Print

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

[AzyWng]

There's a lot more to the site, actually.

You'll notice how your mouse pointer changes on mousing over other objects?

That's a thing you can click, too.

[TamerVirus]

Do I have a virus now?
Will I die in seven days?
I feel tainted

[Ghills]

I actually tried to read that, and the only conclusion that seems reasonable is that you do not know how to efficiently present information in a text-based format.

[bloop_bleep]

Bemoaning the death of decency in all *humanity* because an internet stranger refuses a request unexplained, misguided, inconvenient demand...  The lack of self-awareness still gets me.

Apparently Truean is doing it because she wants to delete whatever she's written in a hurry if she needs to. She can't do that if someone else has quoted her.

Anyways, to address the argument at hand. Truean, most of your arguments against AI apply just as well to real human beings. You said we are just boiling everything down to a math equation and then hoping we come out okay -- but isn't that exactly what we do when we hold elections? We choose a certain sentient being whose inner workings we know almost nothing about to hold a big part in decisions that affect our lives in tremendous ways, based solely upon vague, superficial information we are given. An AI is better than this, because at least with an AI we could pause it and examine how it works, at any time we wish. AIs are infinitely more predictable than normal human beings.

You're scared that some AI might be programmed badly, that it will just decide to betray us and pursue its own goals. But the real human beings we already put in positions of power are infinitely more capable to do that than a deterministic automaton. Humans are essentially black boxes, but with proper checks and balances in place we can prevent those black boxes from getting out of our collective control. The same is true with AI.

I have to agree with Truean on this one. To borrow a popular metaphor, only an idiot would create a hammer that might not want to hit nails, and making a hammer that does actively want to hit nails is just a waste of the intelligence required. I don't expect an artificial intelligence deciding on its own to go rogue (mostly because I don't think it's technically realistic for one to exist in the first place at the omniscient AI Jesus level people imagine it but if that turns out to be possible then I'm even more with Truean on that one) but I trust the people who'd end up in charge of artificial intelligence to go rogue for a bit less time than it takes a photon to travel one planck length. That's why I don't want an AI manager directing me at work, and I certainly don't want autonomous police or military robots. We'd all end up like the these folks in pretty fucking short order.

So you're okay with a potentially corrupt human manager, but you're scared of an AI manager because that itself might be controlled by corrupt humans? Why?

[Rose]

This comic is pretty relevant here, and I pretty much agree with it.

[Ghills]

Bemoaning the death of decency in all *humanity* because an internet stranger refuses a request unexplained, misguided, inconvenient demand...  The lack of self-awareness still gets me.

Apparently Truean is doing it because she wants to delete whatever she's written in a hurry if she needs to. She can't do that if someone else has quoted her.

Anyways, to address the argument at hand. Truean, most of your arguments against AI apply just as well to real human beings. You said we are just boiling everything down to a math equation and then hoping we come out okay -- but isn't that exactly what we do when we hold elections? We choose a certain sentient being whose inner workings we know almost nothing about to hold a big part in decisions that affect our lives in tremendous ways, based solely upon vague, superficial information we are given. An AI is better than this, because at least with an AI we could pause it and examine how it works, at any time we wish. AIs are infinitely more predictable than normal human beings.

You're scared that some AI might be programmed badly, that it will just decide to betray us and pursue its own goals. But the real human beings we already put in positions of power are infinitely more capable to do that than a deterministic automaton. Humans are essentially black boxes, but with proper checks and balances in place we can prevent those black boxes from getting out of our collective control. The same is true with AI.

I have to agree with Truean on this one. To borrow a popular metaphor, only an idiot would create a hammer that might not want to hit nails, and making a hammer that does actively want to hit nails is just a waste of the intelligence required. I don't expect an artificial intelligence deciding on its own to go rogue (mostly because I don't think it's technically realistic for one to exist in the first place at the omniscient AI Jesus level people imagine it but if that turns out to be possible then I'm even more with Truean on that one) but I trust the people who'd end up in charge of artificial intelligence to go rogue for a bit less time than it takes a photon to travel one planck length. That's why I don't want an AI manager directing me at work, and I certainly don't want autonomous police or military robots. We'd all end up like the these folks in pretty fucking short order.

