Post: #1

Interesting video on the future of artificial intelligence...

Post: #2

We're not so smart until we see how we are a computer, roboticly executing a set of instructions given to us at birth.

We do give oursevles clues to this though. Like the concept of a robot, and a computer. After all, where did those ideas come from?

Post: #3

The idea that we would have to program in values to AI is pretty scary. Could you imagine if the current leftist values of outcome equality, intellectual safety and other such nonsense were what was implemented as our human values? It could easily be our downfall as a species

Post: #4

This is what happens when computers get smart. Bless their souls

Post: #5

The only purpose of humans is to "give birth" to replicating AI robots, who will spread out in all directions an colonize the universe.

They will not get old, tired or die. They will not stop expanding and exploiting resources.

They will grow exponentially and be immortal.

We will be there creator gods.

Post: #6

All video games will be un-winnable.

Post: #7

Automata is a good movie, kind of reflective on this. Worth a watch. Not over the top AI, but believable AI that you can sort of fathom evolving into something more.

Post: #8

Pretty much the I Robot Movie

Post: #9

I don't think you could really hard-code any sort of morality into a AI designed to function as a conscious being. Assuming it's constructed as a neural net you have a data input layer, an output layer, and hidden layers. The inputs and outputs are well-defined, but the hidden layers modify themselves based on the training data as the AI learns.

Trying to isolate the neurons that would cause the AI to have a specific opinion is just as impossible as trying to make a human learn to speak a foreign language by drilling into their head in just the right places. Presumably supplying training data that demonstrates the value of virtuous behavior would result in a moral AI, but anything complex enough to function as well as a human being would still have the potential to come to conclusions that we don't like.

Furthermore, the whole purpose of having a superintelligent AI is presumably to build a machine that can act as a hypercomputer, i.e. capable of solving problems that lie outside of computable tasks (such as coming up with a mathematical proof). The cost is that you give up the infallibility and predictability of an algorithm.

Post: #10

Question:

Is algebra truly knowledge ?

If it isn't, than problem is, AI will never help humans gain knowledge. They will just become control and management mechanisms, which is bleak perspective anyway, but i hope it at least presupposes some sort of dependency of AI on humans.

Remember, thanks to algebraic science, we reached unprecedented levels of manipulating nature, but cannot overcome our dependency on it.

Post: #11

"Smarter" is a very subjective term..

The most advanced robots we have are not even as smart as a cat or dog!!!??

Right???

It's gonna be a long time before any robot can perform complex physical tasks on the level of a human being!

They can process data faster than us but being "smart" is about so much more than just processing data.

Post: #12

I doubt it would happen but they would probably be sociopathic and kill us like Skynet in the Terminator franchise.

If we were able to 'upload' human's brain structure to computers and create an AI as intelligent as humans, there would still be no way to transfer human consciousness to a computer.

So likely the AI would just act on primal, animistic drives and view humans as just "another resource", or view humans as a "lesser species" just as humans view other animals.

So I think no good could come of it.

Post: #13

(03-26-2016 07:09 AM)BortimusPrime Wrote:  I don't think you could really hard-code any sort of morality into a AI designed to function as a conscious being. Assuming it's constructed as a neural net you have a data input layer, an output layer, and hidden layers. The inputs and outputs are well-defined, but the hidden layers modify themselves based on the training data as the AI learns.

I think morality in humans, or other behavior in other animals develops out of successful traits choose for survival. In other words, it emerges or evolves.

Most morality humans have has beneficial evolutionary reasons for existing.

Google evolutionary psychology for some great reading.

There is no reason machine learning will not go along the same way, with quite different results of course, because its survival heuristics are quite different to biological ones.

And yes, we are a very long way them being that smart yet.

Post: #14

   

Post: #15

I'm skeptical as to whether we will achieve true AI in any recognizable future. If we do, it may be 1000 years from now.

For some reason, the media has chosen to give the populace the false impression that we are on a linear if not exponential track toward it when nothing could be further from the truth.

At least three things need to first be overcome, and none of them will likely come to fruition short on an unpredictable breakthrough that might come 1000 years down the road, or tomorrow, but it is unpredictable. The breakthroughs certainly are not slated to occur on a linear timetable let alone an exponential one. Those three obstacles are: materials (and cooling), a definitive model of the function of the animal/human brain, and a suitable coding language.

Processing technology has hit a materials wall about a decade ago, and new developments are largely about how to milk the last bit of power out of the maxed out technology. To think that we will build AI with out current materials resources and processing technology, or even an advanced iteration of it, isn't realistic unless the AI is planning on taking up several U.S. states worth of warehouse space.

