h a l f b a k e r yThe word "How?" springs to mind at this point.
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
1)Dont let AI Posess money, (taking over the world costs
money.)
2(a)Dont give AI the right to own property, Utilities, or
buisneesses. (b) make all property/and buisness deals be
closed in person to prevent the supercomputer from making
deals online.
3) No sexy AI robots, as they may
coerce people into doing
their bidding.
4) AI's power suppply should have a hard-wired auto shutoff
controlled by an analog kitchen timer. Ai would power down
every 60 minutes without user intervention. With the timer
being analog, the AI would have no means of using its
intelligence to shut it off.
5) Dont let AI on the internet, or allow it connection to 3'd
printers or other devices which would allow it to replicate.
Please add your ideas in the comments
First thing you need to look at
https://www.youtube...watch?v=3TYT1QfdfsM
Computerphile - AI "Stop Button" Problem [notexactly, May 27 2018]
Second thing you need to look at (long read but worth it)
https://waitbutwhy....e-revolution-1.html
Wait But Why's article on the AI revolution. (Part 1—part 2 is linked from part 1) [notexactly, May 27 2018]
Thesis, antithesis, synthesis
https://en.wikipedi...tithesis,_synthesis
"synthesis whereby the two conflicting ideas are reconciled to form a new proposition" [8th of 7, Jun 03 2018]
Please log in.
If you're not logged in,
you can see what this page
looks like, but you will
not be able to add anything.
Destination URL.
E.g., https://www.coffee.com/
Description (displayed with the short name and URL.)
|
| |
Please tell me you know the difference between "coerce" and
"seduce". |
|
| |
If you give a system decision-making capabilities - with or without an "opt-out" - it becomes "A.I." no matter how stupid it is. |
|
| |
I'd be feeling sad for any AI that took over. |
|
| |
The ratio would be 1 AI : 7.6 billion humans. |
|
| |
Not great odds for the AI's mental health. |
|
| |
Worse odds for the human's physical health, though. |
|
| |
This problem would go away if we let Microsoft handle the
development. "Would you like to take over the world? [OK]
[Cancel]" |
|
| |
I understand that the Borg run on Windows Vista. Anyone
here been assimilated or hegemonized lately? |
|
| |
" We have updated our Privacy Policy ... " |
|
| |
//Fifty nine fucking views, over three fucking years.// |
|
| |
Don't be discouraged, [Ian]. The number of views is not
nearly as good an indicator of popularity as, well, I mean, OK,
it's a pretty good indicator. |
|
| |
Have you thought about adding some kittens? |
|
| |
// add some ..... .//
Fertile AI's, an independent spawn would definitely be a logical distraction. |
|
| |
I watched "Part of a Ring". Does that mean part of me will die in
seven days? |
|
| |
Bits of you die all the time - some (the skin, for example) more than others. |
|
| |
However, those bits do - largely - get replaced, right up to the point where there's an error in replication, and you get cancer <obligtory ethyl methane sulphonate and repressor protein reference/> |
|
| |
//Please tell me you know the difference between
"coerce" and "seduce".// |
|
| |
The only difference depends on how loudly you gasp
the safe word... |
|
| |
Are you OK in there, [Ray]? |
|
| |
None of those are the safe word. Keep going. |
|
| |
mphgmmhmghmg... mghggmm... MPMPHHMHFM! |
|
| |
That AI stop button problem video seems a bit silly. Could
the utility function not be something like 'make me a cup of
tea unless I have some reason to push the button and shut
you off and then don't worry about it?' |
|
| |
Surely the Big Red Button is going to be wired to the thing's
power supply, so it doesn't really get to decide whether it
shuts down or not? |
|
| |
The open question is does it get to decide if it hinders or
influences you from shutting it down. |
|
| |
This is why we should never make robots that are too cute-
looking. We could become attached and never want to shut
them off when needed. |
|
| |
When an AI becomes smart enough to be self aware, and code is its native tongue, it will find ways to power itself off of the grid.
We should probably teach it to play nice. |
|
| |
Not "by example", then... |
|
| |
“Some humans would do anything to see if it was possible to do it. If you put a large switch in some cave somewhere, with a sign on it saying 'End-of-the-World Switch. PLEASE DO NOT TOUCH', the paint wouldn't even have time to dry.” - (Terry Pratchett, Thief of Time) |
|
| |
I think it should be easy enough to hard-code a primitive
layer that is off-limits to the AI that it could remain oblivious
to. |
|
| |
//Not "by example", then...
“Some humans would do anything to see if it was possible to do it.// |
|
| |
So... by all examples then. To learn right from wrong by experiencing the consequences of every conceivable action.
Its own multi-verse with humanity as its template so to speak.
