SciTechTalk

Lover of paradoxes - Talk Radio YT TV Station SciTechTalk: science, technology and general topics-Audio: TV

[SciTechTalk @scitechtalk]

Whitepaper News (AI recent publications from ArXiv, summarized and aggregated)

December 23, 2025

Decoding arXiv AI, One Headline at a Time!

Digital Edition

https://t.co/9IWtj8oBAn

[SciTechTalk @scitechtalk]

@IntuitMachine When asking a LLM to analyze X through the lense of Quaternion Process Theory: What are reasonable X's? What can X be / not be in full generality ? What are good examples and bad examples of what X may be for getting relevant answers? I have no clear idea at the moment.

[SciTechTalk @scitechtalk]

@IntuitMachine So the DSM list becomes less a cabinet of separate disorders and more a map of human process configurations under strain.

[SciTechTalk @scitechtalk]

@IntuitMachine Through the lens of Quaternion Process Theory, all DSM-5-TR diagnoses and conditions can be understood as recurrent disturbances in how mind-life processes:
1. recognize experience,
2. transform state,
3. organize pattern,
4. integrate wholes,
5. and reflect on themselves.

[SciTechTalk @scitechtalk]

@IntuitMachine I am curious!

[SciTechTalk @scitechtalk]

@IntuitMachine OK. Does that mean there will be a separate book QPT 3.5 or even higher versions?

[SciTechTalk @scitechtalk]

Wrote an article about:

Reality Isn’t Made of Matter. It’s Made of Information.

https://t.co/zgJuyUtXyM

[SciTechTalk @scitechtalk]

@PhysInHistory @threadreaderapp unroll

[SciTechTalk @scitechtalk]

@PhysInHistory The incompatibility between General Relativity and Quantum Mechanics. Are there any solutions or at least ways of thinking about how to go about solving that discrepancy?

[SciTechTalk @scitechtalk]

@PhysInHistory @threadreaderapp unroll

[SciTechTalk @scitechtalk]

@PhysInHistory Leonard Susskind thinks Quantum Mechanics is more fundamental!

See:

https://t.co/xtKIwd8jeD

https://t.co/BWFGBgn7Ei

[SciTechTalk @scitechtalk]

@sukh_saroy I tried all 12 prompts with Claude Opus 4.6 on a synthetic use case. It in general gave good, deep answers. But I let another model do independent verification on the generated answers and there was detectable Hallucination in it (but not very much).

Can a changed prompt help?

[SciTechTalk @scitechtalk]

Wat Aldous Huxley in 1946 al zag aankomen: In “Tijd van de oligarchen” waarschuwde Aldous Huxley al in 1946: macht die zich concentreert, technologie die gedrag stuurt en leiders die te vaak handelen vanuit ego in plaats van verantwoordelijkheid.

https://t.co/tCa5ion9kQ

[SciTechTalk @scitechtalk]

@sama Skills (triggered instruction bundles with progressive disclosure) would be a huge upgrade for ChatGPT.
Especially for structured workflows like research and document generation.

[SciTechTalk @scitechtalk]

@sama Will you Implement Skills (like Claude Skills) in ChatGPT? I think it would be a very good feature to have!