Artificial intelligence dangers !!!

I am prepared to talk about the failure of intuition but many of us are often encountered.It really is a failure in detecting a certain danger.I will describe a scenario I think just horrible just might happen,and this is not a good combination when it happens.But instead of fear, most of you will feel that what I am saying really nice I will describe how the things we reap from artificial intelligence (AI) can destroy us.Actually, I think it is difficult to see how they can not destroy us or make us their self-destruction.But if you're like me,you will find that it is very fun to think about these things.And that reaction is part of the problem.
OK? That reaction will make you anxious.And if I convince you of this talk that we may have to suffer a global famine,or due to climate change or some other disaster,and that your grandchildren, or their will probably have to live like this,then you will not think,"Great.I like this talk. "Famine is not happy where.But deaths from science fiction to humorous.but one of the things that worries me most about the development of AI that we can not seem to arrange an emotional response appropriate for the dangers ahead.I can not sort this reaction, so I give this talk.It is as if we stand before two doors.After the first door,we stopped the construction of intelligent machines.Hardware and computer software does not improve even better for some reason.Now take the time to think about why this might happen.I mean, based on the value of intelligence and automation,we will continue to improve the technology if we absolutely can.What might cause us to stop doing this?A nuclear war a large scale?A global pandemic?Meteorite impact?Or Justin Bieber became the US president?The problem is, something will have to destroy civilization as we know it.You have to imagine it would be worse how to prevent us continue to improve technology forever,generation after generation.Almost by definition, this is the worst thing happened in human history.So the only option,and it lies behind door number two,that we continue to improve the intelligence machinery year after year.To a certain point, we will create more intelligent machines themselves,and once we own the machines smarter than you,we will start to improve.Then we ventured to what mathematicians call Ij Good an "explosion of intellectual" and that process can be far away from us.And now, this is often a caricature, as I have here,fear before an evil robot army will attack us.But that's not the scenario most likely to occur.Not our machines naturally becomes evil Real concern is that when we create machines talented than us too much then, even if only a small difference between us and our goals alone can also destroy us.Think of the relationship between us ants.We do not hate them.We do not intentionally harm them.Actually sometimes we try not to harm them.We walked past them on the sidewalk.But every time their appearance extreme conflict with one of our goals,For example, when constructing a building like this,we destroy them not to worry.Means that someday when we created the machine whether they are conscious or not,will treat us in a similar regardless.Now, I suspect that this may seem too remote to many people I bet there are some who doubt that super intelligent AI can occur,more or less inevitable.But then you have to find something wrong with one of the following assumptions.And only 3 assumptions.Intelligence is a matter of processing of information in the physical system.Actually, this is a little more than assumption.We have created the limited intelligence of your machine and then,And a lot of machines have been in operation at the level of intelligence and super Vietnam.And we know that the only material can create opportunities for "unlimited intelligence,"an ability to think flexibly in many fields,because our brains manage it. Right?I mean, only the kernels in here,and as long as we continue to build the atomic system to exhibit intelligent action more,eventually we will, unless we interrupted,eventually we will build intellectual unlimited for their machines.Need to realize that progress is not important.because any process that is enough to get us into the end zone.Moore's law do not need to continue. I do not need to develop exponentially.We just need to keep walking.The second assumption is that we will continue to go.We will continue improving its intelligent machine.And based on the value of intelligence -I mean, intelligence or the source of everything we value or we need it to protect everything we value.It is the most precious resource of ours.So we want to do this.We have certain issues to resolve daredevil.We want to cure diseases such as Alzheimer's or the cancer.We want to master the economic system. We want to improve climate science.So we will do it if possible.Train has left the station, and can not pull the brake again.And last thing: we do not stand on the pinnacle of intelligence,or may be somewhere near it.And this is an important insight.This is what makes our state of instability,and it also makes our intuition about risk becoming unreliable.Now consider a most intelligent person in the world.Almost in your list will named John von Neumann.I mean, the impression that von Neumann created for everyone around,and it also includes mathematicians and physicists at the time great of him,handed down pretty good.If half the stories of his right only 50% incontrovertible he was the smartest person ever lived on earth.So let's review the distribution chain intelligence.Here is John von Neumann.Next is you and me.And then a chicken.Sorry, a chicken.No reason I should do this talk worse than necessary.However, it seems likely that very large distribution chain intelligence extends much farther than what we imagine,and if we make machines smarter than you,We may explore this distribution strip in a way we could not imagine,and far