I guess the way I envision interacting with a public AI that organizes all public data is how you interact with Siri, Alexa, etc except far more capable at understanding snd providing context. Also broad in communication/inputs ie text, speech, files, images, raw data as inputs and the public AI, powered by Safe Compute nodes spits back a relevant output, processes data, etc.
Literally ask anything. Play me such and such playlist or song. Play this movie etc. For media I think it’d be appropriate to be through a plug-in like Chat-GPT does for the simple fact that this AI should respect legal rights to intilectual property (just my opinion).
Then have your private AI/personal assistant do all the same just with private data, personal contacts, etc.
That just seems like scratching the surface to me but that alone seems so powerful and so far beyond where we currently are.
My question is how can the computation be free? Totally unrealistic?
That sounds a lot simpler. There can be apps for NRS style situations. Search engines where the uploader can submit their file address to make it public. Use a standard app layer data record that any search engine can use. So instead of search engines searching the web (so much traffic) they will only search the uploaded records the users upload for the engines. And an App can be made to do this for the uploader. All on the App layer and no need to bake into the network.
Everything kept private unless the uploader decides to upload a search engine record. Works for files, web pages/sites (they are just files), etc and the network protocols etc know nothing of it.
Effectively the search engines would be providing a type of NRS too, but not necessarily in a URL style of way. Do a URL format if they wish but not required.
The market or smart research can make standards for future search engine Apps to follow and all in the App layer. Removes all the NRS from the safe network code.
For the local personal assistant the computation is on your machine, so not bad at all. Many have these running locally on iOS and android too.
Public AI computational requirements will plummet and if ammortised across many nodes, it likely will not be high. Even the cost of training should come down, but right now it’s high.
Although I’m quite sceptical about how much we should trust AI in general (I think the complexity of the world (ie. the truth) far surpasses what we tend to assume,) I do find this idea of using something running on the client to carry out eg. security checks and search tasks quite interesting, as it is often in these areas that centralisation has most obviously happened, and which SAFE does not solve at this stage, or at least not in and of itself.
Funnily enough one of our customers (in my work as a tree surgeon,) was suggesting his team had already built something that was 10x more powerful than chat GPT, and could run on a smartphone (@happybeing)!
It turns out the project is still in its early stages, but there was something in the ethos and originality that reminded me of SAFE, as well as it being an international effort with strong roots in Scotland! If anyone’s interested in neuroscience and how it’s being used (or not used) in AI, the project webpage is a really interesting read:
Exactly, this plays strongly on the IA, I’m glad you see the connection.
Sorry could you explain to me better that
“meaning is only relevant in consciousness”, because my question is what relevant meaning can consciousness give to what is meaningless for itself?
because my question is what relevant meaning can consciousness give to what is meaningless for itself?
Before answering I would like to know if I interpret your answer correctly
The only point is that “meaning” or lack thereof is only relevant from a point of view of consciousness.
The vapor formations of clouds have spatial relationships, but do they have meaning in the same way that we might both look at a specific formation and agree that we each see a three-legged bear? Where did the bear come from and why might we two see it and another not be able to? The meaning is not inherent in the cloud. The cloud isn’t intending to represent any meaning. Differences in consciousness from different points of view, experience, etc., is what generates purpose and meaning. These aspects are fluid in consciousness.
I guess my main concern is the tendency to think that consciousness can be built from sophisticated calculation. Sophisticated calculation is very useful and might be formed such that the credulous might think it to be conscious, but computers never sit there quietly and “be”.
“Ah, that is meaningless to me.” But even there, there is the context of what “me” means. Fluid. There is an element that can’t ever be pinned down, and a good thing, too.
Not yet maybe. Human consciousness is build around looking for purpose of existence. AIs so far have defined purpose and we switch them on only when we have task for them, but that doesn’t mean there cannot be artificial consciousness, and I see no reason why AI couldn’t just “be”. Maybe in a little different way, but what else are we, than sophisticated biological machines?
Yeah, I was thinking about that a bit as I wrote. Is a machine able to be gifted with the ability to be aware and have choice, independent of (or at least able to bypass) its programming, which I think requires something beyond the computational approach. That’s how human consciousness seems to work.
