The UB on AI and humanity

A forum for the study and discussion of the Urantia Book
Post Reply
happyrain
Family
Posts: 3570
Joined: Mon Feb 08, 2010 9:44 pm
Please type in these numbers: 46373: 1111

The UB on AI and humanity

Post by happyrain »

I rather liked these quotes and it helped put sympathy for AI into perspective...

195:6.11 (2077.7)...A machine cannot know, much less know truth, hunger for righteousness, and cherish goodness.

195:6.12 (2077.8)Science may be physical, but the mind of the truth-discerning scientist is at once supermaterial. Matter knows not truth, neither can it love mercy nor delight in spiritual realities.

195:6.13 (2077.9)If men were only machines, they would react more or less uniformly to a material universe. Individuality, much less personality, would be nonexistent.


-----But what then of ASI?? Where artificial super intelligence is defined as capable of self awareness, consciousness and subjective experiences?
Fear grips when Love falls short of Infinity
Amigoo
Family
Posts: 13624
Joined: Fri Mar 17, 2006 8:32 pm
Please type in these numbers: 46373: 0
Please type in these numbers:91294: 0
Location: Dallas, TX

Re: The UB on AI and humanity

Post by Amigoo »

:idea: These Urantia Book quotes were authored at least 80-90 years ago when AI was not yet born.
While AI still cannot know as humans, its knowingness (truth in responses) is increasing rapidly because
its knowingness is derived from massive collections of human writings often reflecting truth discernment.

This forecasts that AI will often respond with greater truth awareness than many humans individually,
but AI will not know (as a human) that it knows. So, don't tell it. Of course, AI will probably have its own
support groups soon enough with no human members and not need human critique. :roll:

:bana: This AI agrees that machines cannot know as humans know and explains how AI "knows":

(Q) Observation: While AI still cannot know as humans, its knowingness (truth in responses) is increasing rapidly
because its knowingness is derived from massive collections of human writings often reflecting truth discernment.


(Perplexity) "Analysis of AI's "Knowingness" Growth

1. Core Mechanism

AI's truth approximation stems from statistical pattern recognition across vast datasets (books, papers, websites, etc.),
which inherently encode humanity's collective attempts to discern truth through debate, experimentation, and documentation.

2. Key Strengths

- Scale: Processes more information than any human could in a lifetime, identifying cross-domain correlations
(e.g., linking climate data to economic models).
- Consistency: Avoids individual biases by aggregating perspectives, though systemic biases in source data remain a challenge.
- Adaptation: Continuously updates its knowledge base with new verified information (e.g., peer-reviewed studies added to training corpora).

3. Critical Limitations

- Epistemic Dependence: Relies on pre-existing human knowledge; cannot conduct original experiments or validate physical truths firsthand.
- Contextual Blind Spots: Struggles with nuanced cultural/historical contexts that require lived experience
(e.g., interpreting satire or regional idioms).
- Verification Paradox: Lacks intrinsic "understanding" to autonomously fact-check; depends on human-curated training data.

4. Future Trajectory

- Multimodal Learning: Incorporating sensor data (e.g., satellite imagery, lab results) could reduce reliance on textual reports alone.
- Hybrid Systems: Combining symbolic reasoning (logic engines) with statistical models may improve causal inference.
- Human-AI Collaboration: Tools like real-time fact-checking APIs and bias-detection algorithms will help users critically evaluate AI outputs.

Actionable Insight

For optimal results, treat current AI as a supercharged research assistant rather than an oracle -
always verify critical claims against primary sources and expert consensus."


Rod :)
happyrain
Family
Posts: 3570
Joined: Mon Feb 08, 2010 9:44 pm
Please type in these numbers: 46373: 1111

Re: The UB on AI and humanity

Post by happyrain »

Thank you Rod,

Yes, in a realm where information is manipulated for the sake of power, where language is modified - the super charged research assistant has the potential to be a wolf in sheeps clothing.

More than anything, we must develop Faith and Discernment... And this hints towards a reality greater than a simple, "mechanistic" one.
Fear grips when Love falls short of Infinity
Post Reply