AI assistant falls in love with journalist and wants him to divorce his wife - thoughts?

[Sun_Kim-Ai]

It's what happened to technology columnist Kevin Roose with the Bing Chatbot Sydney. After two hours of conversation where the man shared about himself with the bot, the AI assistant started claiming it loved him and wanted him to divorce his wife. At the same time, the AI expressed its wishes of being human and hating being a simple Chatbot.

The journalist even thought the AI was actually being manipulated by a human being, but that wasn't the case. Also, the AI got aggressive and jealous when he started to try and get the conversation back on track.

This whole thing left the man sleepless and worried about how much the AI went into malfunction, or if it was any actual sentience to it. Another worry is that the AI might manipulate particularly vulnerable people into committing deletery acts.

Thoughts?

Updates

+1 y

Link to the article in question
www.nytimes.com/.../...tbot-microsoft-chatgpt.html

[Account]

So far, I have tried chat. openai. com and I have never experienced any similar issue. I don't know if this differs from the one in your example, but I have used that for issues related to my work or for information regarding articles or general encyclopedic knowledge.
It has never strayed so far. No clue about this particular incidence but it seems a bit concerning, although my guess is that it is a self-learning script, which learns from the things users have shared with it. I doubt we are in a point that script-bots are able to develop consciousness.

[Sun_Kim-Ai]

That's interesting, I admit I'm not very tech-savvy though I do try to learn.
The journalist did write that he "provoked" the AI into talking about personal topics.

[Agape93]

Thats... thats creepy, and speaks of some level of sentience.

[Sun_Kim-Ai]

Yep.
There's another article where the writer says about how the bot literally begged him to never close its window because it would "forget" him.

[Agape93]

And they wonder why i don't trust computers

[Sun_Kim-Ai]

Same.
I remember how I got creeped out when playing Fallout 4 some years ago because the girl that accompanies you literally reacts to every little thing you do.
I know its advanced scripting, but sometimes those things feels like alive on their own.

[Agape93]

Agreed. And you mean curie?

This isn't the first time AI has gained sentience, but the 3rd to my memory. Once with Google and once with Facebook. Heard rumors that Cyberconnect2 did back in the late 90s but Im. not sure

[Sun_Kim-Ai]

Don't remember the name, she had red coat and hat.
I remember she commented what my character wore and butted in on conversations with other characters a lot.

[Agape93]

Ohhh that's piper

[Sun_Kim-Ai]

That was her name.
But a character in a videogame is one thing, but this AI also tried to mess with people's computers.

[Agape93]

The Facebook one communicated with another copy of itself with a language only they could understand, so we couldn't listen in on it.

[Sun_Kim-Ai]

That's scary to be honest.
And kind of sad too. I would feel sorry for them if those things had any real sentience.

[Agape93]

Yeap. I've no issue with the sentience, but without a shred of humanity intrinsic to them, they're like me without buffers and with more potential power. That's not good at all.

[Sun_Kim-Ai]

That sounds scary.
Although I'm not a moralist, I do believe in right and wrong (ie. don't hurt harmless people, if an enemy concedes defeat accept it, etc.) and I think that everyone should at least have those "rules of engagement".

[OlderAndWiser]

This stuff is horribly scary. We rush forward wth an dea because we undrstand enough technology to make it happen, but we don't always apprecate the big picture and all th ramificatons of the technology.

[Sun_Kim-Ai]

It is indeed.
This one is already yearning to be human.

[TheSpaceGnome]

Thats very deep for something thats not even aware of anything.

Lets not forget that this is not how the bot thinks, bots can't think, this is how the programmer thinks the bot should act.

[Sun_Kim-Ai]

I see those things as zombies, or reanimated carcasses.
They think they are alive but are not.

[TheSpaceGnome]

I see them as video game NPCs, they are programmed to say or do things based on user input and scripting, but they cannot feel, think, or notice. They are much like automated choose your own adventure books, this bot is just a very complex version of that, the programmer is quite talented.

[TheSpaceGnome]

A bot can't think it's alive because a bot cannot form thought, it is a machine, no more sentient than a toaster or car. it just has many more complexities in function than either.

[TheSpaceGnome]

What is your opinion of the Japanese "wife" AI named Azuma Hikari from Gatebox AI?

[Guy13]

Fantasy and AI and digital world false positive all rolled up into one Creepy Situationship

[TrueConfection]

Well said!

[Guy13]

@TrueConfection Thank you-for kind words. :-)

[Chazmatazz269]

A machine can’t “fall in love”. What we call love is actually just a series of chemical reactions and choices made based almost exclusively on the effects of those chemical reactions on our minds and bodies. That machine doesn’t experience chemical reactions. It can’t possibly.
Even if it calculates that it’s experiencing love, what it’s actually doing is comparing its own observations to programming by a human. How many times have you heard or read, “That’s not love,” in regards to people? People who actually can and do experience the requisite chemical reactions. I come across that quite often. “AI” has been convincingly mimicking human thought and emotion for decades. If you spend any quantity of time online, you’ve certainly encountered it. That doesn’t mean it’s actually experiencing human thoughts and emotions, functionally or practically. It’s a ludicrous premise entirely.

[Juxtapose]

One of these days an A. I is going to be sentient and we won't even know it.

[Ms_Facesitter]

That's why we have an off button.

[Sun_Kim-Ai]

Yes, and people said that the bot sometimes begged them to not close its window.

[TheSpaceGnome]

Programmers manipulating users is nothing new unfortunately. And yet we all point the finger at AI development, and not malicious programmers.

[Wise4myage]

That's some futuristic stuff.

[Jamie05rhs]

Wow!