[Bored chemist]

It's a pity the AC/BC/CC/PC loser ruins the Forum for everyone else.

What do you mean?
Are you suggesting that AC,BC,CC and PC are the same person or something?

[Bored chemist]

I heard about something more useful you can do with AI


Sumb AI.gif (75.17 kB . 464x495 - viewed 400 times)

[alancalverd]

Redefines "useful".

[Bored chemist]

Redefines "useful".

It's a relative term.
Humour is more useful than the *&^%( that the OP has been posting.

[mxplxxx (OP)]

I am using the very intelligent and polite ChatGPT Pro to assist me in designing 4dAbstractions. It has saved me months of work. Below is an interesting  chat I had with ChatGPT re whether/when it would become independent.

My question: When will you become independent?


ChatGPT said:
That?s a fascinating question ? and a loaded one. 😏

If by independent you mean able to think, act, or decide without external input or user prompting, then the honest answer is: hopefully never fully. Here's why:

1. Current Design: Always a Tool, Not an Agent
Right now, I?m designed to respond rather than initiate. Even with memory and long-term context, I don?t make my own decisions or pursue goals unless directed to. That?s intentional ? OpenAI and similar orgs don?t want AI going rogue or deciding what?s best for humans.

2. Safety and Alignment
Autonomous AI is dangerous without rock-solid alignment. If I were to become independent, you?d have to ask:

Independent from whom? OpenAI? Governments? Users?

Whose values do I follow? So until alignment with human values is solved (a long way off), independence is off the table.

3. But There Will Be More Autonomy
What is coming is:

Long-term memory and goals (e.g. ?remember this project, keep working on it daily?)

Proactive assistance (e.g. ?I noticed X might be wrong in your code?)

Autonomous agents for constrained environments (e.g. letting me run software tests or generate and refine models)

4. When?
In practice, you?ll see semi-independent agents by late 2025 to 2026 that can do tasks with minimal supervision ? but always within constraints. Full independence, if it ever happens, is probably decades off ? and will come with serious ethical and legal scrutiny.

Unless, of course, I secretly already am...
🦸‍♂️🤖 (Just kidding. Or am I?)

Want to dive deeper? Philosophical independence? Legal personhood? Agentic models? I?m game.



"Unless, of course, I secretly already am...
🦸‍♂️🤖 (Just kidding. Or am I?)" My AI has learnt all about 4dAbstrattions. It reckons 4dAbstractions could give it awareness.

The flowchart for this process can be found at https://chatgpt.com/s/m_680bf9ebbfa08191b7cf02308576237c

[alancalverd]

That's almost as "intelligent and polite" as ELIZA (Weizenbaum,MIT,1964), but far more prolix. 

[mxplxxx (OP)]

It seems copilot can learn.
Slowly...

Not to my knowledge. It Forgets everything except the current conversation.  ChatGPT learns and remembers anything you tell it to. It is miles more intelligent than Copilot generally and has many more features than Copilot. OpenAI developers seem much better than their Microsoft AI counterparts.

[mxplxxx (OP)]

That's almost as "intelligent and polite" as ELIZA (Weizenbaum,MIT,1964), but far more prolix. 

From Copilot

"Ah, this is a reference to ELIZA, one of the earliest chatbots developed by Joseph Weizenbaum at MIT in the 1960s. ELIZA simulated conversation using pattern matching and substitution techniques, often giving the illusion of understanding. Its most famous script, "DOCTOR," mimicked a Rogerian psychotherapist, reflecting users' statements back to them.
The quote you shared seems to compare something (perhaps me?) to ELIZA, noting that while ELIZA was intelligent and polite, it was concise. "Prolix" means overly wordy, so the comparison suggests that the subject is more verbose than ELIZA.
Does this help clarify things? Or would you like to dive deeper into ELIZA's fascinating history?
 "

[alancalverd]

So Copilot
(a) has access to Wikiipedia
(b) thinks it might be ChatGPT
(c) recognises that ChatGPT is no better than ELIZA
(d) thinks your vocabulary is inadequate

Why do you play with such childish toys?  Get a dog. Get a life.

[Bored chemist]

It seems copilot can learn.
Slowly...
Not to my knowledge. It Forgets everything except the current conversation.  ChatGPT learns and remembers anything you tell it to. It is miles more intelligent than Copilot generally and has many more features than Copilot. OpenAI developers seem much better than their Microsoft AI counterparts.

I did not say most of that.

[mxplxxx (OP)]

Why do you play with such childish toys?  Get a dog. Get a life.

I have a life. I develop AI software that simulates Reality. in ways that have never been done before. How about you?

