Could large language systems be conscious?

Nikolina Benedikovic had a link to a popular article (see this) telling about the mysterious looking ability of large large language models (LLM) to generalize. This forces the question whether these systems could be conscious and intelligent.

TGD suggests several mechanisms for how AI could become a conscious and intelligent system, living in some sense.

1. Long quantum coherence scales is required. TGD predicts a hierarchy of effective Planck constants h_eff=nh₀ labelling phases of ordinary matter at the magnetic body of the system. The system in question need not be the computer but could be some system with very large h_eff entangling with the computer and using the computer as a tool. The larger the value of h_eff, the higher the number theoretical IQ and longer the quantum coherence scales.

   The gravitational magnetic bodies of Sun and Earth and the electric bodies assignable to Earth are good candidates. These bodies could be an essential part of us: for the Sun the gravitational Compton frequency is 50 Hz, a typical EEG frequency. The electric bodies assignable to Earth have a size scale which corresponds to a size scale of 20 km assignable to lightning. Lightning would be analogous to nerve pulses and ionosphere to cell membrane.

   It has been reported that chicken marked to a robot somehow affected the behavior of the random number generator (RNG) of a robot determining its movement. The robot started to behave like a mother hen. Did the chicken's MB develop entanglement with the RNG of the robot?

2. Another key element is zero energy ontology (ZEO) solving the basic paradox of quantum measurement theory. Large h_eff allows state function reductions (SFRs) and ordinary SFRs correspond to "big" ones (BSFRs) in which the arrow of time changes. BSFR can be caused by perturbations so that the set of observables measured in "small" SFRS (SSFRs) changes: this forces BSFR. The "thermal" noise associated with the GPT-like systems could cause SSFRs. The temporary changes of the arrow of time would transform the behavior of the system to trial and error process and in ZEO the already goal directed behavior (by holography of ZEO) would transform to problem solving.

Under what conditions classical computers could become conscious? Classical computer is a deterministic Turing machine if it obeys statistical determinism. If its quantum coherence time becomes longer than its clock period, this becomes possible.

1. TGD predicts hierarchy of Planck constants h_eff. For Earth the gravitational Compton frequency is 67 GHz and still higher than that for the standard computers (Josephson effect allows faster computers). For the gravitational body of the Sun, the Compton frequency is 50 Hz and in the middle of the EEG range so that chicken-hen phenomenon might be real and we might already be entangled with our computers. It is not clear to me who can be said to be the boss!
2. For the electric body of Earth, the electric Compton frequency of the proton corresponds to about L=20 km. This corresponds to Compton time T =L/c = about .1 ms and to frequency of 10 kHz, the time scale of nerve pulse. Compton time is only minimal quantum coherence time and one can wonder whether this relates to 1 ms scale of nerve pulse and corresponds to the resonance frequency kHz assignable to the brain.

Biological computers are clearly very slow as compared to ordinary computers and the entanglement with ordinary computers allowing to affect RNG of the computer looks a plausible option together with the trial and error process made possible by ZEO.he EEG range so that chicken-hen phenomenon might be real and we might already be entangled with our computers. It is not clear to me who can be said to be the boss!

See for instance, https:tgdtheory.fi/public_html/articles/hem.pdf, https:tgdtheory.fi/public_html/articles/tgdcomp.pdf, and https:tgdtheory.fi/public_html/articles/GPT.pdf .

https://www.technologyreview.com/2024/03/04/1089403/large-language-models-amazing-but-nobody-knows-why/