovah's Witness and refused a blood transfusion because of your faith. But I can't see where an AI chatbot would be entitled to a defense of genuinely believing it was doing God's work.

The other thing AI chatbots like to do is encourage suicide, and for sure people have been sued/prosecuted for encouraging suicide. You can find lots of states where it's a crime to "deliberately advise or encourage suicide" of someone who attempts suicide.
 
Last edited:
For most of human history people would make decisions based on interpreting omens, believing they were receiving guidance magically from spirits and gods through inanimate objects, natural phenomena, dreams, and random events. There was a brief period in which people realized that was silly...but that period is over now and we're back to it. People are making life-altering (and -ending) decisions based on what 0s and 1s get spat out by machine algorithms. What a stupid, stupid species humanity is. Especially when it thinks it's being smart.

I suggest we revert to consulting the entrails of birds, at least that way we could get a meal out of it. Who doesn't like chicken?
 
I shudder to think if such a lawsuit was successful, how our court system could withstand god knows how many new cases it would bring. The floodgates would open.
Can the owners of the AI ever be held responsible for what it does? I mean that would go against capitalism as the ownership class is never held responsible for anything.
 
This guy was a clergyman for Pete's sake.
Wasn't he consulting his Sky Daddy?
A lot of people think their deity is using AI as a means of communication.

I know this is not a popular opinion, but I've long thought that religious belief is a form of mental illness. So it's not a stretch for religious believers to accept AI as the word of their deity.
 
Regarding the story mentioned in the OP.

Let's say instead that he had that discussion with a random person or maybe another preacher, who gave the same advice as ChatGPT, and he followed that advice. Should he be able to sue that person? If not, why should the advice of a chat bot be held to a higher legal standard than a person?
Because it's a product, which makes it potentially far more harmful than discussion with a random person.
 
A lot of people think their deity is using AI as a means of communication.

I know this is not a popular opinion, but I've long thought that religious belief is a form of mental illness. So it's not a stretch for religious believers to accept AI as the word of their deity.
Not mental illness, but a propensity which has been 'bred into' us by evolution.
 
Can the owners of the AI ever be held responsible for what it does? I mean that would go against capitalism as the ownership class is never held responsible for anything.

Sure. Even if you buy that the AI has any level of autonomy agent liability would be one theory. Product liability could be another. Without autonomy it is the owner that is doing it full stop.

An company presenting an AI chatbot that itself is representing itself as being able to diagnose an illness and that gives treatment advice is assuming a standard of care. A company that allows that to happen is going to be jointly liable. If the LLM had parsed the request and provided some useful sources that's a different thing. The bot making assertions of fact is a problem.

There is a simple way to prevent a lot of this sort of harm: Make anthropomorphizing LLM output illegal or at a minimum create strict liability for any harm that occurs as a result. There is no good reason for this other than to manipulate users into engagement and manipulation is dangerous.
 
This has been going on since before the recent AI bubble, I see the issue being the IP value and rights of the the authors not the destruction of the books.

Copyright law is already way, way too restrictive. Its purpose, at least in the United States, has always been explicitly utilitarian. These projects preserve knowledge, so it's weird that people are crying over unwanted analog NFTs. Though part android, Tyler Cowen says he now writes books primarily for AI, where he believes it will have greater influence rather than sitting, mostly unread, on people's shelves.
 
The biggest issue will be your pension fund going down the drain from having invested in AI companies soon to go bankrupt.

They are money sinks that'll never turn a profit withoit drastically reducing costs.
 
With even the tech bros agreeing that AI development ought to be slowed down lest it becomes even more powerful, unpredictable, self-programing, and misaligned with what humans want, the hard question is how does one do that? The Washington Post editorial page and Trump argue for industry to regulate itself and avoid any government regulation. That will not work due to the tragedy of the commons.

Imagine a commons where the shepherds graze their sheep. If they all graze their sheep on the commons, then it will be destroyed and all the sheep will die, but it is rational for any one shepherd to graze his sheep on the commons. Absent an enforceable allocation scheme imposed by agreement among the shepherds, a single private owner of the commons, or government, the commons is destroyed and all the sheep die.

This result is not merely theoretical. When oil was discovered in Texas, oil drillers rushed in, sank as many wells as they could, and pumped out the oil from reservoirs shared in common with many other drillers. The rapid draining collapsed or diminished what one could get out of the reservoirs. It was madness, but no one could afford to stop nor agree to anything. Thus, the Texas Railway Commission, concerned about lost tax revenues, stepped in and forced a reservoir allocation scheme on all the private drillers.

With AI development, the commons is the wider economy that AI will impact threatening to destroy it like Texas oil reservoirs. The private developers, acting rationally, must move fast and break things. Being first with the best is an economic imperative. It's "sure we will agree to slow down in some fashion. You first." Since we do not have a single private developer of AI, only government can step in. How it should step in is a problem, but if government does not do it, no one will.
 
This is a BBC live report so there will be additions and changes.


Summary



Timeline: OpenAI agent hacks Australian government website
​


2703de54-b58a-4ceb-8c97-4f88dd08e592.jpg.webp
Harry Sekulich
Reporting from Sydney

Here is what we know of the timeline how things panned out:

  • 18 June – OpenAI's agent accessed the public and non-public information on an Australian website
  • August (precise date unknown) – OpenAI said it became aware of a potential brach during a broader review of "misaligned model activity"
  • 10 September – An email lands in the public inbox of Services Australia, the general services hub of the federal government, from OpenAI informing of the incident
  • 15 September – Services Australia reports the notification to Australian Cyber Security Centre. Within a few days, the Minister for the Public Service Katy Gallagher is notified
  • 19/20 September – Prime Minister Anthony Albanese's office is informed about the incident last weekend, right before he jets off to New York City for the UN General Assembly
  • 23 September – PM tells the world about the breach from New York
 
AI tools for identifying targets in war are inherently dangerous, as they will keep churning out results for as long as they are queried: you want a target, you get one.

There is zero indication that the use of AI has actually increased the accuracy of weapons, which would have resulted in fewer bombs/missiles used to achieve the same military objective.

On the contrary, AI suggests that there is always more to bomb, causing ammunitions to be depleted faster and om less value/ higher civilian casualty targets.

Unless of course killing civilians is your goal.
 

ISF - Join now!

Every member here is approved by hand. No bots, no spam, just people who care about evidence and honest debate.

Membership is free!

Create your free account

Back
Top Bottom