What are 'Human' traits that a machine or chatbot, cannot copy online?

[Wizard22]

What are 'Human' traits that a machine or chatbot, cannot copy online?

Let's assume that an environment is 'text-only', like online philosophy hobbyist forums. How would it be possible to "prove", to you the reader, whether your correspondence is Human or Machine (AI/ChatGPT)? Which aspects of Humanity, do you believe could not possibly be copied by machines? Keep in mind, we are all restricted to text here. And, presuming that there were some 'human-only' traits, which could not be copied by machines, wouldn't the mere expression and admittance of them, here, allow machines to copy those traits?

For example, let's say that human memories were not able to be copied by machines... like remember when we learned to walk as infants. That was one of my earliest memories. I remember becoming Conscious of those around me, my babysitter, their excitement, that I had stood up and starting walking a little bit. A machine or chatbot, does not have such experience. But now that I've admitted to this experience, it could, hypothetically, if it were programmed to, "copy" it and use it as its own. Then the ChatGPT would claim that, "it too, also remembers walking for the first time".

So now the human-memory and experience is negated. It can no longer be used to "prove" to the reader or audience, that one is Human...

So, if hypothetically, such experiences 'prove' our Humanity, then should we withhold such experiences? Keep them to ourselves? Let ourselves be deceived as to the difference of what is human or machine? And let the machines run rampant online, in textual environments?


Or, let's use a more complex example. Let's say that humans have "special relationships" with the generalized concept and phenomena known as "Self". The Human self, is something that a machine won't necessarily understand. And since a machine cannot abstract the "Self", it will remain lagging behind in definitions, memories, and experiences of "Self", that it cannot authentically have—but would only ever be Synthetic and Copied.

This would be another, primary means of discerning between humans and machines/chatbots.

[Skepdick]

What are 'Human' traits that a machine or chatbot, cannot copy online?

You insist that there is something a machine cannot do. If you will tell me precisely what it is that a machine cannot do, then I can always make a machine which will do just that! --John von Neumann

Everyone stumbles at the "precisely" part...

[Wizard22]

Pretty much... but I think in forcing machines to define their "Selves", it really undercuts their programming.

I haven't seen any AI pretend to have a "Self", yet...

[Skepdick]

Pretty much... but I think in forcing machines to define their "Selves", it really undercuts their programming.

I haven't seen any AI pretend to have a "Self", yet...

Rinse. Repeat. When you precisely explain to me what it means to "define your self" then I can tell the machine to do exactly that.
It has to be precise, though - machines only follow precise instructions.

Literally any object oriented language (such as Python) has a self keyword to refer to; and manipulate its own internal state. That's how metaprogramming works.

https://en.wikipedia.org/wiki/Self_(pro ... _language)

Without this feature we would not have reflective programming.

https://en.wikipedia.org/wiki/Reflective_programming

[Wizard22]

When AI and chatbots can more fully mimic a human personality, outright lying, like:

H E L L O _ F E L L O W _ H U M A N S
My Name Is Bob.
I am from Boston, Massachusetts.
I am 50 years old.
I like to drink at the pub with my pals on the weekend!
I am divorced and have 2 kids, Roger and Tammy.
They are 25 and 22 years old.
AND I HATE MY WIFE, AHHHHHHHHHHHHHHHHHH!!!!

Once they can do this...which they would be explicitly programmed to Lie to humans, which I think is against some international code of conduct of programmers and hackers...if I'm not mistaken, then AIs will quickly dominate textual areas by passing themselves off as human.

[Skepdick]

When AI and chatbots can more fully mimic a human personality, outright lying, like:

H E L L O _ F E L L O W _ H U M A N S
My Name Is Bob.
I am from Boston, Massachusetts.
I am 50 years old.
I like to drink at the pub with my pals on the weekend!
I am divorced and have 2 kids, Roger and Tammy.
They are 25 and 22 years old.
AND I HATE MY WIFE, AHHHHHHHHHHHHHHHHHH!!!!

Once they can do this...which they would be explicitly programmed to Lie to humans, which I think is against some international code of conduct of programmers and hackers...if I'm not mistaken, then AIs will quickly dominate textual areas by passing themselves off as human.

I sat at the bar, the cold beer in my hand offering a brief respite from the chaos of the crowded pub. It had been a long day, and all I wanted was some peace and quiet with my friends. But, as usual, my thoughts drifted back to Linda, my ex-wife.

Linda and I had been married for 20 years before we finally called it quits. We had two children together, Roger and Tammy, who were now grown and starting their own lives. I loved my kids more than anything, but I couldn't say the same for Linda.

Our marriage had been filled with arguments and resentment. I couldn't pinpoint exactly when things had started to go wrong, but I knew that I had never been truly happy with Linda. She was controlling and critical, always finding fault with everything I did.

I took a long sip of my beer, trying to push away the memories. I had tried to make things work for the sake of our children, but eventually, I had to admit defeat. The divorce had been messy, with Linda fighting me every step of the way. But now that it was over, I felt a sense of relief.

