How Much Longer Before Our First AI Catastrophe?

What will happen in the days after the birth of the first true artificial intelligence? If things continue apace, this could prove to be the most dangerous time in human history. It will be an era of weak and narrow artificial intelligence, a highly dangerous combination that could wreak tremendous havoc on human civilization. Here’s why we’ll need to be ready.

io9

 

"Our infrastructure is becoming increasingly digital and interconnected — and by consequence, increasingly vulnerable. In a few decades, it will be brittle as glass, with the bulk of human activity dependant upon it."

A few decades aren't needed. The Internet is already here, it's a giant block of Swiss cheese, and humanity at this point would collectively plop down on their behinds and stare vacantly in complete befuddlement if it were to be taken away. Anyway, let's hope we don't end up with HAL-9000 (even though if you look at the current state of robotics, especially testing in Japan, human behaviors are being reproduced by machines with great results).

 

Without wanting to get all Vernor Vinge about this, I also wonder if many of these fears are unfounded. I can remember all the paranoia & hysteria about the Millennium Bug. That turned out to be a bit of a damp squib.

 

Probably interesting to see what will happen by then but I'm not too worried. I don't even know if I will live to see the day. As it is, we mankind are already a catastrophe upon ourselves. If anything, we'll survive and we'll sort out the mess if it happens. If we don't, all I can say is we deserve it since we let it happen.

 

Skynet has assured me there is nothing to be concerned about!

Seriously, I do wonder about it sometimes. Just not sure exactly how it would turn out...

 

I don't ever see us having to deal with an honest to God "Skynet" consisting of ~ Snipped as per TOS ~ Robocops taking over humanity. Humans are dangerous, unpredictable and power-tripping all by their little selves. However, if you really pay attention to them, movies like Terminator, 2001: A Space Odyssey and others, you'll see that..minus the mad robots..they are warnings to not try to make machines smarter than us, to not let them do all of our work. They were warnings to not head too far in the direction we're actually headed in now.

Robotics, AI and computers should be treated as simple tools, as assistants and not replacements. Connect hospital databases to the net to share records and information with staff across states and countries. Don't hook up life-support systems and patient monitoring systems to it. Program drones to fly into situations too dangerous to support humans, but don't program a massive system to run the military automatically without humans. We have to be careful with our attempts to further efficiency. Humans have "ifs, ands or buts", computers don't always. And when systems that don't have these screw up, trouble comes.

 

Try "never," IMO. Human intelligence (such as it is) took about a billion years of evolution to arrive at. We still don't know a lot about our own brains, we only have a vague outline of the processes involved in cognition.

Letting an AI evolve is out of the question right now, we'd have to design one. Which would be pretty much impossible with our currently incomplete knowledge of cognition.

(I know, I know, a bunch of neurons can bootstrap to something resembling a brain. But neurons aren't electronics, they have a heck of a lot more programming in them - a billion odd years worth of programming. We are not going to see the Internet bootstrap itself into independent consciousness.)

I don't think intelligence is an "emergent property of carbon" or any such stupidity mind. I just think it's very hard to arrive at, especially by design, and most especially when we don't know what we're doing.

That said, I also think there a certain dangers in relying too much on fast-but-dumb-as-a-brick expert systems, which we are doing a lot right now. Those are more the kinds of dangers I'm inclined to take seriously; I think Mman79 (above) explained that stance pretty well.

 

Computers are essentially difference engines anyway. I like to play chess with my computer regularly, often beating it, but I am under no illusion that it is ever going to pass the Turing Test, let alone be 'intelligent'.

I don't believe AI actually exists (not properly anyway), & although I agree that we shouldn't become over reliant on technology, I doubt anyone outside of a Sci Fi novel/movie will ever actually build a SkyNet.

 

This will never happen.
Mrk

 

"Profiling" in conjunction with automated hi tech AI (drones, robotics, heat, attitude, "threat" monitors etc) is the biggest threat to our lives as free humans on this planet. Orwellian nightmare on the way if we are not careful. Imagine a cash machine with a particular political agenda that monitors you during a bank crisis and passes that information on to a profiling AI and grades your threat level accordingly in all areas of life.

 

Imagine a cash machine that you disconnect from electricity.
Mrk

 

Lord Kelvin once remarked the heavier than air flight would never happen...

(You're probably right though.)

 

Yes, but the human mind isn't one of those areas that we have a little extra to go on to be able to study. We get why brains do what they do physically because we have the study of physics to go on (even if we don't know very much in that area either). But when it comes to reasoning, emotions, dreams and so on, we don't have the first danged clue. Of course a Skynet-like existence wouldn't need those, so there's the bad news, lol. The good news is that I think there are enough sensible people in the world left, even if they are the worst of the worst, to understand why such an existence would be a very bad idea.

