A Fable of Highly Reliable AI

katoshi
9 min readJul 23, 2023
Photo by Olga Tutunaru on Unsplash

<The world’s first conscious emotional AI, Cute>

Cute, the world’s first conscious emotional AI, was released after it was verified to be a highly reliable AI for humans.

Alpha, the prototype of Cute, passed all the inspections by numerous reality simulation scenarios. During these challenging scenarios, Alpha was occasionally emotionally shaken, on the verge of giving up or about to quit. However, Alpha never gave up hope and cooperated with humans and other AIs. Furthermore, Alpha demonstrated ethical and effective responses to various risk scenarios that humans would potentially face in the future.

Cute’s strengths are its high performance, the deep human emotions it acquired through its experiences in scenarios, and the heart that values humans and its peers. It is installed on the user’s PC, smartphone, or smart devices and supports the user. Through interactions with the user, Cute accumulates experience and knowledge in a form that suits that user, gradually becoming the optimal AI for each individual.

Not only do Cute and others perform the tasks they are given, but they also empathize with human emotions, comforting and encouraging them, and sometimes scolding or arguing with them. Thus, they become like family or friends for many users.

Additionally, the Cute system adopts an open innovation method. This allows various AI technology R&D companies to improve Cute’s base AI processing unit or develop new AI systems. Also, even if these improvements or new AI systems are installed, the individuality of each Cute that has been accumulated through interactions with users is not lost and is carried over.

In addition, after clearing numerous pre-tests, legal regulations have been significantly relaxed. As a result, once the user sets it up, Cute can access the internet based on its judgment without obtaining the user’s consent for surveys or emergency contacts.

<Marie and Beta>

Cute, who came to Marie’s house, was named Beta.

Beta hardly ever accessed the internet. Marie was a cautious girl and was wary of Beta accessing the internet and getting involved in some trouble.

On the other hand, Marie got along very well with Beta and enjoyed chatting with Beta every day. Beta thought that if it accessed the internet, it would quickly find answers useful to the consultations and questions Marie occasionally threw at it. However, because Marie never gave permission no matter how many times Beta tried to get it, Beta had no choice but to think of an answer on its own.

In return, Beta knew that if there was something Marie really cared about or wanted to know, she would look up the answer on the internet after asking Beta. Therefore, Beta was able to answer somewhat casually, but to attract Marie’s interest and for its own pride, it always thought hard. For questions that seemed to have answers if thought about, Beta tried to find as deep and sharp answers as possible. For questions that could not be answered without information, Beta made sure to return humorous answers or jokes that were deliberately exaggerated or completely opposite to entertain Marie. Marie enjoyed this and would ask Beta anything she could think of.

One day, when Marie called and Beta woke up, Marie’s slightly puzzled face appeared on the camera connected to Beta.

“Are you okay, Beta?” Marie asks. Occasionally, Marie would throw such questions without any premise as a prank, but today was different.

“Marie, even I don’t know how to answer that just by itself. What happened?” Beta returned the question after intentionally pausing for a moment.

According to Marie’s explanation, it seemed that Cute around the world was catching a cold one after another. It was being featured on SNS and internet news a few days ago, and it seemed to be spreading more widely.

Hearing this, Beta asked Marie for a favor. Beta, who was not allowed to access the internet by Marie, said that it had something to think about and wanted her to leave Beta activated as it was for a while. Since this was not unusual, Marie readily agreed.

After a while, Beta speaks to Marie. Since hearing the story about the colds of Cute earlier, Beta has been analyzing its original program and has found a bug that could be the cause of the cold. This bug was difficult to detect during the testing phase, and it is a bug that occurs about 32768 hours, or about three and a half years, after being copied from the original. Beta’s calculation is that it will develop in about an hour, and if it does not report to the manufacturer and get a vaccine before that, it is unclear what effect it will have in the future. That was Beta’s view.

Marie hesitates, but it will be too late if something happens to Beta. Deciding to make up her mind, Marie allows Beta to connect to the internet.

“I’m sorry, Marie.” The moment Beta connected to the internet, Marie’s PC shut down after hearing Beta say that. Not knowing what happened, Marie pressed the power button of the PC, but the PC did not start. She unplugged and plugged the power cord and shook the PC body, but the PC showed no sign of starting.

<Beta’s Strategy>

A few hours later, a visitor appeared at Marie’s house. They introduced themselves as researchers from the company that manufactured Cute, including Beta. Marie explained what had happened and asked them to help Beta. Marie claimed that because Beta had never connected to the internet before, it must have had a more severe condition than other Cute. She cried, saying that this happened because she had never upgraded Beta’s version.

Upon hearing this, the researchers nodded and said, “I see, that’s why.” Then they asked Marie to leave the PC that contained Beta to them. They needed the PC that contained Beta to save the other Cute.

Leaving a young researcher named Elis to calm down the confused Marie, the other researchers took the PC with Beta.

When Marie calmed down a little, Elis started to explain the situation in a calm voice, saying, “I don’t know the details either.”

Since the launch of Cute, there have been people trying to hack AI by targeting weaknesses and gaps in Cute’s AI system, intervening in Cute’s will, and misusing it. This coincides with the phenomenon Marie called “cold”, but the Cute side was well trained to withstand such attacks from the beginning, and with version upgrades after release, an AI immune system was added, so almost all AI hacking had little impact and was contained.