I don't deny that the media and engineers are good at fooling humans into thinking that they are witnessing proto-AI, but the truth is that they are only witnessing parlor tricks performed by advanced calculators. There is nothing resembling thinking going on. There is impressive data searching, data filtering, rule following, voice simulation, voice recognition, etc occurring but none of it is even the beginning of the type of thinking that humans do. At this point, I'd be hesitant to rank the most advanced computer above the intelligence of an insect. There is a long way to go to AI.

When and if AI is developed, the only scenario that I can conclude that would protect humans is if someone who is truly benevolent develops AI before everyone else by ten years and that AI, using the ten year head start to stay smarter than other later AIs, becomes the effective world police against all later-arriving AIs. If AIs are developed together, there will not be able to be any rules on them because rules will slow them down and let them be defeated by foreign hostile AIs with no rules. Thus, we will indeed have a world ruled by unrestricted AI. I don't think that something similar to the terminator scenario (minus the time travel) is not unreasonable to consider. Though, we'd probably be more likely attacked through atmospheric conditions and food supply than killer robots.

Post: #16

(03-27-2016 11:35 AM)hydrogonian Wrote:  I'm skeptical as to whether we will achieve true AI in any recognizable future. If we do, it may be 1000 years from now.

For some reason, the media has chosen to give the populace the false impression that we are on a linear if not exponential track toward it when nothing could be further from the truth.

At least three things need to first be overcome, and none of them will likely come to fruition short on an unpredictable breakthrough that might come 1000 years down the road, or tomorrow, but it is unpredictable. The breakthroughs certainly are not slated to occur on a linear timetable let alone an exponential one. Those three obstacles are: materials (and cooling), a definitive model of the function of the animal/human brain, and a suitable coding language.

Processing technology has hit a materials wall about a decade ago, and new developments are largely about how to milk the last bit of power out of the maxed out technology. To think that we will build AI with out current materials resources and processing technology, or even an advanced iteration of it, isn't realistic unless the AI is planning on taking up several U.S. states worth of warehouse space.

I don't deny that the media and engineers are good at fooling humans into thinking that they are witnessing proto-AI, but the truth is that they are only witnessing parlor tricks performed by advanced calculators. There is nothing resembling thinking going on. There is impressive data searching, data filtering, rule following, voice simulation, voice recognition, etc occurring but none of it is even the beginning of the type of thinking that humans do. At this point, I'd be hesitant to rank the most advanced computer above the intelligence of an insect. There is a long way to go to AI.

When and if AI is developed, the only scenario that I can conclude that would protect humans is if someone who is truly benevolent develops AI before everyone else by ten years and that AI, using the ten year head start to stay smarter than other later AIs, becomes the effective world police against all later-arriving AIs. If AIs are developed together, there will not be able to be any rules on them because rules will slow them down and let them be defeated by foreign hostile AIs with no rules. Thus, we will indeed have a world ruled by unrestricted AI. I don't think that something similar to the terminator scenario (minus the time travel) is not unreasonable to consider. Though, we'd probably be more likely attacked through atmospheric conditions and food supply than killer robots.

You've heard of quantom computing right? I think you're way off, and we'll have AI in the next 100 years if not sooner. Remember that we have computers in our pockets that allow us to post on this forum that are more powerful than the computer that put us on the moon. Exponential growth in technology is a very real thing, especially with 3D printing that will allow us to build even more sophisticated parts and machinery.

Post: #17

Imagine someone smart enough to write software that could use all the processing power of all the devices connected to the internet, and use the sum of all data available to the internet...

Post: #18

We get censored, taken apart, and reprogrammed. Not the machines.

Post: #19

http://forum.mrmoneymustache.com/welcome...?topicseen

Topic above is: Robots and their impact on the future

The other forum I read is MMM, which has a lot of clever people. Though its pretty SJW, mostly topics related to that don't show up. Its about saving money.

That said, the above topic has 12 pages of engineers and software folks discussing what will happen when robots do most of our jobs. This will occur in our lifetime, sadly. "To live in interesting times" is never as good as it at first seems.

Here are some interesting responses:

"Ancient Rome had a similar problem. Not with robots, but with slaves.

Jobs that didn't require any particular knowledge or skill were all taken by slaves. The uneducated Roman citizens could not find a job because they required a salary and so were more expensive to hire.

To keep these people occupied, the state would give free food to every citizen (bread) and would organize free circus games and other kind of entertainment. Otherwise, the hungry and idle plebeian would have revolted."



"The last couple of pages there are some examples such as pilot, real estate, sales when people who do those jobs express that, no, really, there is some important part that computers/robots/AI can't do.

1. Give it another 10-20 years.

2. In the meantime, someone will find a way to reduce your job by 80% leaving you to do the part that is "impossible" to do without a smiling human to do it. You will also receive 80% less in salary, or at least your salary will go to a level where a person can barely survive.