ahead of us in a way we could not imagine.And it is important to realize that this is true for the attractiveness of promptness.Right? Just imagine if we created a super-intelligent AI without talented group of researchers than average Stanford or MIT.Well, the electronic circuit operation about a million times faster compared with biochemical circuits,Should this machine can think about a million times faster compared brain that created it.Then you install it run for a week,and it will take 20,000 years to work at the level of human intelligence,week after week.Even how we understand, though less restrictive,a type of brain perform such a process?Another thing truly worried,That is, imagine a best case scenario.What we imagine finding a design for super intelligent AI without the safety issues.For the first time we get a perfect blueprint.It's as if we received a prophecy happened exactly as intended.Well, this machine will be a great device to help save manpower.It can design a machine capable of creating other machines perform physical work,powered by sunlight,more or less than the cost of raw materials.So we're talking about an end to the drudgery of man.We also talked about the cessation of mental tasks.So monkeys like we would do in this situation?Yes, we are free to play Frisbee and messaging to each person.Add some and few options for clothes suspicious,and the whole world will look like at Burning Man.Now, that might sound very good,but try to ask yourself to see what will happen under the economic order and our politics now?It appears that we will witness a distinguished level of wealth and unemployment never seen before.Lack voluntarily to immediately put this new wealth to serve all mankind,several "billion billionaire" was honored on the cover of the magazine Economic while the rest of the world are suffering from starvation The Russians and Chinese will do when they heard the news several companies in Silicon Valley preparing to deploy a super-intelligent AI?This machine has the ability to conduct war whether on the ground or on the network,with unprecedented strength.This is the prospect of "winning king".Six months ahead in this race equivalent to be 500,000 years ahead,at least that.So even though it looks just rumors for such groundbreaking style but was enough to get exasperated man.And now, one of the most frightening thing,I think at this point,is what researchers say about AI when they want to be assured.And time is the most common reason for us not to worry.You know, this is a long road.There may be 50 or 100 years.One researcher said,"Worry about the safety of AI as well as worries about a population explosion on Mars. "This is the version of the Silicon Valley on the "do not do your little head to worry about it."Seems no one noticed that a review of the scope of time is completely unfounded.If intelligence is only a matter of information processing,and we continue to improve the machine,we will create some form of super intellectual.And we do not know how much time it will take to create the conditions to do so safely.I repeat.We do not know how much time it will take to create the conditions to do so safely.And if you have not realized, never be the same 50 years This is 50 years in several months.This is the time we created the iPhone.This is the time "The Simpsons" was shown on TV.50 years is not too much to perform one of the biggest challenges that humanity has ever faced.Again, we seem unable to give an emotional response appropriate for which we have sufficient reason to believe it will happen.Computer scientist Stuart Russell has a very good analogy here.He said, imagine we get messages from alien civilizations,stated that:"Humans on Earth,We will respond to the planet of the people in 50 years.Prepare to go. "And now we are just counting down the months until the mother ship landed?We will feel much more urgent.Another reason tells us not to worry This is the machine does not do anything but share values ​​with usbecause we really will be an extension of us.They will be implanted in our brains,and we basically become our emotional system.Now consider that road safety and the only complete forward cautiously,proposed,is implanted directly into the brain of this technology ourselves.Now it is probably true that the road safety, the only prudent,but usually concern a person's safety technology must be calculated quite carefully before you attach it to his head.Deeper problem is to build a super-intelligent AI individual seem easier is to create a super-intelligent AI and get nervous sciences complete which allows us we integrate his mind and it.And assuming companies and governments do this may recognize themselves are in a race against humanity,assuming that this race is going to win the whole world,as long as you do not destroy it in the next time,it seems that no matter what easier will be completed before.Now, unfortunately I have no solution to this problem,except to suggest we should think about it more.I think we need something like the Manhattan Project the topic of artificial intelligence.Not because to create it, because I think we certainly will do,but to understand how to prevent an arms race and build it in a manner consistent with our interests.When we talk about super-intelligent AI can change,it seems we only have one chance to make the right initial conditions,and even then we'll need to do suffer the economic and political consequences of doing it properly.Moments we recognize that process information is the source of intelligence,that a suitable computing system is the foundation for the intelligence and we recognize that we continue to improve these systems,and we recognize that the horizon of perception can be very far compared to what we know,we must recognize that we are in the process of creating a kind of god.Now would be a good time to make sure that it is holy spirit that we can live together.
Thank you very much