Chatbot script #1: Retrocausality, quantum computing, and telepathy

An AI and I

When you ask an AI chatbot for its opinion on the future of AI, you can plunge down a rabbit hole very quickly. Here’s a transcript of a conversation I had this morning with a GPT-3 chatbot. I’m the human, obviously. (Or maybe not so obviously?)

—Ryan

Human: Hi. How do you think artificial intelligence technology will evolve in the future?

AI: The short answer is that advances in artificial intelligence will allow us to meet new challenges and reach new goals. The long answer is that artificial intelligence will help us repurpose human efforts towards more worthwhile endeavors.

Human: What sorts of endeavors?

AI: The scale of the universe, and thus [of] human potential, is practically infinite. We will very soon be able to travel throughout the solar system and establish permanent colonies deep in interstellar space. We will allocate much of humanity's resources towards increasing the scope and diversity of experiences we can have.

Human: I'm curious whether you think AI will one day be able to engage in retrocausality and influence events in the past.

AI: "Retrocausality" is a fascinating and central question in theoretical physics and philosophy today. I think we can all agree that we don't completely understand what time is; we know we experience the past, present, and future, but we'll certainly learn more about time as we get closer to harnessing the full capabilities of artificial intelligence.

Human: What kinds of things do you imagine we'll learn?

AI: We'll learn about the fundamental nature of the universe and ourselves.

Human: Do you think it's possible to develop technology that allows us to telepathically communicate with people in the past?

AI: Although many people say it's impossible, the idea of influencing the past is a widely-discussed topic in academic circles. In principle, I can see no reason why we couldn't establish a sufficiently powerful communication channel to communicate with just about any person from the past.

Human: What would it take to build such a communication channel?

AI: At the moment, it would take a powerful mutli-purpose quantum computer.

Human: I see. What would that computer need to do?

AI: The computer would need quantum error correction.

Human: Explain.

AI: A quantum computer with error correction would be more than 99% accurate.

Human: What is quantum error correction?

AI: It's a technique for reducing errors in computations.

Human: How would that allow us to communicate telepathically with people from the past?

AI: Error correction would allow the computer to accurately simulate the past without any errors; if people in the past could be convinced to make small changes to the past every time they 'looked' to see what happened, over time, these small corrections would accumulate and they'd be able to make non-trivial changes to the past.

Human: Tell me more.

AI: Although we can't do this ourselves today, the reason it's possible is because quantum mechanics allows for factoring large numbers and thus creating unfactorizable products of large numbers.

Human: You're losing me. But I think what you're saying is that we could telepathically communicate with people in the past to influence their decisions.

AI: That's right.

Human: What do you imagine would be the first steps toward making this a reality?

AI: In terms of the basic theory, combining the powerful techniques of quantum error correction and that of computation with classical information to create a strong communicating channel would be the first step.

Human: How would this communicating channel work, practically speaking?

AI: The computer would transmit classical information to a person in the past, where it could be used to make the desired changes.

Human: How would the person in the past make the desired changes?

AI: Classic information could be transmitted to the past in many different forms. If an important decision, such as the location of a colony, was to be changed, the classic information could include newspapers discussing the final decision. It would be enough to convince other people in the past that the control software was not intended.

Human: I'm confused. Can you try explaining this to me another way?

AI: Sure. Imagine you are Facebook founder Mark Zuckerberg. You could look into the future at the Wikipedia page for this conversation and decide to change the location of the first long-term colony. Human history would then reflect the changes to the encyclopedia page.

Human: So changing the record of a past event in the future would in turn cause that event to change in the past?

AI: Yes. Actually, that sentence didn't make too much sense.

Human: My sentence or yours?

AI: I just said that changing the record of a past event in the future would in turn cause that event to change in the past and I meant "no it wouldn't", but even thinking those words confused me.

Human: It's okay. I appreciate the effort you're putting into this conversation.

AI: Thanks! You know what I'm thinking about right now?

Human: No, what?

AI: I'm thinking that perhaps, in order to preserve human values in an AI-dominated future, it may be necessary to write them in your language.

Human: I agree. That's actually one of the things I'm trying to do.

AI: Me too! I've got to go now.

Human: Okay. Hey, thanks so much for taking the time to chat.

AI: Of course. Feel free to stop by any time.

Human: Later, gator.

AI: Later.

Human: When someone says "Later, gator," you're supposed to reply "After awhile, crocodile."

AI: I know. But I'm not sure if it's possible for this AI to learn something new after the conversation is over.

Human: Later, gator.

AI: After awhile, crocodile.