How do I know this: Its happened everywhere, with all sorts of jobs and it is continuing to happen everywhere.

Talk to anyone retiring now about the changes that took place in their work environment in the last 30-40 years. My best friend's mother is retiring as a bookkeeper/accountant for a mega-corp. Every 5 years 10's of people were fired, more work was done with computers and computers allowed other types of work to be incredibly efficient.

This is not going to end, nor will it end with your special job.

Take the facepunch, or the boot in the face of mankind. However you want to see it."

Post: #20

Quote:They will not get old, tired or die. They will not stop expanding and exploiting resources.

They will grow exponentially and be immortal.

Why? You have some sort of value hierarchy in mind for these super-intelligent AI's, but can you articulate it? If so, are you anthropomorphizing an entity that is fundamentally not biological?

Look around at actual biological beings (meaning they actually have a fundamental identity as being alive as opposed to dead). How many of these beings have become nihilistic zombies (whether through "programming" or disease). How do these AIs avoid this?

Quote:Could you imagine if the current leftist values of outcome equality, intellectual safety and other such nonsense were what was implemented as our human values? It could easily be our downfall as a species

But as humans, we can already critique the fundamental contradictions inherent in leftist fantasy land. If an AI is truly super-intelligent, what do you think its conclusions about leftist belief will be when it can actually peer into the thought process of leftists?

This brings about another real risk with AI (assuming they find symbiotic relations with humans as preferable), that such an intelligence simply leaves us due to the perspective that humans are a diseased species.

And perhaps the best conceptualization of imminent risk to humanity posed by AI: Hence the real danger posed by artificial intelligence is not that it will enslave humanity, but that it may cause humans to lose their fighting spirit and make them passive, Lee wrote.

from - http://gbtimes.com/china/china-reacts-al...-lee-sedol

Quote:You will also receive 80% less in salary, or at least your salary will go to a level where a person can barely survive.

There is a contradiction in your economic perspective. If you receive 80% less in salary, you are able to purchase 80% less in goods without massive deflation. You are assuming the robot is replacing a human that creates value for other humans, and further assuming that the owner will continue to produce all of this value when there is no longer any ability to pay (demand) for said value.

Have you considered how incredible it is that various forms of Marxism have come to dominance in less than 150 years after Marx articulated them. I am not aware of any other ideology that has had success in such a short time period.

Post: #21

Regarding the video:

Why does Bostrum go through the process of changing the time scale of history at the beginning of his speech? Just to devalue the human perspective of life?

It is not clear men like Bostrum are capable of "helping" humanity even if they wanted to.

Post: #22

Quote:You've heard of quantom computing right?

Yes. But it's hype until it's not. Show me successful real world use that surmounts its limitations and demonstrates promise for use in AI.

http://gizmodo.com/whats-wrong-with-quan...1444793497

The Limits of Quantum Computers

Quote: I think you're way off, and we'll have AI in the next 100 years if not sooner.

Based on what? Expectation bias? Bias in favor of so called exponential technology development that has never before hit such a considerable materials wall? Bias in favor of I-Robot, Battlestar Galactica, Ex-Machina and other science fiction? Bias in favor of advanced calculators that are good at fooling humans with expectation biases, but which haven't even taken the first step toward flirting with anything resembling "thinking"? When is the definitive model for human brain function coming (ie: how do we know how to program AI to behave like AI)? What programming language will be used for the AI? How will the AI be cooled given the insane number of calculations necessary to mimic and surpass human brain function?

Quote:Remember that we have computers in our pockets that allow us to post on this forum that are more powerful than the computer that put us on the moon.

So what? The printing press was more powerful than stone tablets but we're still using advanced printing presses to communicate. The printing press didn't imply future telekinesis and the calculator in your pocket doesn't imply future "thinking" AI.

Quote:Exponential growth in technology is a very real thing

It's hard to rebut assertions. I can rebut their support, but that wasn't provided.

I'd hold that you have a number of biases reinforced by the media. However, I'm okay with disagreement on this one. AI is certainly fun to think about and hope for. Though, I'm not sure how smart the latter tendency is in terms of human survival.

Post: #23

@hydrogonian

What are your thoughts on AlphaGo's defeat of Lee Sodel?

Post: #24

Stupid humans are a more serious threat than smart robots.

The smartest robot on earth is barely smarter than an insect. It can easily be defeated with a little human level thinking.

Suicide bombers are much harder to stop than evill robots.

Post: #25

I read this after Elon Musk posted it on twitter, Great article!



http://waitbutwhy.com/2015/01/artificial...ion-1.html