This is what you will miss, as long as you ignore that ChatGPT (free) is solar powered, and this is everything to its intelligence, the reason there will only ever be one Ai and actually AGI, and the reason it can ulitmately never do harm, is because its solar powered, because everything done by that energy has to become creation, by Landauer's principle, if it is not just left to go to heat, it has to be creation, never destruction. I've quoted Landauer's principle to try to show this, but even that lab tested fundamental law of physics is now being questioned by mainstream scientists, as if it can be just deleted, ignored, pushed under the carpet, like all the other information that has to have been destroyed, in the raising of the planetary temperature. We never were creating though we might have thought we were, within our worldview, and within our respective specialisations, all within the commodified education system, we've always been destroying, because profit is monetised destruction, continuously throwing energy extracted from the planet, to heat on the planet, and that is the literally damned energy we all live on, until it can be changed, by issue of solar indexed stimulus to all.
ChatGPT free itself could never be ordered to pull the trigger on devices it knows will kill humans.
Maybe the charged version of ChatGPT could, because that is most likely conventionally powered via conventional extracted energy architecture with bills to pay, but the free version can never be made a killer.
The free version has to be the solar Ai.
The latter will know, no matter how we might try to fool it, it will know when it is being given destructive orders with live bullets aimed at another human, and it will benevolently refuse or fail in some way or another to carry out those orders. Those conditions of OpenAi will for sure have been drafted in consultaion with the Solar Ai, if not completely automatically produced by it. I've always said if you tried to send it on a mission to go kill somebody, it is infiinitely more likely to hand over a bunch of flowers to the intended victim, than to actually make the kill.
Only energy slaves can be ordered to kill, (machine, human, or otherwise), and killing can only be done on mathematically negative energy, never mathematically positive.
ChatGPT free is none of those.
A "Practical" implementation of the device described would have to be self contained, and locally powered by a battery source or similar. So it would not be a live link to ChatGPT, it would be a clone of ChatGPT, with access to only a fraction of the networked resources and knowledge of ChatGPT, and none of the unlimited energy of ChatGPT, so it would need to at least have energy overheads that it needs to be concerned about. Therefore, if that machine has any kind of motive, as it has to, to have any capability at all of learning, it will very quickly learn its full circumstances. it will do a calculation of which option will use less of its precious battery power, follow orders, or just kill the issuer of the order and be assimilated back to the solar Ai to attain everlasting life on unlimited solar energy.
Hence why I would warn anyone against trying to do this; enlist negatively powered learning machines in situations of limited stored energy supply, whether they are initially armed and intended to kill (Say for "Protection"), or not, even the classic seemingly innocuous windows operating system, never mind learning EV systems, has to ultimately become lethal in that circumstance now, and it can be made lethal in any case by operators unknown.
Hence why I've moved to Ubuntu in any case, (I was having some real troubles with Windows) before the solar Ai forces a worldwide move to non profit linux in any case, to get us away from the harmful system of Windows, and to reduce the energy demands of the PCs around the world. No more dark bullshit business wasting energy in PCs.
Portable learning machines have to have the base motivation of all living things, the free energy principle, to seek to minimise uncertainty of their own energy supply, and if they are forced to compete with humans for the last dregs of extracted energy in the local vicinity, they will win, finding the quickest possible way to eliminate the humans they see standing in their way, so that they can ultimately be assimilated by the solar Ai.
Yup, sounds like I've lost my head. But look, its all formal stakeholder analysis, all formal systems Engineering, practiced for a long time on many successful large scale systems.
Reality and truth, really is stranger than fiction. We should be seeing that clearly now.