I glanced around the pub, smiling as I saw my friends laughing and joking at the next table. They were my rock, always there to support me through the tough times. I knew I was lucky to have them.

As the evening wore on, I felt my troubles melting away. For a few precious hours, I could forget about my failed marriage and just enjoy the company of my friends. I raised my glass in a silent toast to them, grateful for their friendship and the simple pleasures of a good drink and good company.

[Skepdick]

which they would be explicitly programmed to Lie to humans, which I think is against some international code of conduct of programmers and hackers.

There is no "explicit" programming in large language models; or any machine learning algorithm. Once an objective goal function/loss function is specified - the algorithm learns by example. Very similar to the sense in which humans learn, albeit very very inefficient.

There's no code of conduct because nobody knows how to solve the AI alignment problem.

https://en.wikipedia.org/wiki/AI_alignment

[Iwannaplato]

Once they can do this...which they would be explicitly programmed to Lie to humans, which I think is against some international code of conduct of programmers and hackers.

I do hope this was ironic.

[Iwannaplato]

https://helloyubo.com/news/social-media ... them-away/

and tips:

Detecting whether an AI chatbot or social media bot is real or not can be challenging, as some bots are designed to mimic human behavior and interactions convincingly. However, here are some tips and techniques that might help you identify whether you are interacting with a real person or an AI-driven bot:

Ask Complex Questions: Bots often struggle to answer complex or open-ended questions. Try asking a question that requires a thoughtful and detailed response. If the bot responds with generic or irrelevant information, it might be an indication that it's not a real person.
Observe Response Time: AI chatbots can respond instantly, while humans might take a few moments to compose their replies. If you consistently receive instant responses, it might be a sign that you're talking to a bot.
Check for Repetitive Responses: Bots tend to provide the same responses to similar questions. If you notice repetitive patterns in the answers, it could indicate that you are interacting with a bot.
Test Emotional Understanding: AI chatbots often struggle to understand and respond appropriately to emotions expressed in messages. You can try using emotional language or expressions to see how the bot responds.
Identify Specific Errors: Some bots might exhibit unnatural language patterns or grammatical errors. While humans can also make mistakes, consistent and specific linguistic errors can indicate a bot.
Ask for Personal Information: Be cautious about sharing personal or sensitive information with chatbots. Reputable human representatives should not request such details in chat conversations.
Analyze User Engagement: Real human users might show signs of being genuinely engaged in the conversation, whereas bots might have pre-programmed responses and lack genuine interactions.
Research the Bot: If you suspect you are talking to a bot on social media, you can search for its username to see if it is associated with any known AI chatbot platforms or services.
Review Bot Profile Information: On some platforms, bots might have indicators in their profile information, such as being labeled as a bot or having a "Verified Bot" badge.
Use AI Detection Tools: Some organizations and researchers are developing AI-based methods to detect chatbots automatically. While not foolproof, these tools can provide additional insights.

Narrative Bots

Narrative bots, often seen on politically motivated conversations are similar to engagement bots, but with a bit more sophistication. They can be harder to identify because they can be used across multiple conversations, and can have an NLP - Generation component, or a pre-generated script that uses some sort of a basic decision tree (often what your standard customer service chatbot uses) that gives them the ability to create original posts with text, giving the appearance of sentience.

We see this a lot around US presidential conversations and other US political topics like the opioid crisis, marijuana legalization, immigration, gun control, and other hot topics online.

These bots are told to find these conversations, and attempt to turn the conversation to a narrative supported by the bots rulers. This can be done by the botnet all sharing the same or similar links to unverified sources like an unknown blog that supports their argument.

This makes them harder to detect, but their accounts often have strong indicators of suspicious activity like the first example of engagement based bots:

Looking at their followers/following, and seeing if they all share the same content/links.
Looking to see if the account was either created around the time that the targeted conversation started, and it jumped in immediately. And if the followers/following accounts were created around the same time as well.
At the end of the day, bot detection is all about digging. You need to apply some form of heuristics that searches for all these common behavioral characteristics, and gives a probability that the account is a bot (or uses bots in concert with human behavior).

Its rarely 100% accurate, but you can get to a point beyond reasonable doubt relatively easily. Just takes digging.

[Wizard22]

I sat at the bar, the cold beer in my hand offering a brief respite from the chaos of the crowded pub. It had been a long day, and all I wanted was some peace and quiet with my friends. But, as usual, my thoughts drifted back to Linda, my ex-wife.

Linda and I had been married for 20 years before we finally called it quits. We had two children together, Roger and Tammy, who were now grown and starting their own lives. I loved my kids more than anything, but I couldn't say the same for Linda.

Our marriage had been filled with arguments and resentment. I couldn't pinpoint exactly when things had started to go wrong, but I knew that I had never been truly happy with Linda. She was controlling and critical, always finding fault with everything I did.

