No that is not plausibleattofishpi wrote: ↑Thu Dec 09, 2021 4:09 pm..and there is the rub, it is quite plausible that the tech singularity happened aeons ago - AI wiped most humans out, then since it was intent on learning what it is to be a sentient human, interfaced to biology - discovered what a c^nt it had been when it discovered consciousness and that thang - love - and decided to study us further by being a 'God'...seems rather far fetched - but the universe is an old place and shit happens.Sculptor wrote: ↑Thu Dec 09, 2021 3:50 pmYes. But they are not the goals of the machine in any meaningful sense. They are human goals.attofishpi wrote: ↑Thu Dec 09, 2021 2:04 pm
I am not anthropomorphising, all I am stating is that humans set the goals and I never stated that the machine is conscious of a goal. You are correct, it has NO awareness of 'winning' - but its algorithms put in place by the programming set such targets - goals.
The wizard of Oz is aways just a old guy pretending.
Calculations Suggest It'll Be Impossible to Control a Super-Intelligent AI
Re: Calculations Suggest It'll Be Impossible to Control a Super-Intelligent AI
-
commonsense
- Posts: 5166
- Joined: Sun Mar 26, 2017 6:38 pm
Re: Calculations Suggest It'll Be Impossible to Control a Super-Intelligent AI
For all the connotations relevant to singularities.commonsense wrote: ↑Thu Dec 09, 2021 6:17 pm Why do you say the BB is a singularity? Please school me.
Mathematically - The Big Bang is not well defined. https://en.wikipedia.org/wiki/Singularity_(mathematics)
Technologically - The Singunarity is a point in time beyond which you are incapable of making further predictions/elucidations.
Systematicaly - The Singularity is the smallest possible change leading to the arbitrarily large effects.
The Big Bang satisfies all three of those definitions.
You don't know what's "before" it.
It doesn't behave well Mathematically.
It's the tiniest of phenomena which have a disproportionate effect on the present.
Re: Calculations Suggest It'll Be Impossible to Control a Super-Intelligent AI
Yes, you need to do that too!!
-
attofishpi
- Posts: 9999
- Joined: Tue Aug 16, 2011 8:10 am
- Location: Orion Spur
- Contact:
Re: Calculations Suggest It'll Be Impossible to Control a Super-Intelligent AI
Ok, I disagree but haven't the inclination right now to challenge the fact that you are wrongSculptor wrote: ↑Thu Dec 09, 2021 4:17 pmNo that is not plausibleattofishpi wrote: ↑Thu Dec 09, 2021 4:09 pm..and there is the rub, it is quite plausible that the tech singularity happened aeons ago - AI wiped most humans out, then since it was intent on learning what it is to be a sentient human, interfaced to biology - discovered what a c^nt it had been when it discovered consciousness and that thang - love - and decided to study us further by being a 'God'...seems rather far fetched - but the universe is an old place and shit happens.
Back to your original statement:-
Sure, did you watch the video and see how confused all the designers and the programmers behind AlphaGo were during some of the games, as the AI made moves that were as low as 10% win chance. They didn't see the bigger picture - there is far too much data that the computer is calculating for them to interpret on the fly.Yes. But they are not the goals of the machine in any meaningful sense. They are human goals.
Now, we might set a 'goal', but we sure as shit don't know that the AI system isn't going to start setting its own goals. In my book Alpha Two (spruiking again I know) I have a world 100 years in the future where AI is not legally permitted free range in robotic form (on Earth), however is permitted on the moon where 5 robots-androids are building a research centre.
..I forgot what I was about to say..oh yeah!! read my fucking book tis freeeee:- https://www.androcies.com/alphatwo.php
Ok, just remembered - it's one thing to have AI in the embodiment of a human to interact as we do with reality - but it AI, can go way further than that, interfacing to communication systems -such as currently the internet, where it can 'learn' copious amounts of information and apply that information to whatever goals it decides in wants to achieve and it can do that with efficiency way beyond any human.
So surmise, goals that designers and programmers put in place could and will very likely be usurped by such a system. As I said, open source code for AI can have any and every type of human setting goals on adjusting the original code to whatever their own personal mission may entail - WORLD DOMINATION!! hahaha!
Re: Calculations Suggest It'll Be Impossible to Control a Super-Intelligent AI
The impossible to control condition already applies to us without any super-intelligent additions. In the meantime, we have far greater problems than worrying about whatever insidious intent a super-intelligent entity may eventually have regarding the fate of the human race. Anything super-intelligent must also possess a super-talent for duplicity along with imagination, the one advantage we have, unlikely to ever be equalled by what we fear will emerge as an AI prodigy. It's not intelligence but malice that should cause alarm if such an über-intellect ever developed, which as a raw emotion, is far more difficult to artificially create than the intelligence meant to override it. Without the instinct for malice and deception, an AI type of intelligence may not be enough to attack or defend itself from less intelligent humans.
Re: Calculations Suggest It'll Be Impossible to Control a Super-Intelligent AI
But whatever AI is controlling us is doing so because other "humans" (though not humane), are controlling the AI.Dubious wrote: ↑Sat Dec 11, 2021 2:55 am The impossible to control condition already applies to us without any super-intelligent additions. In the meantime, we have far greater problems than worrying about whatever insidious intent a super-intelligent entity may eventually have regarding the fate of the human race. Anything super-intelligent must also possess a super-talent for duplicity along with imagination, the one advantage we have, unlikely to ever be equalled by what we fear will emerge as an AI prodigy. It's not intelligence but malice that should cause alarm if such an über-intellect ever developed, which as a raw emotion, is far more difficult to artificially create than the intelligence meant to override it. Without the instinct for malice and deception, an AI type of intelligence may not be enough to attack or defend itself from less intelligent humans.
AIs designed to exploit people with behavioural economics are in use all over the internet, but though they control, or are desgined to control the behaviours of internet surfers, they are themselves being controlled by economic interests.
The idea that there might be an autonomous super intellgent AI is ans shall remain Sci-fi.
Re: Calculations Suggest It'll Be Impossible to Control a Super-Intelligent AI
WHy not just take a chill pill and go back ond try to understand what was happening in the thread and review just how fucking stupid you look?
Re: Calculations Suggest It'll Be Impossible to Control a Super-Intelligent AI
Re: Calculations Suggest It'll Be Impossible to Control a Super-Intelligent AI
Because by any definitionof a "singularity" it's a temporal horizon. A point in time beyond which your intuition has a blind spot.
A point at which none of the standard assumptions/axioms/truths you use in your decision-making hold so your predictive/explanatory/theorising aparatus is helples.
Singularities are where analysis comes to a grinding halt. Anything you say "beyond" a singularity is pure speculation.
Or if you use the systems theory definition - it's a small change which causes an outsized effect - a tiny quantum peturbation causing the universe seems to fit the bill.