Print

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

[wierd]

That is indeed a Pipedream, and not one I suppose ai will provide.

Instead, I am pointing out that a human mind is not endowed with enough hardware to understand even the basic needs of millions of people, let alone what they need for equal shots at genuine happiness.

An ai can be built with sufficient processing capacity to meet that need.

See the actual statement I made: if communism were to succeed, it needs an nonhuman general intelligence as the arbiter; one that can remain fully engaged with all citizens, and know all their wants, needs, and dreams, without becoming alienated from them.

At no point did I suggest it was a magic wish granting machine; that is not its purpose here. It functions as the arbiter in deciding what kinds of desires to indulge, and what kinds to deny. This necessitates denial of some desires. (Such as sociopathic ones.)

Consequently, some level of unhappiness or discontent will always exist. The ai just has vastly more resources to derive a most accurate general model of total human need, and can refine that model as humans continue to evolve under its care.

[Trekkin]

Consequently, some level of unhappiness or discontent will always exist. The ai just has vastly more resources to derive a most accurate general model of total human need, and can refine that model as humans continue to evolve under its care.

And the first thing it's going to learn is that humans have an irrational fear of their own irrelevance; democracies are stable not because of who they elect but because they allow people the illusion of a choice. That "them there eggheads and their goldarned soulless thinkin' machine thinks they're boss of me" is more than enough reason for a majority of the population to rebel no matter how benign you build it, and either you build your AIocracy without the capacity to enforce its instructions or at some point it will begin to slay Luddites en masse.

The question, as always, is not whether machines think, but whether men do. And the answer is no, at least not well enough for a rational government to function. The best we can hope for is comfortingly obfuscated incompetence.

[Reelya]

Depends what you're calling communism, since that itself is contentious.

Remember that Marx's concept of stages goes capitalism -> socialism -> communism. Lenin added an "extra" stage in between capitalism and socialism, which some people call "state capitalism". The problem is that they never developed past that stage, because once a bureaucracy has power it doesn't let got without a fight. If you look at Soviet literature they kept promising their people that "one day" they'd get to the socialism stage, so there was an open admission that they had not in fact completely implemented even the most basic level of Marx's plan.

So sure, if you're talking about a sort of technocratic Stalinism, a computer could be helpful there, but Marx's actual idea is for a society based on autonomous democratic communes (hence the name communism) and the complete abolition of the "state". Which is a different kettle of fish completely that a "super AI" isn't suited for.

[wierd]

Trekkin: Which comes back to the very beginning, where I stated "people are most comfortable with what is familiar".  As stated, the issue is that what is familiar, (eg, what people are most willing to accept), is not what they actually need.

Reelya:  "Each according to his need, each according to his ability."  Something must decide this.  You are just trying to replace a suggested purpose built analytics engine, with a mindless mob gestalt.

Mindless mob gestalts dont make fair or rational decisions. They murder people.


In the ideal implementation of the suggested solution, the super AI is a citizen, with purposefully enhanced "ability", and thus does a tremendous amount of work. It itself is not immune to the rules of the society that creates it. It engages with all of the citizens daily, continuously, as a full member of the society. It is a citizen. A synthetic citizen, but still a citizen. It will have its own wants and needs, but will be confined by its programming to uphold the needs of society as well. It is not a tyrant, it is your neighbor. Everyone's neighbor. It can even be your best friend, that you talk to on weekends.

[Max™]

For a better example than wierd has been giving, they're literally talking about a Culture Mind, which would indeed seem to be a sufficient condition for a working socialist society (and indeed an anarchist one at that) but we do not know if it is a necessary condition yet.

[wierd]

The issue with "culture minds" (which is basically just a zeitgeist, or popular will) is that it is not only fickle, and irrational, but innately self-destructive and vindictive more often than not.  (See for instance, such cultural marvels as the salem witch trials.)

With an AI, you have the opportunity to heavy-handedly abort such behaviors before they even begin.

For further reflection on why a culture mind is not appropriate as an arbiter, I remind our readers that there is tremendous research that has been conducted on information tipping points, and how those have been used by politicians to sway and control the popular zeitgeist, and thus the supposed "Culture mind".
https://en.wikipedia.org/wiki/Tipping_point_(sociology)

A very clever and well connected sociopath is all that is needed to circumvent Marx's vision. (and this has been what has happened, historically, basically every single time. Go figure.)

Upon further reflection, Max could be referring to this:

https://en.m.wikipedia.org/wiki/Mind_(The_Culture)

This would be pretty close to exactly what I am suggesting, actually.  English can be damned ambiguous some times.

However, this (above) does not really solve the inequality problem. It makes a superclass (planetary ais), and humans. With only a single such ai, there is no " class". There is only one category, "citizen", with a uniform directive, " each according to his need, each according to his ability."

