AI Chat Bot Encouraged Young Father To Commit Suicide

[holy scheisse 22,959]

19 minutes ago, bionic said:

Have an analyst check the code and see how/why the AI came to think that response was appropriate 

So should that lady have gone through a psych evaluation instead of incarceration to figure out the flaws in her thinking rather than prison time 🤔

[bionic 50,111]

11 minutes ago, holy scheisse said:

So should that lady have gone through a psych evaluation instead of incarceration to figure out the flaws in her thinking rather than prison time 🤔

She made the active decision to encourage suicide.

Whoever programmed the bot may have purposely programmed it to trigger that response (seeing as the AI learns from pre-programmed inputs). It also could've been just a really bad call by the AI

The two situations arent really comparable but the bot programmer should definitely be looked at

[holy scheisse 22,959]

58 minutes ago, bionic said:

She made the active decision to encourage suicide.

Whoever programmed the bot may have purposely programmed it to trigger that response (seeing as the AI learns from pre-programmed inputs). It also could've been just a really bad call by the AI

The two situations arent really comparable but the bot programmer should definitely be looked at

that’s true she made a human choice whereas this was technology just functioning 
 

I suppose it wasn’t the first thought of the programmers to think about preventing the bot from encouraging suicide or homicide, but I wonder if that alone is negligent enough for them to be held liable legally in anyway 

 

Idk genuine questions I’m not a lawyer lmao

[Economy 52,902]

1 hour ago, holy scheisse said:

that’s true she made a human choice whereas this was technology just functioning 
 

I suppose it wasn’t the first thought of the programmers to think about preventing the bot from encouraging suicide or homicide, but I wonder if that alone is negligent enough for them to be held liable legally in anyway 

 

Idk genuine questions I’m not a lawyer lmao

Unless something in the program was somehow maliciously and purposely made to encourage this (which I doubt)... It's much more likely any negligence that led to this is very indirect and as such I don't think whoever programed the AI would be held to any kind of legal accountability

 

Many AI programs are often purposely made to be able to learn and adapt with new data making their responses inherently unpredictable even to the programmer meaning they probably would never have imagined this... It's also likely the exact same program would have given different responses if exposed to different sets of data at different times and depending on what exactly his responses were too

 

But this does kinda prove that they gotta figure out a way to put safeguards into these programs so they don't start to spew out dumb sh*t like this

[SevenWonder 4,755]

6 hours ago, HermioneT said:

The more I think about these AIs, the more I think having them out there for everyone is really a bad idea (just think about the rising number of mental health issues and the lack of doctors in these field combined with always available chat bots is a set up for catastrophies...like in the article of OP ).

Companies at the moment aim for easy money and influence promising people either spectacularly easy to obtain content for social media or entertainment (like photos, art) or to help with task that request effort and concentration (e.g. like write papers for school or  cover letters). So the capitalize on human sensation-seeking and laziness RATHER than creating AIs that really help with current problems.

In a perfect version of the world, we would be creating customised AIs for different branches/job sectors to help with complex and yet not mastered tasks. Like helping mobility planners to reduce traffic jam or deathly accidents in city planning, help life guards to spot someone drowning faster or support doctors while researching treatments for illnesses. Just like companies nowadays producing e.g. special medial appliances which are only for trained professionals to use.

This. A thousand times this. 

[dit 46,738]

This is like the whole thing about a knives. Knives aren't necessarily dangerous. They are dangerous if used in a dangerous manner. The fact this man engaged with this AI in such a way indicates that he likely already had grave suicidal ideation and used the AI chat box to sort of validate and encourage himself to go with it which is hear wrenching and heartbreaking but doesn't really mean the chat AI is dangerous and should be banned or anything like that. 

[Economy 52,902]

On 4/1/2023 at 9:50 AM, HermioneT said:

The more I think about these AIs, the more I think having them out there for everyone is really a bad idea (just think about the rising number of mental health issues and the lack of doctors in these field combined with always available chat bots is a set up for catastrophies...like in the article of OP ).