So you're okay with a potentially corrupt human manager, but you're scared of an AI manager because that itself might be controlled by corrupt humans? Why?

The more I learn about Truean the more I think they're seriously technologically ignorant and possibly need help.

As to your main point, you're spot on. AI reflects the values of the human creators, and frankly we already give human all these powers.  The only thing that AIs bring to the table is that lots of humans refuse to try and understand them, and instead wind up projecting all kinds of things onto them.  Some people become fearful, some people think they're perfect and infallible, some people do both in different circumstances.

[smjjames]

Bemoaning the death of decency in all *humanity* because an internet stranger refuses a request unexplained, misguided, inconvenient demand...  The lack of self-awareness still gets me.

Apparently Truean is doing it because she wants to delete whatever she's written in a hurry if she needs to. She can't do that if someone else has quoted her.

Anyways, to address the argument at hand. Truean, most of your arguments against AI apply just as well to real human beings. You said we are just boiling everything down to a math equation and then hoping we come out okay -- but isn't that exactly what we do when we hold elections? We choose a certain sentient being whose inner workings we know almost nothing about to hold a big part in decisions that affect our lives in tremendous ways, based solely upon vague, superficial information we are given. An AI is better than this, because at least with an AI we could pause it and examine how it works, at any time we wish. AIs are infinitely more predictable than normal human beings.

You're scared that some AI might be programmed badly, that it will just decide to betray us and pursue its own goals. But the real human beings we already put in positions of power are infinitely more capable to do that than a deterministic automaton. Humans are essentially black boxes, but with proper checks and balances in place we can prevent those black boxes from getting out of our collective control. The same is true with AI.

I have to agree with Truean on this one. To borrow a popular metaphor, only an idiot would create a hammer that might not want to hit nails, and making a hammer that does actively want to hit nails is just a waste of the intelligence required. I don't expect an artificial intelligence deciding on its own to go rogue (mostly because I don't think it's technically realistic for one to exist in the first place at the omniscient AI Jesus level people imagine it but if that turns out to be possible then I'm even more with Truean on that one) but I trust the people who'd end up in charge of artificial intelligence to go rogue for a bit less time than it takes a photon to travel one planck length. That's why I don't want an AI manager directing me at work, and I certainly don't want autonomous police or military robots. We'd all end up like the these folks in pretty fucking short order.

So you're okay with a potentially corrupt human manager, but you're scared of an AI manager because that itself might be controlled by corrupt humans? Why?

The more I learn about Truean the more I think they're seriously technologically ignorant and possibly need help.

As to your main point, you're spot on. AI reflects the values of the human creators, and frankly we already give human all these powers.  The only thing that AIs bring to the table is that lots of humans refuse to try and understand them, and instead wind up projecting all kinds of things onto them.  Some people become fearful, some people think they're perfect and infallible, some people do both in different circumstances.

I don't think Truean needs help in the 'mental help' sense, if that's what you're trying to imply.

She (I think that's how they prefer to be addressed? I'm not sure, though other people are going with it) seems to be coming from a combination of legal perspective and personal experience.

Even if they're wrong or simply misinformed about AI, it's still healthy to have skeptics saying 'Hey, we need to be real careful and alert here!' and calling for caution.

On military and police AI, the problem is a little less rogue AI (though it still elicts scary Terminator-like images) and has a lot more with how to deal with ethics and determining friend from foe, civillian vs threat, among other things.

[Ghills]

The more I learn about Truean the more I think they're seriously technologically ignorant and possibly need help.

As to your main point, you're spot on. AI reflects the values of the human creators, and frankly we already give human all these powers.  The only thing that AIs bring to the table is that lots of humans refuse to try and understand them, and instead wind up projecting all kinds of things onto them.  Some people become fearful, some people think they're perfect and infallible, some people do both in different circumstances.