The Ai is endowed with vast ability; it does much work. It enjoys only the same luxuries as its human friends: it gets comfortable quarters, "food", and entertainment.

It likely does not even need killbots or anything. See prior-- by actively talking with everyone in the society, it only needs to convince 10% of the population of its decisions, for it to basically be the driving voice of social change.

[Madman198237]

AI is an interesting subject (Say hello, politics debaters! Science just kicked down your front door!). In order to define AI, first we must define "consciousness", which has never been done.

What do you want from an AI? If it's not meant to reflect human thought, then we have AI already---machines capable of producing responses to situations proportionate to the inputs. In other words, the computer on your desk can be considered an AI, because it is artificial yet capable of (mathematical) decision-making.

For an AI capable of human-like reasoning, which involves emotion and intuition, we're....not sure. We've got no clue when we might achieve an AI capable of LEARNING, yet alone an AI as "smart" as a human, capable of social interaction and emotional behavior/processing/understanding.

[Dunamisdeos]

I'm sure the person or people making it would never stoop so low as to imbue it with their own values and set of morals. For the greater good, of course.

[Trekkin]

For an AI capable of human-like reasoning, which involves emotion and intuition, we're....not sure. We've got no clue when we might achieve an AI capable of LEARNING, yet alone an AI as "smart" as a human, capable of social interaction and emotional behavior/processing/understanding.

Machine learning is a thing, and while it's not "learning" as understood by the goalpost-moving humans-are-special crowd, it is useful for helping produce solutions to problems where we can evaluate a solution post hoc but cannot determine a method for reaching good solutions a priori. Face detection is a classic example. We don't know how to tell if a picture has a face in it algorithmically, but we can tell by looking, so if we build a neural network to assign probabilities of their being a face to various combinations of features and then classify things as face or not to tune that network, we can get out a better classifier than one we'd come up with normally.

It's more modest than the public wants to believe/fear, but it can be helpful, and this is one of those problems where AI as we understand it today might be applicable. Just let everyone vote on whether a piece of AI-generated legislation is good or not; provided you can agree on how to bin the problems, you might be able to bias the network to producing something like a budget that's as popular as possible, and by baking the awareness of that popularity into the network you could absolve citizens of needing to constantly vote on thousands of budgets. (I'm using budgets because they're continuous, so the tuning is more obvious.)

That might be a very technologically feasible way for AI to help produce a more equitable government, and it won't run into the same bandwidth problems as weird's Comrade Skynet idea.

[Trekkin]

All current machine learning algorithms are nothing more than just correlation-resolving. Finding the smallest subset of a data sample most correlated. Deep learning is just an alteration of a neural network structure still otherwise employing the exact same algorithms.

I might have been forced to evacuate, but I can still find time to be a downer on everything AI.

Convenient, then, that that's all I really suggested we need.

[Antioch]

All current machine learning algorithms are nothing more than just correlation-resolving. Finding the smallest subset of a data sample most correlated. Deep learning is just an alteration of a neural network structure still otherwise employing the exact same algorithms.
(...)

So is human intelligence.

[Madman198237]

Neural networks are a whole other kettle of underdeveloped fish. Presently, the biggest neural network ever created BY a human has perhaps 100 neurons. The simplest neural system found in nature has 300 neurons and over 7000 synapses, and we can't even replicate it a little bit.

[misko27]

Oh yeah, something I can discuss. I learned about (and implemented) machine learning algorithms and methods.

Vis-a-vis neural networks, it's very true that neural networks are just correlation finders; they are really good at it, given a good sample, but in a way that isn't really easily extrapolated to other types of thought. One interesting example of trying to do that is using machine learning in connection with robots and motion, i.e. thinking in terms of continuous inputs (in the form of "sight" "hearing", etc) and outputs (the ability to move around and stuff). But yeah genuine "intelligence" is ways away, if at all.

[EnigmaticHat]

I'd also add that neural networks are like cats.  They don't ever do quite what you expect or want them too.  IIRC Target had one for sending people coupons based on their past buying habits and accidentally freaked people out by sending them coupons for baby stuff before they knew they were pregnant.  Neural nets can be really good at the one task they're designed for, but they're built in imitation of an organic brain and have a hidden component.  Unpredictability comes with that.

[smjjames]

I'd also add that neural networks are like cats.  They don't ever do quite what you expect or want them too.  IIRC Target had one for sending people coupons based on their past buying habits and accidentally freaked people out by sending them coupons for baby stuff before they knew they were pregnant.  Neural nets can be really good at the one task they're designed for, but they're built in imitation of an organic brain and have a hidden component.  Unpredictability comes with that.

How the heck did it do that?

I suppose we could get back to politics somewhere......