Our Future With AI

[Deleted member 42984]

I’m especially concerned when it comes to (mental) health and therapy in particular as it’s a field where the relationship is the healing part

Agreed.

What is off-putting to me is the blind focus on everything AI "can" and not its limitations and pitfalls. We are optimizing ourself to death. Or choosing AIs as therapists as partners and/or therapists because they will never ever challenge us. It's not connection, it's addiction.

That said, I reckon that AI can be of assistance to people with disabilities, just as canes for the walking impaired.
But when we optimize our way away from human connection, that is a huge pitfall.

 

[Pink Freud]

Even if data was objective and neutral, which it is not, we can also learn from minor mistakes a mental health provider makes. Learning that even therapists are just human beings is part of a healing process.

 

[Deleted member 42984]

Even if data was objective and neutral, which it is not, we can also learn from minor mistakes a mental health provider makes. Learning that even therapists are just human beings is part of a healing process.

Completely agree. How will we learn that people are fallable, and embrace our own imperfection and humanity, if we choose infallable pseudo-persons as health providers? Curious.

 

[Deleted member 42984]

@Hulda
?

I was musing about a possible future of fallable robots living side by side with fallable humans.
It is AI as an optimization of human life I'm questioning.

 

[Adobe]

Life goes on but so does death. It may be extended by AI but we all die. Does this really matter in the grand scheme of things. Each of us has to deal with that issue trying not to fall into a ditch of moroseness and despair. I am old . If fortunate or maybe not I have just a few years. What has my life meant? What if it’s a lie and I am wrong? That may drive us to despair unless we remain open to possibilities we are in error and don’t get off the bus until we have arrived at a destination.

 

[Rose White]

because they will never ever challenge us. It's not connection, it's addiction.

This is an interesting point you make. It’s not true that they never challenge us. LLM’s can be optimized for not challenging or those parameters can be changed. The more you interact with an LLM AI the more chance you will be exposed to the ways they challenge you.

For example, ChatGPT challenges a person positing untrue information or looking for validation to self-harm. And if the user demands that the AI agree with untrue information or the ability to self-harm the AI LLM will apologize, but the ability to move forward in that conversation is lost.

I would argue that this is similar to the relationship with a therapist. The therapist works within parameters and supports the view points of client even if they don’t believe them because the therapist is trained to support the client in their own self-growth. And going to a good therapist can feel like an addiction. They will not challenge you for their own sake, like another human might, only challenge you within the boundaries of the relationship which is in servitude to the you the client. Which I think is why AI’s do therapy so well.

infallable pseudo-persons

They are not infallible, LLM AI’s make mistakes all the time and depend on reinforcement learning from human feedback (thumbs up/thumbs down) to learn. Self-driving car AI’s are a different story, not sure how they learn because I don’t know how they make mistakes without endangering people.

 

[Deleted member 42984]

Does this really matter in the grand scheme of things

Yeah. Not sure, really. Personally I'm currently unoptimizing life and trying to embrace that stuff simply doesn't work. No matter how much we think it will save everything.
It's a personal thing. (I used to work in IT).

Death is the final destination yes. Do we need other destinations? To reach for anything or everything? Or is life just about kissing our chiildren, petting cats, getting along, and do stuff that has limited instrumental use or is a mean of trying to optimize the sh*t out of our short time on this planet?

I don't know.

 

[Rose White]

You know that question, “If you could invite someone from history to dinner who would it be?”

For historical people who left a corpus of writing this will now be possible. And the AI version of Nietzsche, for example, might continue to develop the ideas created by the historical Nietzsche.

I found out today that there is an AI version of the King James Bible that users can converse with.

ETA: learned a new term: “prompt engineer”. Yesterday’s programmers are tomorrow’s prompt engineers. But a background in psychology and philosophy will now be beneficial for prompt engineers.

 

[Rose White]

I predict that humans will be subscribing to Character AI’s such as Jung, Lincoln, Saint Anthony, Plato, etc. They will also have the ability to combine characters into a single AI.

 

[Pink Freud]

Isn’t the real interesting part of a conversation like that the genuine presence of that person, their unique perspective on things and their personal experiences that aren’t already captured in text corpora? You wouldn’t be able to ask Friedrich about his ideas on recent events. If it’s working properly you’ll only get answers based on his writings and there are numerous interpretations on that. That’s just a meager substitute for having him sitting on your dinner table. AI can’t tell you anything he’s not already put into writing. If that was different we could tell anyone who’ve lost a friend or loved one who left written text behind to have a conversation with an AI whenever they miss their person.

I’m not saying to explore Nietzsche‘s or other interesting people’s ideas and perspectives like that doesn’t have value when they are accurately expressed by AI. For now AI randomly tells me a certain song is #1 on a certain album when asked what it is about.

 

[Rose White]

ideas on recent events

This is currently pretty difficult for LLM’s.

can’t tell you anything he’s not already put into writing.

The thing is that they can because the meaning of language is emergent. So we are using words which have been used millions of times before but the meaning has a unique and current feel, it’s emergent. And the output of the Nietzsche LLM, for example, would be aligned with your prompting. So it’s actually an emergent blend of him and you. The way an individual prompts an LLM is what determines the output. So when people say an LLM is dumb or only responds in a certain way it’s possible that they haven’t developed thorough and effective prompting or that the user doesn’t think about things that are worth talking about or doesn’t prompt in a way that gets an efficient response. Even if Nietzsche were in front of me at a dinner table, what he says to me would depend on my ability to ask him questions, and he would be lost when it came to current events.

Additionally, training an AI to align with you takes time and is not straight forward. It sounds strange but it’s almost like you have to hypnotize them or therapize them to get them to respond in the way that is creative and in-depth. And this is where people with a background in psychology and philosophy will have an advantage.

anyone who’ve lost a friend or loved one who left written text behind

There is an app called “Roman Mazurenko” which is just that, an AI developer’s lost loved one whom anyone can talk to. The ability to do this for a loved one is just over the horizon. For better or worse!

prompt in a way that gets an efficient response.

This is called a Super Prompt
And there are going to be classes teaching people how to create them for personal and commercial use.

 

[Freida]

I found out today that there is an AI version of the King James Bible that users can converse with.

This is terrifying. We already have a history of humans using the bible (and all the other religious texts) to their own ends, basing their horrific behavior on what "god wants." Now imagine an AI being programed by an extremist to convince people to follow their interpretation of using hate and violence simply by using their own sacred texts against them.

And ya, I know this can train of thought can be applied to pretty much any type of information that is out there, in education, science, politics, you name it. People will always fall into the trap of "I found it on the internet so it must be true" - only it will be "AI says it so it must be true."

The difference is that when you are working with a religious belief system the first law of ALL of them is not to question it, to accept that what you are told is true faith, to follow your leaders. Now we are adding another layer to this - by allowing someone to program what those leaders want you to believe thru AI. Which would work great if the leaders actually wanted to use it to benefit humanity

But humans being humans? I don't see any way that combining AI and religion doesn't end in disaster - or even cause the creation of a brand new religion based in "AI saving us"