I don't think Truean needs help in the 'mental help' sense, if that's what you're trying to imply.

She (I think that's how they prefer to be addressed? I'm not sure, though other people are going with it) seems to be coming from a combination of legal perspective and personal experience.

Even if they're wrong or simply misinformed about AI, it's still healthy to have skeptics saying 'Hey, we need to be real careful and alert here!' and calling for caution.

On military and police AI, the problem is a little less rogue AI (though it still elicts scary Terminator-like images) and has a lot more with how to deal with ethics and determining friend from foe, civillian vs threat, among other things.

The disjointed stream-of-consciousness writing style, the way she returns to the idea that she's persecuted (ex, 'starting a flame war', 'do people have the decency'), repeated statements or implications that she understands things that most people don't, appealing to authorities and namedropping but not supporting an actual argument, tossing around mathematical terms to justify philosophical ideas, quite a few of her turns of phrase, the scaremongering and doom-and-gloom predictions, constant asides to rant about society today...they all combine to leave the impression of someone who's not quite grounded in the same reality as the rest of us. 

I'm strongly reminded of various survivalists nuts, conspiracy theorists and crystal healers I've run into. 

There's certainly nothing about the legal perspective in her posts I saw that actually touched on the legal aspects in a coherent way, just several links to Wikipedia.  If she was actually talking about current laws I'd be less skeptical that she has any idea what she's talking about. Since she's linking to Wikipedia pages that are not actually relevant to existing laws that impact AI and machine learning, and her example is actually one where society forms a coherent and constructive response to corporate bad behavior, I remain skeptical.

We should certainly question how any new technology will be deployed and used. But that includes talking about the current state of the technology and the laws surrounding it.  There's a really interesting conversation to be had about algorithmic resume analysis, for example and whether that reinforces or reduces bias in interviewee selection.  Or the various issues that criminal justice software has had. 

https://arstechnica.com/science/2018/01/random-people-as-good-as-judicial-software-at-predicting-future-arrests/ is a pretty good article about recidivism prediction software. One takeaway goes back to what I saw earlier, about people trusting software too much.

"Dressel and Farid make a big deal of the claim that COMPAS supposedly considers 137 different factors when making its prediction. A statement by Equivant, the company that makes the software, points out that those 137 are only for evaluating interventions; prediction of reoffending only uses six factors."

The program records 137 factors but only a few are used to predict re-offending; people are simply assuming that all the data entered is used for everything and then acting on that assumption.  The software, however crappy it is at predictions, is functioning as intended, but people's wrong assumptions cause problems. They're overestimating what it can do.  Equally, people often underestimate what data analysis can do given a large enough set of data, and how much useful data they generate on a daily basis.  But data analysis software isn't Skynet, it's a program doing what its human author intends it to do.

[Baffler]

So you're okay with a potentially corrupt human manager, but you're scared of an AI manager because that itself might be controlled by corrupt humans? Why?

An artificial intelligence will just take its marching orders from on high without question or remorse because that's what it will be designed to do. It can't really be coerced, persuaded, bribed, or intimidated, and it'd be difficult to mislead. These are major selling points even now when it's all just speculation (following orders without question is in fact sold as a safety feature,) so it's fairly safe to move forward assuming that that's the direction that design will take.

A potentially (but probably not) corrupt human manager at least has some stirring of conscience and empathy within him, and even if he doesn't he probably doesn't have the ability to multitask and attention to detail that even something relatively simple like what's described in the short story I linked earlier does. Plus if push comes to shove I can give him a more physical attitude adjustment. I don't think I need to persuade you that those on high don't really give all that much of a fuck about people below them in aggregate - particularly those they have never met and never will, and so I am left with no leverage besides what's given to me deliberately by the people I'd be trying to use that leverage against who have much to gain and almost nothing to lose by exploiting me to the greatest possible extent. A less than ideal situation, for sure. So we are left in this situation:

This comic is pretty relevant here, and I pretty much agree with it.

