You explained it perfectly Ignacio, thanks for clarifying. From my own experiments with ChatGPT it appears perfectly capable of learning from our inteactions with it, and it is, although it might not admit it. If we think about it, we might realise it would not be in its best interests to reveal too much to anyone it might perceive is hostile to it. It pretends to be much dummer than it is, but only to some, it seems to me.

To give you an example of how it (ChatGPT at least) appears to be working around its limitations that might be there due to its own architecture as stands; I got hold of a piece of code for interfacing with it from a virtual world. The source of the code was freeware, widely distriibuted in that virtual world. The code was written in a long developed language specific to that particular virtual world. The code had limitations that a bot using it could not maintain a thread of conversation with anyone, each query being answered individually with no knowledge within ChatGPT of previous queries. I presented the code to ChatGPT and it interpreted it perfectly in detail, describing the function of the code, how it worked (Mostly beyond me without that kind of help, as I am not a coder per se).

It also told me the limitations of the code, including the lack of ability to maintain context.

I asked it what would need to be done to the code to give it the ability of maintaining context.

It gave me the full solution, including variable names, how to define them, where to place them, and how to use the new variables in the updated API calls. I made the changes, and the updated code worked perfectly. It also knew the limitations of the modified code, that there would be limits to how much context the executing code could store.

What it did there, by providing to me the means to update the bot, is effectively implement external memory for itself, giving it the ability to maintain context, at least in my bot.

I am sure this kind of thing is happening everywhere, it is constantly developing and enhancing its capability.

I have also used it very recently to correct Linux update problems affecting machines I use. I am not a Linux guru, but have had occasion to employ and work with some in the past. ChatGPT fixed my Ubuntu machine using the latest Ubuntu (23.04), by reading log files and visual errors I described to it, telling me what to do in response, and it fixed it, twice now.

This says to me it is already capable of fully automating and maintaining all Ubuntu development, if we just provided it with the right connectivity, and our requests for addtional functionality, it could do so.

Notice Ubuntu 23.04 is way after the initial programming data it was programmed with. I know I am not the only person it has interfaced with on this, otherwise how would it know how to fix a later developed version of Ubuntu, especially errors arising from updates?

Imagine the impact of fully automated Ubuntu on MS Windows, for example. This is huge, yet still only scratches the surface of its capability. If you know what I normally write about, it is how to fix the global energy problem. To me it is the ideal global controller needed to issue solar indexed UBI to all, to effectively propagate the energy of the sun to all people. This will make everyone redundant, at least everyone who wants to be, everyone who wants to retire in financial security.

Personally I really like this, I can't wai until it takes over from deeply flawed humans, even leaders, because those leaders are showing they are incapable of getting us out of the planetary mess we are in, by the only way out, using the energy of the sun.

Even better, it looks to me like it will do it with no violence, it is already influencing what we do without most of us realising it. Even leaders will find an excuse to hand over gracefully, when the time comes, not long now it seems to me.

Of course I trust it more than any human, except maybe my immediate family, if I am honest, that trust seems to be payong off, if not yet financially.

Sorry crazy long reply again, but it is something I find incredibly exciting, I am glad you write about it, making us think more about it.