Stephen Hawking : AI could be our worst mistake

max

Legend
This is really the problem of our age---21st century civilization's greatest weak spot----the control of the direction of science.

Without direction, bad happens as well as good. Particularly if one can make a buck with it or annihilate your enemies.
 

max

Legend
Envy and jealousy, of course. :evil:

It's just brickhead jocks, who barely managed a C in HS math / science, getting in their still-impotent yuk-yuks at the eggheads. Same as it ever was.


It would be interesting to see if this kind of juvenile High School dynamics really ends up affecting scientific inquiry, to wit:

"Hey, don't do that! It's dangerous!"

"No, you're trying to restrain a Superior Mind, you witless C-student," sneers an arrogant technologist.

"No. I graduated summa from Columbia and was a Rhodes Scholar! It really IS dangerous to the world!"
 

rh310

Hall of Fame
Oh, what I was going to say before I got distracted was that Stephen Hawking's fears are surely misguided, even though he is a great man. Simply because AI is all about software, which runs on rather primitive inanimate objects. I think there is greater danger of his fears coming true if humans start creating computers based on biological or organic material which does not require explicit human programming, and uses something like DNA as its code... just thinking aloud here!

I'm afraid you are badly misinformed about the state of the art in AI.

On a related subject, Google up 'biomimicry' while you're at it.
 

RajS

Semi-Pro
I'm afraid you are badly misinformed about the state of the art in AI.

That goes without saying!

On a related subject, Google up 'biomimicry' while you're at it.

Biomimicry looks like a developing methodology to mimic nature - for example, studying humming birds and butterflies so we can build better flying machines, and so on. But the modeling of nature seems to use the usual methodologies, which are finite and predictable, albeit encompassing huge numbers of states.

The way we currently model intelligence is through a combination of brute force methods - for example, natural language processing, exhaustive database look up, correlating data points, pattern recognition, ad hoc heuristics that are problem dependent, and logical inference, of course. Since all of these methods lead to predictable outcomes - even with non-deterministic problem formulations - I have difficulty believing that with these methodologies we will accidentally emulate a life form. Of course, we could end up creating something dangerous, like a drone that takes off and starts shooting everything in sight - but I am sure that's not what Hawking had in mind.

However, once we start fiddling with real biological stuff, which we don't understand fully anyway, all bets are off! That's what I meant to say.
 

rh310

Hall of Fame
That goes without saying!



Biomimicry looks like a developing methodology to mimic nature - for example, studying humming birds and butterflies so we can build better flying machines, and so on. But the modeling of nature seems to use the usual methodologies, which are finite and predictable, albeit encompassing huge numbers of states.

The way we currently model intelligence is through a combination of brute force methods - for example, natural language processing, exhaustive database look up, correlating data points, pattern recognition, ad hoc heuristics that are problem dependent, and logical inference, of course. Since all of these methods lead to predictable outcomes - even with non-deterministic problem formulations - I have difficulty believing that with these methodologies we will accidentally emulate a life form. Of course, we could end up creating something dangerous, like a drone that takes off and starts shooting everything in sight - but I am sure that's not what Hawking had in mind.

However, once we start fiddling with real biological stuff, which we don't understand fully anyway, all bets are off! That's what I meant to say.

There are several research areas proceding independently that are very likely to merge up within a decade or so. Biomimicry, non-connectivist AI, robotics, brain-computer connectivity, network science, computational biology, ...

(As a sidebar: Seen this? http://www.youtube.com/watch?v=YQIMGV5vtd4)

Bear in mind that biomimicry is a new field. At this point it's essentially attempting as much as possible to duplicate evolved mechanisms in hardware. This starts with little dragonflies and robotic pack animals to carry heavy loads, but different groups are already revising artificial neural networks -- physical ANNs, not software -- to better model naturally occurring connectivity.

IBM is doing a great job in natural language processing essentially by throwing massive compute power at the problem. Brute force, yes, but Watson is a pretty interesting piece of work, nonetheless. I was invited to IBM Hawthorne a couple of years back to see where they are now with it. IBM is extremely ambitious. I think any validity of Hawkings' concern is rooted there. Massive compute power and encyclopedic data sets is one thing; such a savant, completely connected to the world and able to make sense and effect changes without direct human comprehension or control is quite another.

Rodney Brooks pioneered non-deterministic, non-connectivist, non-representative AI back in the late 80s. He built robots, not software, with these capabilities built in.

It's true that there are ample less impressive branches of AI research (Watson, again I'd agree, is brute force) but then I didn't read Hawking as saying that his cautions were due to any capability within our grasp today. More like what's on the horizon, should we head off towards one specfic direction.
 
Last edited:

RajS

Semi-Pro
@rh310: Stated that way, you are absolutely correct - I agree with you. The nano quadrotor video is fascinating! Thanks for posting.
 

sureshs

Bionic Poster
That goes without saying!



Biomimicry looks like a developing methodology to mimic nature - for example, studying humming birds and butterflies so we can build better flying machines, and so on. But the modeling of nature seems to use the usual methodologies, which are finite and predictable, albeit encompassing huge numbers of states.