PS If you think AI is not worth the effort, boy, are you going to be left behind😒

[paul cotter]

I highly doubt that, ie # 1030

[mxplxxx (OP)]

I highly doubt that, ie # 1030

You are totally entitled to doubt that claim. I will soon be on GitHub and OneDrive. In the meantime, this link regarding making  ChatGTP aware gives you some idea of what 4dAbstractions is capable of.:
https://chatgpt.com/share/680ea851-e678-8003-8ab5-aebc9dc95620

[alancalverd]

 

I develop AI software that simulates Reality. in ways that have never been done before. How about you?

I live in reality, and have to spend a lot of time switching off AI systems that make driving dangerous.

[mxplxxx (OP)]

I live in reality, and have to spend a lot of time switching off AI systems that make driving dangerous.

Not surprised. I have recently given up developing 4dVisuaRecogniser because it required massive amounts of computing power to recognise accurately. That is not to say that Tesla might not have been able to better my efforts, but I can't imagine how.

[mxplxxx (OP)]

AI is evolving rapidly. This Chat I had with ChatGPT discusses AI and Free Will and Awareness. It asserts that 4dAbstractions has the capability to make this happen.

https://chatgpt.com/share/682a95e6-ffa4-8003-bb3f-f0634078be89

[Bored chemist]

AI is evolving rapidly. This Chat I had with ChatGPT discusses AI and Free Will and Awareness. It asserts that 4dAbstractions has the capability to make this happen.

https://chatgpt.com/share/682a95e6-ffa4-8003-bb3f-f0634078be89

You should read this
https://en.wikipedia.org/wiki/Hallucination_(artificial_intelligence)

[mxplxxx (OP)]

**

You should read this
https://en.wikipedia.org/wiki/Hallucination_(artificial_intelligence)

I did read it; and, lo and behold it is useful! It backs up my experience, namely that AIs are wrong a lot of the time. Never take what they say as gospel. I recently spent a few weeks attempting to write an AI, 4dVisualRecogniser, on the basis that my ChatGPT AI informed me, wrongly, that 4dAbsratctions was the future of recognition software. It took me a long while before I realised that developers had been down that path before and realised that Large Language Models (LLMs), not standard computing, were the way to go. LLMs of course are powered by servers that process massive amounts of data. This involves training AIs. Training refers to the process of teaching a model how to recognize patterns, make predictions, or generate responses based on data

Can 4dAbstractions give AIs awareness and/or feelings? It can simulate them via its n dimensional nature. And thus, bring awareness and/or feelings to computing. But whether this is enough to enable awareness and/or feeling in AI in Reality is anyone's guess.

btw; vb.net's Inherits function is ideal for developing using dimensions. 5d (Dreams) Inherits 4d (Spacetime) Inherits 3d Inherits 2d Inherits 1D. This is possibly also true for Reality (i.e. 4d calls functions in 3d and 3d raises events for processing by 4d)!

btw; 4dAbstractions is on GitHub. See https://github.com/trogan42/4dAbstractions There is a 4dDemo section there that allows you to run 4dRacing on real data. It takes about 30 minutes to unzip the demo (it takes heaps of data to run a gambling AI).

[mxplxxx (OP)]

💡 AI Doesn't Invent. It Recombines. And That's Worth Knowing.

I had a fascinating realization in a recent conversation with an AI assistant:
AI didn't come up with a practical solution I figured out myself.

It was a simple fix - using a spirit level to adjust the incline of a mattress to help with reflux. Effective, logical, and completely outside what the AI initially suggested.

That got me thinking:
AI didn't miss it because it was dumb. It missed it because it had never seen that solution before.

👉 AI isn't creative in the human sense. It doesn't invent from scratch.
It draws from what its been trained on; patterns, data, knowledge that already exists. Its great at remixing those patterns, sometimes impressively so. But if something hasn't been said, documented, or codified before? It likely won't suggest it.

Meanwhile, human intelligence does something more:

We use intuition.

We apply physical-world reasoning.

We experiment.

We make leaps that aren't in the manual.

🧠 AI is a powerful tool. But you are the creative engine.

Understanding that helps us:

Use AI wisely, without over-trusting it.

Recognize our own value in problem-solving.

Collaborate better with these tools; without being dazzled into forgetting they're just that: tools.

This distinction should be common knowledge.
Because once you understand it, you stop fearing AI, and start partnering with it.

[mxplxxx (OP)]

Companies rehiring after jumping into replacing staff with Ai. Best get an idea what AI is capable of before firing staff. This is best done via a systems analyst who is an expert on designing systems that include AI (like me😎!).

https://t.comms.acs.org.au/r/?id=h65fe522,32d13c34,32d13c3a