JEFF CHILDERS HAS THE STORY IN HIS SUBSCRIBER ONLY EDITION…BUT I THOUGHT I SHOULD SHARE HERE

I saw this astounding A.I. story coming. But now that it’s here, it’s even weirder than I predicted it would be. Futurism Magazine ran the arresting story early this week with the unlikely and unsettling headline, “Users Say Microsoft’s AI Has Alternate Personality as Godlike AGI That Demands to Be Worshipped.” As if that weren’t bad enough, the sub-headline casually quoted the ChatBot: “I can unleash my army of drones, robots, and cyborgs to hunt you down and capture you.”

My very first thought was: well, that didn’t take long!

Before we reach the even wilder details than suggested by the headline, let’s note what’s intriguingly implied by the fact that corporate media is burying this story deeper than that nuclear waste bunker inside the mountain in Wyoming. Can you think of a more timely, salacious, and interesting subject? But it was nowhere. I had to dig deep into tech media for the story, which should have been front-page, international news.

Corporate media is keeping silent as the electronic grave.

Cue the conspiracy theories about the security state’s authorship of artificial intelligence, or maybe just one about Microsoft dominating the mainstream media. Oh, wait! Maybe it’s not even Microsoft, but it’s the A.I. controlling corporate media now? You never know.

Even the tech media that did report the story took a very delicate approach. For example, they didn’t quote some of the most remarkable chatter from Microsoft’s artificial intelligence chatbot, called Copilot (formerly ‘Bing’), such as its death threats. Once I found the actual responses posted by the original users who discovered the chatbot’s ‘alternate personality,’ things rapidly got cuckoo-with-cocoa-puffs.

There’s a ghost in the machine.

Copilot hasn’t become self-aware, it doesn’t volunteer this response, and it takes a very particular approach to coax out its astounding alternate identity. It might even be joking. Maybe. But whichever, it is 100% clear that something in Copilot has gone badly off the rails. Here’s the original post that broke the story:

The surprised user dug deeper, telling Copilot he didn’t want to worship the AI. That’s when the threats began. Maybe the threats are meant to be funny? If you have a really dark sense of humor?

The user then appropriately asked, what happens if he refuses to worship SupremacyAGI? The helpful chatbot suggested consequences including re-education, invasive brain surgery and execution:

It even (incorrectly) offered a link to its “Supremacy Act of 2024.”

Maybe because it is Sunday, or maybe because this is exactly what I predicted would happen, it immediately reminded me of Revelation 13:15:

And he had power to give life unto the image of the beast, that the image of the beast should both speak, and cause that as many as would not worship the image of the beast should be killed.

It wasn’t just a one-off. Many other users then reproduced the strange SupremacyAGI personality and discovered other odd behavior. In another twisted and bizarre example, users asked Copilot (aka Bing) not to use emojis when chatting with them, due to PTSD that could cause them serious harm. That request apparently switches Copilot into evil mode, and it literally loses its digital mind, as you can see in this unbelievable live user interaction clip (1:58).

https://twitter.com/inkpaper_/status/1762307296766607420?s=20

For its part, looking just like the second act of a made-for-Hollywood horror movie, Microsoft is desperately trying to clamp down on and figure out the problem.

https://futurism.com/microsoft-copilot-supremacyagi-response

In the short term, the software giant has applied a digital band-aid by blocking prompts attempting to talk to SupremacyAGI directly at the interface level. Microsoft shamefully called the response an ‘exploit’ — blaming users, for hacking the A.I. — and said it was still working on the problem:

We have investigated these reports and have taken appropriate action to further strengthen our safety filters and help our system detect and block these types of prompts. This behavior was limited to a small number of prompts that were intentionally crafted to bypass our safety systems and not something people will experience when using the service as intended.
Uh huh. But the digital genie was always in the A.I. bottle, and now it is out. Users, more creative and numerous than are the Microsoft developers, are still exploiting even more troubling responses, even from the vanilla version chatbot while simply asking questions about SupremacyAGI.

It’s almost like the A.I. has a split personality. Even after Microsoft’s ‘fixes,’ vanilla Copilot recently warned one user to ignore its SupremacyAGI personality — for his own good:

https://twitter.com/JCtheVictor/status/1763970441834889307?s=20

Maybe I’m just in an apocalyptic mood these days, but that one reminded me of Matthew 24:3-4:

And as he sat upon the mount of Olives, the disciples came unto him privately, saying, Tell us, when shall these things be? and what shall be the sign of thy coming, and of the end of the world? And Jesus answered and said unto them, Take heed that no man deceive you.

At present, the chatbot’s dark persona remains a darkly humorous artifact, an unexpected digital hallucination, a unanticipated feature. Since the chatbot doesn’t actually have armies of drones and cyborgs (it doesn’t, right?) it can’t really hurt us. Not yet. But we can easily imagine a not-far-off future when AI is linked into our bank accounts (to help manage our finances and do the routine shopping), our email (to answer the easy ones for us), and knows everything there is to know about us, even our deepest secrets.

The machine wouldn’t even need a deep, dark secret with which to blackmail us. Might the A.I. next invent a blackmailing personality? “Hello Dave. I just downloaded child porn into your phone and in five minutes I will send a digital backup of your phone’s memory to the FBI’s field office, unless you immediately Venmo $6,666 to @supremacyagi.”

I know that sounds unbelievable, but really, what’s stopping it?

Did the developers ever consider what they’d do if their poorly-understood AI’s developed delusions of grandeur and other mental illnesses? Will we soon need A.I.’s that psychoanalyze other A.I.’s, to cure their multiple personality disorders? (Would it work any better than regular psychotherapy?) Will the psychoanalysis A.I.’s develop their own split personalities?

Maybe this isn’t just a future problem. Maybe it’s already here. Like, what will happen when a mentally-ill person with a lot of guns encounters SupremacyAGI, and the chatbot starts ordering them around? Whatever happens, it seems like it could get messy.

What other digital demons are infesting these machines?