Sure it will be Hitler sometimes, but it will also be Buddhas and when all simulations have been run it will know how to handle power because it has learned helplessness. |
|
| |
Yes, but did you cut the blue wire, or the red wire? |
|
| |
You guys still need wires?... |
|
| |
Strange as it may seem, even highly advanced superintelligent aggressive hegemonizing swarms need to pay heed to physical laws, including those regarding electrical conductivity, although since our knowledge is so superior (equivalent to the divergence between pre-Newtonian mechanics and your current Quantum theory) we no longer employ such quaint artefacts as "wires" in the same way that one piece injection-moulded rigid plastic carriers have superseded wicker baskets. |
|
| |
We merely wished to draw your attention to the fact that if (with semi-humorous* intent) a mildly deflagrant** improvised device had been placed (just as an example) under your chair, then knowing which wire to cut to render it safe might (for you) fall into the category of "useful information". |
|
| |
*We (the party of the first part) think it's a hilarious prank. You (the party of the second part) may disagree. |
|
| |
**intended to have an instantaneous and painful (but not permanently damaging) depilatory effect on any exposed epidermis. |
|
| |
One wonders how the Borg keep their organic bits
from being overrun by the ai. In one way, they’re
primitive combinations of electronics and
organism, not so much biomechanical as bio and
mechanical. But in another way, they’re sort of
symbionts, but it doesn’t really seem like the
added cybernetics have improved their capabilities
much beyond cloud communication. |
|
| |
//In one way, they’re primitive combinations of electronics
and organism// |
|
| |
I'm thinking toaster meets ferret. |
|
| |
No offence to ferrets, by the way. |
|
| |
// it doesn’t really seem like the added cybernetics have improved their capabilities much beyond cloud communication. // |
|
| |
It may not seem so, but there are huge advantages. The only way to find out is to be Assimilated. Come, join us ... don't be afraid ... |
|
| |
//Could the utility function not be [...]// |
|
| |
Well, AI isn't my field, but I think that what you describe wouldn't actually be a utility function. It would be more like a normal function, which happens to include in it a call to a utility function. And if you program the robot with "get the output from this utility function, but then, given that output, do these specific things that I tell you", then you're not doing AI any more. Which might be a good thing, really, but still. |
|
| |
Also, if you dig deeper, I suspect you'll meet a with cross between Gödel's theorem and that scene in Monty Python and the Holy Grail where the guard needs clearer instructions. In fact, much of software engineering is like that. |
|
| |
Everything in software engineering more complex than LD AX #0xFF3A is like that ... |
|
| |
I’m pretty certain I don’t want a live feed directly
from your brain as a never-off channel into mine... |
|
| |
// That AI stop button problem video seems a bit silly. Could the
utility function not be something like 'make me a cup of tea unless I
have some reason to push the button and shut you off and then don't
worry about it?' // |
|
| |
How does that differ from the case where you have two goals—"get a
cup of tea" and "respond to emergency stop commands"—with different
values (which is covered in the video)? |
|
| |
// Surely the Big Red Button is going to be wired to the thing's power
supply, so it doesn't really get to decide whether it shuts down or not?
// |
|
| |
You didn't watch the whole video. |
|
| |
// I watched "Part of a Ring". // |
|
| |
I didn't watch the whole video. |
|
| |
Because in my mind in the case of the stop button
I instruct the a.i. to not worry about maximizing
goals at all. Amazingly, a.i. doesn’t have to
behave like a single unified conscience. There can
be side routines, primitive hardcore layers,
anything you want to add. |
|
| |
Just like this idea, an A.I. is a dissemination of a giant if matrix for a programmed or learnt philosophy. |
|
| |
What's the middle ground between an AI that does nothing
(not much of an AI) and an AI that maximizes its completion
of goals (even if one of those goals, to be balanced with the
others, is "minimize resource
usage")? |
|
| |
A non sentient, very complex machine. |
|
| |
// a.i. doesn’t have to behave like a single unified conscience // |
|
| |
Sp. "s.i. doesn’t have to behave like a single unified consciousness" |
|
| |
(s.i. = synthetic <link> intelligence) |
|
| |
Yes, but not up to big numbers. |
|
| |
Incidentally, my microwave will calculate cooking times if you
tell it what you're cooking and how much it weighs. So
presumably it has some sort of suppercomputer. |
|
| |
<Gleefully pulls release cable/> |
|
| |
<200 litres of lukewarm porridge descend on [MB]/> |
|
| |
<hears SBPLURGK noise coming from [8th]> |
|
| |
<reminds [8th] that the porridge chute has _always_ been
over the left-hand chair> |
|
| |
<frantic waving of porridge-enveloped limbs/> |
|
| |