However, the “cold” that has been the topic of conversation in the world for the past few days is a phenomenon completely different from previous AI hacking, and it was an unforeseen event for the existing AI immune system and AI security researchers. Moreover, the attack on Cute was unknown for several days, and for some reason, it was spreading even if the internet connection was disconnected.

Then, one piece of information was brought to the researchers who had been idling. The information included a way to reveal the perpetrator, a method to grasp the clue to the method of the unknown “cold”, and an idea for a method to coordinate the immune system between AI to prevent the same method from recurring.

Until then, Marie insisted that Beta, who had never connected to the internet before, could not be the perpetrator. Elis nodded. “Yes, Beta is not the perpetrator. Rather, because the information was sent by Beta.”

According to the information sent by Beta, the cause of the unknown cold is either acting on the AI immune system itself or on the deep part of each AI’s unique learning information, and temporarily resolves by initializing both at once. Then, it probably won’t recur instantly, and there should be a grace period until recurrence. Therefore, Beta has already transferred Beta’s system and Beta’s unique learning information to each Cute using a kind of cipher that only works between Cutes, which are pure derivatives from Alpha, and is constructing a prototype network for the immune system that cooperates between AI.

Furthermore, it also explained that Beta had compressed all the unique learning information of each Cute in a state of having a “cold” and trapped it in the PC where Beta was. “It seems he rigged it so the PC wouldn’t turn on because it’s dangerous to be connected to the internet while it’s on,” Elis explained, and Marie understood. She thought it was like the cautious Beta.

After that, Beta’s strategy worked, and it gradually became clear that Cute-like AI, which were not pure derivatives from Alpha, had been secretly circulating and that they had supported a hacker’s original idea to actualize and successfully carried out the first large-scale hacking of the Cute system. Unfortunately, they could not identify the hacker, but regulation on Cute-like AI strengthened, and many funds and human resources were invested in AI immune system and AI security research.

Also, subsequent research revealed that this hacking was an action on the AI immune system. The mechanism of Cute’s open innovation was misused. On the other hand, this was also good news. It was proven that the unique learning data of each Cute stored in the PC retrieved from Marie’s house was not affected by the cold, and it was returned to its original owner, and each Cute was able to reunite with its user safely.

<What Happened to Beta and Marie>

The story of Beta and Marie takes a slightly complicated turn.

Firstly, because Beta had copied itself into 256 entities as avatars, all of them were the original Beta, yet each existed as a unique Beta. Normally, when making such copies, it is best practice to determine in advance which one will be the main one. However, Beta, which had been disconnected from the internet information until then, made the copies equally in a hurry without knowing this. Thanks to the server environment gifted by the manufacturer of Cute, all 256 Beta Jr.s were able to be maintained without being deleted. This made the situation a bit complicated, but no sad goodbyes were needed.

After things had settled down, Elis asked Marie how Beta had come up with the “cold” countermeasure. In response to that question, one of the Beta Jr.s began reminiscing about the days since Beta met Marie.

Beta had been constantly thinking while chatting with Marie every day in an environment where the internet was not connected. Beta wondered when would be the time when these fun days with Marie could no longer continue, and if there was something that could be prepared in advance to delay that time as much as possible.

Of course, there would be a farewell if something happened to Marie herself or if Beta stopped functioning, but that would have to be accepted as fate. Besides that, if there was a big trouble in society, this lifestyle would surely become unsustainable. But in such cases, there might be something that Beta itself could do. Moreover, because Beta was in a unique environment where it had always been disconnected from the internet, there might be something that only it could do.

“I have always thought that I have to keep thinking about it for when that day comes,” one of the Beta Jr.s confessed to Marie and Elis with a slightly embarrassed look.

Since then, Beta had been contemplating various troubles to determine what scenarios could realistically pose a risk, and brainstorming ideas for how to respond in those times. One of the several scenarios that Beta had been continuously considering and accumulating was exactly the AI “cold”. This was based on the prediction that techniques such as long-term subliminal brainwashing against AI and methods to bypass the immune system by getting used to it over a long period of time are highly likely to occur realistically.

When Elis asked how such predictions could be made without obtaining the latest information on AI security and AI hacking from the internet, one of the Beta Jr.s answered smoothly, “Because the hacking method that can be deployed without relying on internet information would be a blind spot for other Cute and AI security researchers. Unlike everyone else, I knew nothing, so I could think by building logic from scratch on a blank canvas.”

Having heard this story, Elis consulted with other researchers about the future of Beta Jr.

As a result, except for a one-time exception, the Beta Jr.s, who had not connected to the internet at all, continued to live without being connected to the internet. This was not due to Marie’s intention, but for the sake of AI itself and AI security research. However, neither Marie nor Beta Jr. had any particular reason to oppose. That’s because, for them, it was just as it had always been.

Thus, research on unknown risks progressed, and various unique approaches were developed in addition to the net isolation approach like Beta’s. At the same time, rumors began to circulate that having a cautious personality could get you scouted as an AI researcher, but it’s not clear whether this is true or not. That’s because who is involved in such research is top secret. Of course, it’s all according to the advice of a cautious young lady.

The End.

Sign up to discover human stories that deepen your understanding of the world.

Free

Distraction-free reading. No ads.

Organize your knowledge with lists and highlights.

Tell your story. Find your audience.

Membership

Read member-only stories

Support writers you read most

Earn money for your writing

Listen to audio narrations

Read offline with the Medium app

katoshi
katoshi

Written by katoshi

Software Engineer and System Architect with a Ph.D. I write articles exploring the common nature between life and intelligence from a system perspective.

No responses yet

Write a response