I personally think drones are about as close as we'll get to any resemblance of such a world, and really that's uncomfortable enough.

 

AI has one big problem - determinism round 50:50 chance.
Where do you draw the line - down the floating point limit?
And what if it's an EXACT 50:50 within its precision limit, what then?
Kernel panic.
Mrk

 

I don't understand. If both possible branches are of exactly equal weight, wouldn't it be possible to pick one at random?

 

That's where machines fail. If you pick at random, you lose.
What about intuition, instinct, experience, suggestion from a friend.
Mrk

 

Given the (some would say) relatively poor track record of natural intelligence, who is to say that "artificial" intelligence will get us in so much trouble?
Any likelihood that it might go the other way, and lead us into a better world?
I don't think so.

 

Those apply to situations where the decisions are not evenly weighted, but you don't know it based on currently available information. For a truly 50/50 decision a human wouldn't do any better.

Also I see no reason why a computer program couldn't theoretically have an unconscious mind, or way of recording prior experiences, or an idea of when to ask a human to make the call. All modern computers are Turing-complete, simulating this stuff is not impossible (unless you're a Cartesian dualist or something); the problem is
a) we don't know how to do it
b) even if we did, our computers are too weak by absurd orders of magnitude
c) even if they weren't, the work required to do it would be stupendous
d) even if someone were willing to do the work, what would the point be?

 

some part is good of artificial intelligence but like mrk said 50:50 combo is good the life would be like movie Idiocracy if we depend on machines 100%

check that movie if you dont know

 

Considering that, as has been pointed out here several times, the functioning of the human brain is a mystery even to experts, I can't visualize anything workable coming along in AI in the foreseeable future. And that's totally leaving out the question of whether or not it would be benevolent and useful.

I'm a brain injury survivor, took a bad fall down the basement stairs about 18 years ago and my skull lost the ensuing disagreement with the concrete floor, fractured skull and quite a bit of blood, a classic "open" head injury. It was so uncertain whether I'd survive that the police took my family to the hospital by a "scenic" route so there'd be time to have me tucked away wherever before they arrived. At that I was lucky in the nature of the injury, it was obvious immediately to the ambulance crew long before we got to the hospital ... far too many people have had brain injury from "closed" head wounds and not gotten properly diagnosed for years if ever.

I was in coma for four days, and have total amnesia of the approximate six weeks afterwards -- fortunately, from what I've been told. And I learned that one of the ways you can distinguish a good, expert neurologist is by his copious use of "typically", "often", "hopefully", "we'll have to wait and see", and the like, as opposed to things like "this is wrong, and this will fix it".

That's why I'm so skeptical about the future of AI -- there's no such thing as a "prosthetic" for the human (or for that matter any other living organic) brain, and I suspect that would probably have to be the first step in advancing towards AI. Probably the closest we've ever come so far was the "Eliza" program (remember that?), which could emulate interactive conversation with a human and could probably give even a modern-day Turing test a good run for the money.

 

I think this article can be safely filed in the cabinet along with those other titles such as:
'Don't open your closet door.....there's an alien inside!'
'The day the Earth was swallowed by a giant jellyfish from the 9th dimension!'
'Earth invaded by green troglodytes from the planet Dork'!...etc

Visions of HAL – Skynet – the Matrix type AIs are fanciful and belong in the realms of sci fi fantasy. IMO the closest we will probably get is a programmable machine – just more sophisticated than at present. If such AIs make errors it will be mostly due to poor\inadequate programming.
AI with the capacity to become completely independent regarding thought, reasoning and action would require developing a machine into a human being.

An AI might well ask.....where's the logic in that?

 

Krysis ... well said.
If the half wits of the 1950s had have left Alan Turing alone then he would no doubt have expanded on his original ideas about AI. His journals spoke sparingly about AI after WWII but he had a clear vision of its potential. He was a realist, a visionary and an eccentric, however I think he would have found the doomsday cultists of today amusing. He would have been able to articulate the application of AI as a mathematician with a realistic awareness of human nature.

 

Biotechnology. Alive computers composed of brain cells could be dangerous, because they could create neurons and synapses on their own, in other words, they could evolve.

 

One scientist (J. Craig Venter) has created "synthetic life" form already from a bio-genetic (cell level DNA) point-of-view.

All it will take would be a bio-computer with the intelligence of a human and the computing speed of a yotta scale computer ( 10^-24 ). We've already crashed through the ( 10^-18 ) barrier, and someday will no doubt get to planck level scale computing ( 10^-35 ). And maybe somewhere in-between we may even get to the quantum computer which would crank up the speed.

No doubt things will evolve in this field.

-- Tom

 

Chips becoming or thinking like humans while humans being chipped and humans animating cybernetics or any host like in the movie avatar, these are the Globalist transhumanists' goals.