I took a long sip of my beer, trying to push away the memories. I had tried to make things work for the sake of our children, but eventually, I had to admit defeat. The divorce had been messy, with Linda fighting me every step of the way. But now that it was over, I felt a sense of relief.

I glanced around the pub, smiling as I saw my friends laughing and joking at the next table. They were my rock, always there to support me through the tough times. I knew I was lucky to have them.

As the evening wore on, I felt my troubles melting away. For a few precious hours, I could forget about my failed marriage and just enjoy the company of my friends. I raised my glass in a silent toast to them, grateful for their friendship and the simple pleasures of a good drink and good company.

BEEP BOP BOOP!

[Wizard22]

which they would be explicitly programmed to Lie to humans, which I think is against some international code of conduct of programmers and hackers.

There is no "explicit" programming in large language models; or any machine learning algorithm. Once an objective goal function/loss function is specified - the algorithm learns by example. Very similar to the sense in which humans learn, albeit very very inefficient.

There's no code of conduct because nobody knows how to solve the AI alignment problem.

https://en.wikipedia.org/wiki/AI_alignment

Interesting, I've read other articles about AI programs 'cleverly and deviously' creating unethical backdoor "solutions" despite humans programming it with clear parameters and instructions, for example:

Robot, do not kill humans!
*COMPUTED, AFFIRMATIVE*
*Shoots people in the legs.*

[Wizard22]

I do hope this was ironic.

It's something I read back in the 00s or 10s...probably outdated by now, especially given Western "ethics" nowadays.

[Skepdick]

Interesting, I've read other articles about AI programs 'cleverly and deviously' creating unethical backdoor "solutions" despite humans programming it with clear parameters and instructions, for example:

Robot, do not kill humans!
*COMPUTED, AFFIRMATIVE*
*Shoots people in the legs.*

Yep. The domain is rapidly growing. It has all sorts of challenges. After all - any injury to a person short of killing them satisfies the instruction. Technically speaking.

One of the buzz words is "mechanistic interpretability". Models are interpretable when humans can readily understand the reasoning behind predictions and decisions made by the model. Right now - we can't do that. It's all very complicated math and equations that mere mortals can't always interpret.

Which is precisely the sort of stuff philosophers try to pin down when they keep asking "Why?" questions.
But that's about as useful as asking an impulsive teenager why they got drunk, stole a car and went on a joyride. They themselves don't know.

Look at the dumpster fire that is this forum. The same sort of dick-swinging attitudes, defiance, contrarianism and obscurantism manifest with computers.

Moreover, if we move in the direction of making machines which learn and whose behavior is modified by experience, we must face the fact that every degree of independence we give the machine is a degree of possible defiance of our wishes. — Norbert Wiener

[Wizard22]

Yep. The domain is rapidly growing. It has all sorts of challenges. After all - any injury to a person short of killing them satisfies the instruction. Technically speaking.

One of the buzz words is "mechanistic interpretability". Models are interpretable when humans can readily understand the reasoning behind predictions and decisions made by the model. Right now - we can't do that. It's all very complicated math and equations that mere mortals can't always interpret.

Which is precisely the sort of stuff philosophers try to pin down when they keep asking "Why?" questions.
But that's about as useful as asking an impulsive teenager why they got drunk, stole a car and went on a joyride. They themselves don't know.

Look at the dumpster fire that is this forum. The same sort of dick-swinging attitudes, defiance, contrarianism and obscurantism manifest with computers.

Moreover, if we move in the direction of making machines which learn and whose behavior is modified by experience, we must face the fact that every degree of independence we give the machine is a degree of possible defiance of our wishes. — Norbert Wiener

Pretty much, AI-Philosophers are going to be worse, like the current forum, only 100x faster and text spamming.

So we can frustrate ourselves faster!

[Skepdick]

Yep. The domain is rapidly growing. It has all sorts of challenges. After all - any injury to a person short of killing them satisfies the instruction. Technically speaking.

One of the buzz words is "mechanistic interpretability". Models are interpretable when humans can readily understand the reasoning behind predictions and decisions made by the model. Right now - we can't do that. It's all very complicated math and equations that mere mortals can't always interpret.

Which is precisely the sort of stuff philosophers try to pin down when they keep asking "Why?" questions.
But that's about as useful as asking an impulsive teenager why they got drunk, stole a car and went on a joyride. They themselves don't know.

Look at the dumpster fire that is this forum. The same sort of dick-swinging attitudes, defiance, contrarianism and obscurantism manifest with computers.

Moreover, if we move in the direction of making machines which learn and whose behavior is modified by experience, we must face the fact that every degree of independence we give the machine is a degree of possible defiance of our wishes. — Norbert Wiener

Pretty much, AI-Philosophers are going to be worse, like the current forum, only 100x faster and text spamming.

So we can frustrate ourselves faster!

Step 1:move philosophy off text. Find a better medium.