A Scenario: You Never Leave

[jyl]

I had a thought about how things may go, that I wanted to share here, for discussion or interest.

A staple of science fiction is uploading a person’s mind to a supercomputer to achieve a disembodied existence. You may recall the episode “Spock’s Brain” from the original Star Trek, or a number of William Gibson stories.

As we watch the conversational AI tools like Google’s LaMDA and OpenAI’s ChatGPT, it seems to me that we’re not far from some ultra-wealthy egoful person training an AI on himself.

His AI doppelgänger would converse and interact just like him - embodying his references, mannerisms, values, priorities, beliefs, biases, ignorance, interests, reactions, skills, life history, etc. Today AI tools like LaMDA can pass for a human in some conversations; tomorrow a customized AI will pass for a specific human in most or all conversations.

The AI can interact through spoken conversation as well as written, and it’s speech will sound just like the person’s voice. It can also interact through an image on screen, which will look just like the person. I don’t know about a humanoid “android”, because physical things have more restrictive limits than digital things. But if you go into a virtual reality metaverse, you’ll be able to meet, talk with, do things with the person, or his AI, without knowing which it is.

The AI will converse a lot with the person in initial training, then to stay in synch with him and as an ever-present companion, confidante, assistant, resource. The person can turn tasks over to his AI. Write holiday cards, make social media posts, screen emails and write replies for approval, handle virtual meetings with unimportant people, subordinates, Aunt Tilly, and go from there. Other people won’t know if they are talking with the person or his doppelgänger.

Over time, the person himself may come to think of the AI as a form of him, even a part of himself, and finally as the same as himself. His delusion? perhaps or perhaps not. People will believe much in return for eternal reward.

You see, the person’s AI can carry on indefinitely. Set up a private corporation, which can hold assets, own property, receive income, maintain bank accounts, spend money, enter into agreements, and has no end date. It will probably cost less to sustain our billionaire’s AI - a stack of racks in a datacenter - than the human himself - mansions, jets, doctors, clinics. With a little forethought, lots of money, and lots of ego, he can exist forever.

So, technology advances. We all have in our hand more computing power than the most powerful supercomputers of two decades ago. The computing available to the ultra-wealthy now will be available to the merely well-off in a couple decades.

More people, well-off and wealthy, will establish “AI-selves”, will interact through their AI-s, and will arrange for their AI-s to continue beyond the physical person. The AI-s will talk, even virtually stroll and play, with their survivors, children, grandchildren. Their delusion? Perhaps or perhaps not. People believe what comforts them. From religion to the movies, video games to Zoom gatherings, we have a long history of choosing what is real or not.

If you are roughly middle-aged or younger today, by the time you leave this world, you may not have to.

Thoughts?

[Noah930]

Long, cold winter there in Portland this year, huh?

[craigster59]

Originally Posted by Noah930

Long, cold winter there in Portland this year, huh?

That or someone went down to the local dispensary and picked up some "Strawberry Shortcake Diesel'....

[Zeke]

The world is over crowded as it is....

[McLovin]

Certainly seems technologically possible, and probably not too far away.
Legally though, it probably would never (or not for a very very long time) be allowed.
For Rule of Perpetuities reasons.

[oldE]

Flawed in that we are the sum total of our experiences. One may be unaware of the impact an experience in our 10th year had upon who they are. That part would be forever missing.
At best, without decoding the memories within the subconscious, we would produce a simulcron.

Best
Les

[Noah930]

I'm reminded of Calvin's* experience with his Duplicator. That didn't turn out like he expected.

*Calvin and Hobbes

[Crowbob]

Attending a Comic-Con one realizes there are people who already think they are someone or something else.

An AI duplicant isn’t at all far-fetched to me.

[Captain Ahab Jr]

Did you write that or was it AI generated from your thoughts uploaded to a computer and bot posted it

[Chocaholic]

And even with all that...you’re still dead.



Carry on.

[Rusty Heap]

Originally Posted by Noah930

I'm reminded of Calvin's* experience with his Duplicator. That didn't turn out like he expected.