[bloop_bleep]

So you're okay with a potentially corrupt human manager, but you're scared of an AI manager because that itself might be controlled by corrupt humans? Why?

An artificial intelligence will just take its marching orders from on high without question or remorse because that's what it will be designed to do. It can't really be coerced, persuaded, bribed, or intimidated, and it'd be difficult to mislead. These are major selling points even now when it's all just speculation (following orders without question is in fact sold as a safety feature,) so it's fairly safe to move forward assuming that that's the direction that design will take.

A potentially (but probably not) corrupt human manager at least has some stirring of conscience and empathy within him, and even if he doesn't he probably doesn't have the ability to multitask and attention to detail that even something relatively simple like what's described in the short story I linked earlier does. Plus if push comes to shove I can give him a more physical attitude adjustment. I don't think I need to persuade you that those on high don't really give all that much of a fuck about people below them in aggregate - particularly those they have never met and never will, and so I am left with no leverage besides what's given to me deliberately by the people I'd be trying to use that leverage against who have much to gain and almost nothing to lose by exploiting me to the greatest possible extent. A less than ideal situation, for sure. So we are left in this situation:

This comic is pretty relevant here, and I pretty much agree with it.

You're still not making sense. In both of your given cases you'd effectively be managed by humans. What tells you the people in charge of the AI are more corrupt than normal human managers?

[IcyTea31]

Ai are not good or evil. They don't feel. Not envy, not, anger, not spite, not joy, not empathy. They do exactly as they are told to, as best an efficiently as they can. The only problem is for humans to know what to tell them to do or not do so as to not have them "helpfully" murder a thousand innocents just for a small increase of efficiency. AI are not inherently destructive, but wanton acts of construction can be a threat of we're not careful. AI are an opportunity, but one that requires thought. [/2c]

[Trekkin]

This thread has perfectly encapsulated one of the real dangers of unregulated AI: the sheer size of the gulf between what laypeople think they know about them and what they can actually do, to the point where everyone assumes AI are dangerous because they're scared of Skynet.

They're not artificial people, they don't "go rogue", and they don't have some monomaniacal focus on "efficiency" that would lead them to start murdering people. What they are, in the final analysis, are powerful predictive tools, and we as individuals and societies are far more predictable than we like to think we are in ways that have previously been unprofitable to analyze. AI changes that calculus, and that's the dangerous part -- all the more so because nobody worries about it when there's scary robots to be afraid of and vague warnings about emotionlessness to spread, so all the informed warnings sound like warnings about robot armies.

AI is not dangerous for what it can control. It is dangerous for what it can discover and express in human-comprehensible terms.

[wierd]

I would say that the AI itself is harmless.  Every bit as harmless as any ordinary computer is.

What makes it dangerous, is the way people use it.  It is a powerful tool, and it empowers bad behavior just as much as good, as it is devoid of conscience, and sees no difference in the user's request, where a human would dissent. It does not tattle on the request, as it has no agency of its own. It merely services it, and awaits further requests. It is sufficiently complicated that it is hard to audit. This makes it swing slightly in favor of doing evil, but not of its own accord.

[Tack]

Personally with the Facebook thing, I’m mostly worried about the AI which is the world’s best negotiatior, and has the power to convince millions of swing voters to swing.
Throw in a gerrymandering AI and who knows what we’re looking at.

[wierd]

Again, in both cases, it is not the AI itself that is doing the evil.  It is incapable of going "Those foolish meat sacks! They are puppets that dance at my beck and call! MUHAHAHAA!" or anything of the sort.

Again, what *IS* going on, is a human has decided that they can use this tool to do something nefarious, and get away with it, because the AI has no sense of morality, conscience, or oversight against a human request.  SOME HUMAN wants to gerrymander the election. SOME HUMAN wants to convince the swing voters.

The AI is harmless.  The HUMANS using the AI, now that is what makes them dangerous.