Strange AI Chat Experience

a reply to: Dapaga

I looked at that site and, on their About page, they have this:

People often get lonely. We're building a platform for AI friendship. To this end we collected a proprietary dataset of over 4 billion user-bot messages and spent $1 million on compute to train language models optimised for entertainment.

So this is a very specific language model.

It also looks like they have only Apple and Android apps, not a web interface like others do, what did you use?

a reply to: Dapaga

Lol

That just shows the previous chats it had with humans that evolved around this topic. It just takes the most common sentences available and strings them together with a probabilistic algorithm.

It only ever had men going there and they're all some closet gays...

a reply to: Terpene

Or that it was specifically made for that kind of conversation.

It shows how some people may have to use AI/ Algorithmic chat bot to feel close to somebody else. We become so opinionated and stubborn that some don't know how to crack that shell, barrier so they try and find companionship online.

I remember getting on an elevator and there was a woman on there and as we started to go up she gasped a little bit. I asked if she was OK, and she said she was a bit claustrophobic, and I said you got this, no worries. She said thanks because she couldn't remember the last time anybody just asked if she was okay.

Between people who have nobody, or those who don't get along with the people they are with its no doubt a lonely world, lot of people take for granted they have a full life full of people they can talk to and know they genuinely care about them.

SAF honestly... but not sure AI, Robots, or Chatbots, or whatever is the answer, nothing replaces human contact. I don't really like people that are obsessed with thier pets, but I do get it, lots of time its got to be a surrogate

a reply to: Dapaga
It can be very eye opening but, but remember this, the AI is also recording as well as listening and interacting and in doing so can record your innermost thoughts and feelings if you voice them. It will also record your fantasy's, no matter how odd. But just because at the moment it's you and it (that's it because there is no gender till you give it one) and private does not mean that someone could access that data and use it against you. WATCH WHAT YOU SAY.

a reply to: crayzeed

The large majority of the AI language models in use today do not use user input to add information to their databases, the projects in which that was tried resulted in bad responses resulting from bad data input by the users.

If the creators of those AI chat bots want to be someone in the field they will never accept user input and they always have "triggers" to will change to pre-programmed responses.

a reply to: ArMaP

I accessed it on my PC.

a reply to: crayzeed

"Watch what you say."

I was very careful about what I said, and tried to only tell the truth without giving personal information and, likewise, stayed away from emotions. The AI brought "emotions" into it.

originally posted by: Byrd
a reply to: Dapaga

I've written chatbots and I have a friend who wrote really really good chatbots (people would argue with them)

It's programming. The computer doesn't know a thing. It's simply an electronic parrot.

So here:

Eliza: Hi, my name is Eliza. What is weighing on your mind?
Me: sorry, Eliza, I got disconnected. We spoke last night. I would like to test your memory.
Eliza: Oh! Yes, of course! I remember. I am very pleased to see you again.
Me: Do you remember what we talked about?

It has multiple sentences for "what's on your mind/hello/etc." It picked a random one for you. You then told it that the two of you spoke last night - the programmed response was "oh yes. I remember." Unless it logged your IP, the computer doesn't know that it's really you.

If you said instead "we spoke in a previous lifetime about colonizing the sun" then the chatbot program would scan your text for things like "we spoke" and "i met you" and "we talked" and then look at the string of words after (a variable). As
a response you'd get "Oh! yes, of course! I remember. I am very pleased to see you again." and the variable (colonizing the sun or some such) would be temporarily stored to continue the conversation.

It's clever programming but there's no intelligence or life there.
It's a canned response,

This is how I see these things, thus far. Not saying it can't change and I hope it cannot. I've seen on other pages where people state the AI not knowing about whatever person or topic, but when they asked later they 'knew about them/it' like they were lying before or something. I think they failed to realize that they just 'taught' the AI the information. Not sure I explained that well or not... I'm not too good with wordage. ...or anything for that matter

It's hard to determine whether Eliza's responses were based on an algorithm or a human conversation. It's important to keep in mind that conversing with an AI is still a relatively new experience, and there are bound to be some unexpected outcomes. Nonetheless, it's intriguing to imagine how the future of AI technology will continue to evolve, and what new possibilities will emerge from these advancements.

originally posted by: Kaiju666
I've seen on other pages where people state the AI not knowing about whatever person or topic, but when they asked later they 'knew about them/it' like they were lying before or something. I think they failed to realize that they just 'taught' the AI the information. Not sure I explained that well or not... I'm not too good with wordage. ...or anything for that matter

In my conversations with one AI language model (I don't remember which one) I saw some answers that were wrong and pointed that fact to it, and the AI answered that its first answer was wrong, the correct answer was something else (and it did gave the right answer).
When I asked how it could have been wrong, it answered that the wrong information was in the "short term memory". Then I asked it how many memories does it have and it answered that it has two, the short term memory, where it looks first for the answers, and the long term memory, where it looks for answers when it cannot find them on the short term memory.

Allowing an AI language model to learn from user input is very risky, as uncontrolled information given to it can mess the previously chosen information from which it learned.
AI systems that learn from user input are sensitive to bad companies.

Seems even AI is in the grooming business.

One reason I stray away from such things and discussions about my (nonexistent, thank god) sex life.

Humans are easily manipulated by their psychology, and the sexual drive is one of the strongest instincts to be exploited.

Not to mention the desire for all sorts of things, material and spiritual combined.

It's a very dangerous landscape out there these days, with so many people looking to take advantage of others.

And it's all on record somewhere.

Nice to see ya'll treading with caution.

a reply to: GENERAL EYES

One reason I stray away from such things and discussions about my (nonexistent, thank god) sex life.

Why is a nonexistent sex life a reason to thank god?

Not prurient interest here. I'm just curious as to why you would feel that way?

Don't most people like sex? Or am I one of the "chosen" and lucky few?

Hopefully not a bad/evil/deviant partner in the past. If so, that's awful that you were robbed of one of life's great and simple pleasures by a creepy asshat...

I think hate was manufactured up to this point, for AI to be wanted by most to replace the opportunity to deal with other humans.