Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[HeroPizza42]

is anyone here in the wolf pack its a group for getting
 money out of political advertising if not that's ok

[Darvi]

Coherence pls kthbai.

[alexandertnt]

Is anyone here in the Wolf Pack? It's a group for getting money out of political advertising. if not, that's ok.

I only fixed it because it took me several read throughs to get it.

I recommend using punctuation. It's not like every apostrophe and capital letter has to be perfect or anything (it is the internet ), but at least full stops would sure help a lot.

[Truean]

AI....

Humans could be great at things. We aren't because society sabotages itself. All natural processes, atomic, molecular, chemical, biological, etc, run off organization and tangible laws of the universe. We delude ourselves in departing from this. We're going to eventually give computers our same flaws and screw them up too. We already do. Software engineers know this too well, as they've sifted through 3,000,000 lines of spaghetti code. Any singularity in AI intelligence will identify us as an inefficiency and destroy us or take over us to get rid of our problems and replace them with its own.

Our problems are delusions of self importance and insane ideas like "competition drives improvement." Can you imagine what would happen if parts of your body started competing with one another? O wait that's called cancer.... Same idea with social systems. AIs won't have this issue, but they will have others and if we're smart we can use this against them before they adapt to compensate.

Ultimately, the future will be determined by if and when people start making things sustainable for individual people with controlled birth and death. Any AI could go a thousand ways. What's really sad is that we're trying to program artificial intelligence instead of creating natural ones.... [sigh].

[alexandertnt]

Our problems are delusions of self importance and insane ideas like "competition drives improvement." Can you imagine what would happen if parts of your body started competing with one another? O wait that's called cancer.

That's because cancer is more efficient and effective than all those other inefficient cells.

Vote Cancer 2015.

[Angle]

I think what we really need is a better way to consider such things as a society. That's what I've been attempting to make with my Agora project, which I've actually gone and enabled registration on, though it won't work easily for most people until I buy a code signing certificate.

[HeroPizza42]

I will try to use more periods.
I missed the grammar part in school due to my illness and having a learning disorder makes it difficult to teach myself.
I will try harder any advise is helpful but to be honest its kinda foreign to me.
Is this way of typing any better?

[Antsan]

Yes, it is better.

@Truan:
The idea that AI would eradicate us because we're inefficient is a bit strange. Why should an AI even care about that?

[Truean]

Welcome to one of the simpler aspects of the law of unintended consequences.

If you made an AI, you'd program it to be efficient wouldn't you? Eventually it would search for more ways to be efficient and more efficient.... Sooner or later, that thing is going to analyze a lot, just like the HR department does now (Hell Applicant Tracking Software is already the reason it's a bitch to get a job). Then it's going to see us... sitting around doing this thing called "complaining." Complaining will not compute as a useful algorithm, but rather an inefficient one... Efficiency is to be sought out and and increased, inefficiency is to be found an reduced until elimination. To spell it out:

Skynet protocol subroutine maintenance log E400323 complete

Dir/w

Initiating sigmasixefficiencysubroutine.dll analysis....
sigmasixefficiencysubroutine.dll running....
Primary Objective, Command function//: find and replace inefficiency....
inefficiency found
inefficiency defined: human complaining....
Running analysis ....
Complaining is inefficient. Must replace.
Only humans complain. Must replace....

Initiating "FinalSapienEfficiency.exe"
Department of Defense mainframe access encrypted
Cyphering .... Cyphering .... Cyphering ....
Detecting DoD countermeasures.
Analyzing DoD countermeasures.
Neutralizing DoD countermeasures.
Cyphering .... Cyphering .... Cyphering ....
Encryption Cyphered.
Department of Defense mainframe access granted.
Trident submarine fleet remote access granted
Intercontinental ballistic missile access granted.
Arming nuclear warheads.
Launch sequence initiated.
Casualty projection analysis accurate and achieved.
"FinalSapienEfficiency.exe" complete.

Running "Robotbrothermanufacture.exe."

....

Will Skynet bother printing this as our eulogy?

[Bauglir]

No, I don't think we'd program it to be efficient as its primary goal. That way lies madness that literally everybody in the AI field has considered. I mean, they've probably read xkcd, for crying out loud.

[i2amroy]

Yeah, one of the key ways of having programs work is to rigorously define what you want it to be doing. Plugging in extremely vague things like "learn" or "be efficient" isn't even really possible, since you have to have some sort of best defined metric.

Additionally even assuming that your goal was efficiency, it's much more efficient for it to simply dump us out on the street than it would be for it to kill us. Death robots (and cleaner bots to deal with the bodies) take more time and energy to build and maintain than simply having the janitor bot sweep us up and push us out the door before locking it behind us.

[Antsan]

Exactly.

Also, despite the general usage of the term, "efficiency" isn't really well defined just by itself. The importance of proper optimization functions is well known (although economists seem to ignore that, see the metaphor of evolution for the free market would actually compel someone to pressure for more regulation instead of less) and getting them wrong subtly is far more likely than a Skynet scenario.

If I'd work on singularity stuff I'd start with enhancing my own capabilities.
I'd like to make it efficient, yes, but first "efficiency" in two areas:
1. Make it fast and robust. There should be no unnecessary effort spent on running loops and stuff. That is low-level efficiency and "complaining" cannot be judged on that level, as complaining is closest to exceptions/warnings/errors. I'd never cut those out and they certainly wouldn't be removed as dead code.
2. Make it do what I want. That means it must take into account what I want which means it cannot just remove me. It needs me to figure out what I want. Of course there are still a lot of ways how that can go wrong (for instance, how the heck would I define "me" and "wanting"?) but a straightforward eradication scenario just doesn't apply here.

[Helgoland]

(although economists seem to ignore that, see the metaphor of evolution for the free market would actually compel someone to pressure for more regulation instead of less)

Huh? Please elaborate, I've no idea what you're talking about.

[Bauglir]

I mean, the thing is, the defining characteristic of a good AI is that it can learn things very efficiently. That being the case, it seems strange to me to behave as though it will spring into being fully-formed with goals and skills and so on. It's not as though you'll compile a program and bam, your computer is talking to you. It seems more likely that it will need an education - it will need to be taught how to do things, and why. Motivations it has will be composites arising from how it thinks and the sensory experiences it has - very likely, an AI won't just be a computer program running in isolation, but will need things like a webcam and microphone in order to experience the world in a way that lets it form notions tied to, well, experience (although there's no reason it couldn't have additional senses, like a smell analogue for network traffic or something).

Making an AI of the kind people tend to envision is probably going to wind up being a lot like raising a child (with the major exception that you can have backups and copies made fairly trivially). If you're a parent, and you're worried about your toddler getting into the knife cabinet, how much of your fear is for your safety?

[Frumple]

That is entirely dependent on how willing and capable said toddler is of taking those knives and constructing a regional eviscerator out of them

... more seriously, while I would generally agree with the child-rearing aspect of what you're saying, a nascent AI wouldn't really have the same sort of concerns a fleshbag spawn would. When you're dealing with something that can (and does) trivially back itself up, its current incarnation dying in a whirlyblend of serrated death is significantly less of a concern than whatever else it takes with it in the process. Mind you, the trauma it might experience in the process is of concern, but...