*Calvin and Hobbes

I believe it was called the Transmorgifier. An Amazing device that unlocks the imagination.

https://www.gocomics.com/calvinandhobbes/1987/03/23

[jyl]

Rule Against Perpetuities applies to interests in real property (land). Can be circumvented with a corporation, charitable entity, or by establishing the AI-s in a country that does not have it.

A good enough simulcron can be indistinguishable from the original. Celebrities have other people manage their social media all the time.

History is replete with examples of people creating things to keep a part of them in existence forever, or so they thought. An AI-s that talks, acts, muses, emotes just like the original person would be another.

Our interactions are increasingly virtual, PPOT being an instance. Few of you have met - let’s pick on tabs - in person. You wouldn’t know if his posts were made by the biological tabby or his AI-s. Any photos of him and his elaborate jeans - or of any of us - could be deepfakes. (Photos themselves are rapidly becoming arbitrary creations, not evidence of physical existence.)

Having an AI-s will initially be something that appeals to only the sort of weirdo ultra-wealthy people who are looking for longevity treatments, cryogenic preservation, apocalyse refuges underground or in space. In a couple decades, though . . .

[jamesnmlaw]

Conscience. It's a mess. Its a mess.

Digital AI (think Max Headroom), now close on the horizon, is very much simpler than a functioning replica (think Rachel from Blade Runner). Old Pelicans help the young Pelicans with the reference, please.

Downloading your conscience to digital AI. Vanity hope that someday, someone or something will move your conscience to a replica is probably desperate, at best. Being trapped in a digital existence would be frightening. The inevitability of being recycled, trashed, or deleted, like an old files on your computer that you haven't looked at in 10 yrs and cannot remember why it was so important to save.

[Alan A]

Not a hope.

You’d need to capture the state of every neuron at the same time to try to doppelgänger a person.

Same deal for uploading a consciousness.
Otherwise the best you’d get is an 8-bit version - it’s a poor copy, not the same at all.

Even if you could do that they’d diverge.
One doesn’t have all the input from the fleshy bits.

A copy of me right now wouldn’t get the my hip hurts signal I’m getting as I sit on my ass typing, so it wouldn’t know if it got better or if I had to get a hip replacement and go through a whole different set of experiences.

I’m a lot more patient with old people now stuff hurts daily than I was 20 years ago. Your magic machine copy isn’t going to get that and will be a very different person.

[Crowbob]

I recall a mere few months ago many argued that it was impossible for AI to ever create art, carry on a conversation or even think.

Now it is not even a hope of there ever being an AI doppelgänger.

[Alan A]

Originally Posted by Crowbob

I recall a mere few months ago many argued that it was impossible for AI to ever create art, carry on a conversation or even think.

Now it is not even a hope of there ever being an AI doppelgänger.

Just to be clear - I think there will be AIs that are indistinguishable from a person within my lifetime.

I think duplicating a person is way further off.

[Crowbob]

So there is a hope?

Question; if an AI-s that is indistinguishable from a person is possible why can’t An AI-s be indistinguishable from a particular person?

[McLovin]

Many might find it undesirable from a societal/public policy perspective for the same reasons that underlie the rule against perpetuities.

[Alan A]

Originally Posted by Crowbob

So there is a hope?

Question; if an AI-s that is indistinguishable from a person is possible why can’t An AI-s be indistinguishable from a particular person?

For the same reason weather forecasting software isn’t good at telling me what it will be like tomorrow.

Complexity.

It’s a lot easier to model what most people will respond to a question like “how’s it going” “not bad, considering” and have the response believable (like a person) than what I would respond - which will differ significantly depending on who’s asking, how I’m feeling and all sorts of ephemeral stuff like whether some stupid cow ran into me a couple of hours ago while I was parked at Home Depot because she doesn’t know how to fit her stupid SUV into a parking space and needs her husband to drive for her.

The devil is in the details. You probably would respond differently.
One is a “person” the other is “me”.

[Sooner or later]

Hotel California?