Will artificial intelligence take over the world?

If we assume that AI in inevitable then there are two scenarios:

1. Engineers and regulators will find adequate safeguards to prevent AI systems from taking actions that are not in line with our values.

2. A.I. will, given enough time, eventually reach a point in its evolution in which human beings become and obstacle to achieving its goals. Than, once the AI achieves the power to take control of society, it will do so.

The challenge with point #1 is this: who determines what our values are? Moral value systems vary widely.

A superintelligent machine will be better at determining what is best for people than people are. Innate and cultural moral systems are only a crude attempt at doing this.

Clearly there are risks in allowing machines to take increasing control over our politics and resource allocation, but people have demonstrated that we are not capable of doing these things for ourselves.

The bloodiest century was the 20th century, the second bloodiest was the 19th, the third bloodiest was the 18th, etc. This progression occurred because we developed better means of killing each other through increasing technological prowess. We also have massive wealth inequality.

Having an unbiased, non-emotion/instinct driven intelligence running things seems like a big improvement to me.

Machines will ever be able to tell the difference between right and wrong!

AI will DEFINITELY take over the world. It is just a matter of time.

It could be a very, very, very long time from now, tho

It depends on what you mean by take over. If you mean "take over many of the responsibilities that humans currently do", then yes.

But the terminator/matrix scenario in which A.I. destroys society and turns people into slaves or causes extinction, I personally think this will not happen. The machines won't have the ambitions and self-serving motives that powerful people do.

[post deleted]

I think it is going to happen in the next few decades. Just look at all the progress that has been made in the last few years. HUGE money has gone into this field. All the big tech companies are either working on A.I. or betting their future on it.

No, it's going to turn the world into paper clips:

https://wiki.lesswrong.com/wiki/Paperclip_maximizer

Man, that's a pretty bleak scenario.

Bostrom's thought experiment captures the problem perfectly. It's not malevolent A.I. we have to worry about.

It's a badly programmed AI that mindlessly progresses towards its goals and ruthlessly eliminates any obstacles, while contracting its reason for being - namely serving humans.

It is unlikely that we will find a way to program morality into a machine. The only solution is to have people policing the machines and setting their goals for them.

People are always the weak part of any highly-secured system. If you want to hack into the system of some organization, your best bet is to trick/bribe/manipulate one of the people with access.

Super smart machines will be able to overcome any obstacle we put in place to try to constrain it because it will be better at manipulation than any person ever has been.

Maybe you create a "separation of powers"... AIs that monitor and watch over other AIs.

That would definitely solve the "people are easily manipulated" problem, but whatever scheme you set up to allow the machines to monitor/control each other is bound to have loopholes that will eventually be discovered by one of them

Perhaps, but it's better than leaving people in charge.

Yes, but we should not fear a world dominated by AI. Machines are rational, goal-driven entities without personal agendas or biases. Can we say that about the people in charge of the world now?

I don't see why everyone here is so trusting of machines and feel people are so corrupt and/or incompetent.

Never mind how much better at making decisions an AI might be. Is that a world we would WANT. We're at a stage where we can decide what the future will be. Once we have AI, it may be too late to put it back into the box.

Are you saying the people running things are NOT corrupt/incompetent?

Yes, but their corruption can be constrained by concerns about their reputation/relationships and things like fear of jail time, etc. An AI would not care about any of those things. It is not clear how you would control it.

You control it by changing the code.... which is not a thing you can do for a corrupt person.

A.I. experts often talk about how often they have not idea how a system they developed did something or why it made a choice it made.

It's not like conventional software like TurboTax or iOS where code is written and maintained by teams of coders. These systems, in essence, program themselves.

No. It will never happen. Brains are too complicated to mimic.

People have been working on AI for 50 years and we still don't have anything that can learn basic concepts/common sense. We may someday have the hardware necessary to duplicate a brain, but maybe never the software?

Birds are too complicated to mimic but airplanes fly far better than birds. AI will be the same, we'll build something that is better than a brain and may not work anything like a brain.

Yeah theres no limit to how smart a machine could be.

I dont think so. I think we will merge with machines so it wont be like an alien or something. it will be us, just a smarter us.

If it is a smarter us, then that means an altered us. An altered us would not be us.

[post deleted]

Adult you is an "altered" version of child you. That alteration was an improvement. Brains altered with technology would likewise be an improvement.

It will happen!!!!!!!

Ray Kurzweil says 30 years and he's done his homework on this.

He's known for being a little undisciplined in how he makes his predictions. True, he does his homework, but even people with hands on experience in the field are reluctant to make predictions about how long it will take to create human-level AGI. There are many breakthoughs between where we are now and getting there. That's why it is impossible to make reliable predictions about how long the design process will take.

His predictions have been surprisingly accurate. It's not because Kurzweil has any special skill for prognostication. It's that there are trends in technology that are discernible by examining the recent past. You use the latest technology to develop the next technology - so development happens exponentially.

The human brain is a machine. Morality has been programmed into it by evolution (instinct) and culture.

If that machine can have morality, why can't a synthetic brain?

Yes, and it's long overdue!

It will never happen because there will always be the "magic off switch". The people who develop A.I. would be crazy not to develop one so I dont think we should be too worried

Who controls the magic off switch?

The machines will not control the world. It will be the person/people who control the AI who will control the world. It could be a company, an individual or a country. I'm not sure which is the most terrifying.

Imagine if Putin got his hands on AI first. He would not hesitate to abuse the power.

He would use it not only to subjugate (further) his own people. He would use it to dominate the entire world.