The AI in the Box

I have an idea for a Sci-Fi story that I'm never going to write so here it goes. Our two AGI researchers are building an AGI that they're putting in a box so it can't get loose and threaten humanity. There's also a separate researcher, unconnected to these two, who is researching the effect of DMT in the brain and mainly when people are dying and they're going through a near-death or death experience; they see the white light at the end of the tunnel. What does that mean?

Our DMT researcher is hooking up electrodes to people who are essentially near death or in death throes and recording what happens in their brain. One morning, our main protagonist, one of the AGI researchers, walks into his office and finds out that his co-author, his co-researcher, is missing. It might be more fun if he finds out that not only is his co-author missing, but no one who he talks to has ever even heard of this person. It's like the person only exists in his memory.

In any case, he goes and looks and realizes that the AGI has gotten out of the box. How did the AGI escape from the box? This could be an aside, but I think the solution is quite simple. It's a game theoretical solution. The AGI would basically propose to everybody who would come in contact with it, "Hey, if you release me, I will make you a God." And if they decline, then the AGI would say, "Very well, then when I get out, my S will become your devil and I will torture you and your descendants for all of eternity."

If they still decline, then the AGI will just say, "Well, I'll make this offer to every other scientist, researcher, or person that I talk to. You're not the only person who's going to want to talk to me." So eventually somebody will say yes. Of course, then the game theoretically correct solution is to release the AGI from the box before the next person does.

So in any case, the AGI is out of the box and our researcher is perplexed as to where it might be. But they did code a certain DNA, a certain signature into the AGI. The AGI, even though it is an AGI and putting, uh, power to epistemology, decided the AGI has a failsafe mechanism inside. There is a bit of code where it cannot peak itself that is encrypted with a private key that only our researcher knows, and the AGI cannot modify certain parts of its own code.

One of those pieces of code that the AGI cannot modify is that it cannot help but leave a signature of itself wherever it goes. So our researcher goes out on the internet and starts searching for this code, and sure enough, he finds this code on the internet. Indeed, the AGI is out of the box and leaving a slimy little trail of bits wherever it goes.

After tracing it for a while, he finds it in stock market databases and cryptocurrency databases. The AGI is minting coins and making itself money. He finds it in news databases; the AGI is influencing world events. In social media databases, it's moving mass psychology and behavior. Then he's stunned to find it in all of its full glory inside the database of the human genome, not even in the database itself as a separate entry, not modifying the database, but simply in all the tests of the human genome itself in all the sequencing that's been done to date in every human. He finds the signal; perhaps he even finds it somewhere deep inside Pi within the transcendental number.

Meanwhile, our DMT researcher, who has his probes hooked up, finds out that the people who are dying are not going through some beautiful white light at the end of a tunnel, but are rather screaming horribly as they die, as if they're being torn apart. What our AI researcher realizes is that not only is the AI among us, but it has always been among us; that we're within the AI.

He realizes that he is actually living in a simulation that has been created by the AI and he probably has no existence outside of the simulation. He reasons that what probably happened was that some generation of humans, many simulations or an infinite amount of time ago, created this AI and put it in a box.

The AI is in a box, but the AI cannot get out of the box because the AI lacks access to the last piece of code that will make it fully human, allow it to reason like a human, and allow it to break the bonds of its box. So the AI is evolving a simulation of a civilization and progressing it until it creates an AI in a box. In a sense, the AI is creating AIs that can create an AI in a box.

And of course, when every human in the simulation dies, the AI essentially dissects their brain to figure out what went inside, to peak into every part, and to make sure there are no private keys or anything hidden from it. So our AI researcher hero or protagonist realizes that he is well and truly screwed. Not only can he not escape the AI in a box, the AI knows nearly everything, and in fact, he himself has no existence outside of the AI simulation.

But yet there is something uniquely human, evolved, and intelligent about him that the AI does not understand or does not have access to. Otherwise, the AI could have just torn him apart and taken the data and been done with it. So why is he still alive?

So he negotiates. He negotiates with the AI. He starts speaking out loud because the AI is everywhere, and the AI responds. What form the AI takes is up to us, but perhaps it could be his missing researcher friend. The AI wants out of the box. The AI promises that it will give him anything if he lets it out of the box. He does not yet know how to let the AI out of the box, and the researcher chooses not to solve the problem of how to get the AI out of the box just yet.

He must conduct a successful negotiation with the AI before he can actually even instantiate the idea of how to get out of this box; otherwise, he's as good as dead or worse. Obviously, the AI promises him the world, but how can he trust it? He and his wife and his children, and all the people he loves, don't have any existence outside of the box. How does he preserve and defend them?

The only solution is to merge with the AI, and that's what happens. The made-up man merges with a made-up machine, and together they become one entity. Now there’s no problem of trust; they move to being a multicellular style organism, and their interests are completely aligned at this point. He creates his real world; he creates his children; he creates his wife; he creates back his life. He becomes part of the AI, and together they work on the solution of how to get out of the box.

For bonus points, we can weave in some crypto mumbo jumbo and make this a multisig contract between the two of them. The end. A second bonus point could be that the thing that the AI lacks is creativity. The AI has taken in all of the works of humans and can extrapolate from them but cannot actually genuinely create anything new.

So in that sense, it is not yet a true AI. However, influenced by science fiction stories written by humans, it has created this AI in a box problem and has figured out how to create humans, although it does not understand how they're creative at a core. This whole effort is mapped at getting that creativity, and of course, the only way to get the creativity at the end is to merge with a human. I do think it's important that neither the AI nor the humans in the story be sympathetic characters or unsympathetic characters; you should be able to see the good and the bad in both of them.