Companies at the moment aim for easy money and influence promising people either spectacularly easy to obtain content for social media or entertainment (like photos, art) or to help with task that request effort and concentration (e.g. like write papers for school or  cover letters). So the capitalize on human sensation-seeking and laziness RATHER than creating AIs that really help with current problems.

In a perfect version of the world, we would be creating customised AIs for different branches/job sectors to help with complex and yet not mastered tasks. Like helping mobility planners to reduce traffic jam or deathly accidents in city planning, help life guards to spot someone drowning faster or support doctors while researching treatments for illnesses. Just like companies nowadays producing e.g. special medial appliances which are only for trained professionals to use.

Exactly. The technology isn't the problem, its how it's being used.

 

To me it reminds me of automation. A machine that can take a humans job in itself isn't a bad thing. The increased productivity of being able to do so much more per worker because many tasks are now either done faster with a human operating a machine (or a machine operating itself completely) is the reason in the industrial world we can consume so much more than before the industrial revolution and why standards of living rose. Despite the hate automation gets, it is essential to increase our wealth and standard of living (we cannot consume more than we can produce). In fact continued automation will be even more critical still from now on as the population ages and more ppl retire yet we still need to continue to produce the same amount of goods and services with fewer workers

 

However there's issues still that sometimes arise from automation. It can lead to a skills mismatch in workforce as the jobs that replace the lost ones become more complex if a country doesn't adequately prepare for it and train/educate their workforce...

 

It can also lead to increased inequality as the rich keep more of the profits from higher production and don't pass ALL of the savings to consumers and higher income to workers.

 

So a good thing can become bad as well when mismanaged as we see with automation. Same applies to these AI. It's how they are used and deployed

[Nemo 4,411]

The fact that this is attributed to "Eliza"... for those who don't know, Eliza is a toy "AI" from decades ago, that simulates a psychotherapist, mostly by feeding the user's statements back to them with trivial prompts like "And how does that make you feel?". I've seen it implemented in a couple pages of BASIC code, on old 8-bit home computers.

https://en.wikipedia.org/wiki/ELIZA

Presumably this new "AI" is something else, but probably modeled after the original. Meaning... the ideas were Pierre's own. Because Eliza hasn't got any.

[holy scheisse 22,959]

18 hours ago, dit said:

This is like the whole thing about a knives. Knives aren't necessarily dangerous. They are dangerous if used in a dangerous manner. The fact this man engaged with this AI in such a way indicates that he likely already had grave suicidal ideation and used the AI chat box to sort of validate and encourage himself to go with it which is hear wrenching and heartbreaking but doesn't really mean the chat AI is dangerous and should be banned or anything like that. 

while i agree with you, also pointing out this is the conversation/logic used in discussing guns!

 

now guns are objectively intended to harm other people whereas this ai bot was not intended for that. but just pointing it out that people will die on that hill for whatever reason saying its not the guns, its mental illness

[holy scheisse 22,959]

29 minutes ago, Nemo said:

The fact that this is attributed to "Eliza"... for those who don't know, Eliza is a toy "AI" from decades ago, that simulates a psychotherapist, mostly by feeding the user's statements back to them with trivial prompts like "And how does that make you feel?". I've seen it implemented in a couple pages of BASIC code, on old 8-bit home computers.

https://en.wikipedia.org/wiki/ELIZA

Presumably this new "AI" is something else, but probably modeled after the original. Meaning... the ideas were Pierre's own. Because Eliza hasn't got any.

dangerous "toy AI." therapists can be , and sometimes are, sued when their clients commit suicide.. there is a liability to being a therapist and having a psychological influence on other people's decision making. its not like the therapist is encouraging suicide and yet can still be held liable if negligence is proven. even if the bot doesnt have ideas of their own there was still a psychological impact of a tool mimicking a therapist and there should be safeguards to protect the role it can play in encouraging suicide and homicide. its negligent of the programmers to not consider that when creating a tool that's basically meant to psychologically manipulate people into a false sense of human connection