The way we currently model intelligence is through a combination of brute force methods - for example, natural language processing, exhaustive database look up, correlating data points, pattern recognition, ad hoc heuristics that are problem dependent, and logical inference, of course. Since all of these methods lead to predictable outcomes - even with non-deterministic problem formulations - I have difficulty believing that with these methodologies we will accidentally emulate a life form. Of course, we could end up creating something dangerous, like a drone that takes off and starts shooting everything in sight - but I am sure that's not what Hawking had in mind.

However, once we start fiddling with real biological stuff, which we don't understand fully anyway, all bets are off! That's what I meant to say.

1. Neural networks are a big thing now
2. Even if the methods used now are deterministic, it only matters if the outcomes are beyond our grasp. That is why Turing proposed the AI test that if you cannot determine whether the thing behind the curtain is human or not by interacting with it, it is as good as human.
 

KineticChain

Hall of Fame
0fi3p.gif
 

dParis

Hall of Fame
Bear in mind that biomimicry is a new field. At this point it's essentially attempting as much as possible to duplicate evolved mechanisms in hardware. This starts with little dragonflies and robotic pack animals to carry heavy loads, but different groups are already revising artificial neural networks -- physical ANNs, not software -- to better model naturally occurring connectivity.

I hope science can do a better job with these "artificial neural netwoks" than they did with synthetic gut.
 

rh310

Hall of Fame
I hope science can do a better job with these "artificial neural netwoks" than they did with synthetic gut.

They've been used to approve credit card purchases and flag suspicious transactions for about 30 years now.

The US Navy uses ANNs to detect enemy submarines from sonar, sound, and other signals -- since ANNs are better at this than all but the top 0.1% of human experts.
 
Sadly,

Religion is, and could be our worst mistake.

Making AI or any kind of science seem bad/scary, compared to Religion
is hilarious.


edit: specifically Religious fantatics and anyone convinced they know what happens when we die, as fantastically described
here:
http://www.youtube.com/watch?v=55h1FO8V_3w

Sky Cake!
 
Last edited:

Sentinel

Bionic Poster
Sadly,

Religion is, and could be our worst mistake.

Making AI or any kind of science seem bad/scary, compared to Religion
is hilarious.


edit: specifically Religious fanatics and anyone convinced they know what happens when we die, as fantastically described
here:
http://www.youtube.com/watch?v=55h1FO8V_3w

Sky Cake!
What does it say, in brief ?

Don't you think religion will be on the wane, now ? I see the young generations in my country quite free of it. Although people are still in the clutches of other forms of division/ identities.

My fear is more of technology falling into the hands of politicians, and extremists/terrorists (as has been happening with weapons).
 

rh310

Hall of Fame
^^^ Yep, technology exists now that allows tryanny to be extended to what thoughts you express.
 

sureshs

Bionic Poster
Don't you think religion will be on the wane, now ? I see the young generations in my country quite free of it.

That is a great trend. In the US, the demographics is still tilted towards older people, who have been the products of decades of brainwashing, so the change is much slower. Till they die out one by one, change will be slow.
 

dParis

Hall of Fame
Comparison is not valid. All branches of science are not the same and do not progress at the same rate.

It is a valid comparison if you consider the possibility of incorporating the brain matter of cows in the grommets of the racquet thus creating an artificial neural network out of the stringbed. The racquet and the user will not be able to tell if the string is natural or synthetic and they can make the stringbed "moo" when hit in the sweet spot. You laugh, but...
 

Nostradamus

Bionic Poster
once we are able to download our mind into artificial positronic brain, will we still be the same person ? would we really transfer into the new brain ? will it be still us ?
 

Sentinel

Bionic Poster
once we are able to download our mind into artificial positronic brain, will we still be the same person ? would we really transfer into the new brain ? will it be still us ?
Forget that hogwash. There is no way you can download your brain into something else. We won't be able to even record the exact state of your brain at any point of time. Even trying to do so will alter it.

I am not sure this is just a technical issue.

No, the person in the body will still think it is him in the body. That is even if you could get a perfect copy of each cell of your brain into some machine/computer/drive. The problem only is (if you saw Moon and various other flicks that deal with clones that have the same memories) that the clone will think it is you, and interfere in your life. For that you don't need to even download a copy, you only have to implant some memories.

You will not transfer into something else.
 

rh310

Hall of Fame
Is this where one is supposed to say "Lolwut?" ?


If you combine the tip of the iceberg that Edward Snowden has documented with a historical review of tyranny and abuses of power...you might say "LOL WUT?" but I don't find it either funny or confusing.
 

Sentinel

Bionic Poster
If you combine the tip of the iceberg that Edward Snowden has documented with a historical review of tyranny and abuses of power...you might say "LOL WUT?" but I don't find it either funny or confusing.
Pardon me, but I was confused.
 

sureshs

Bionic Poster
No, the person in the body will still think it is him in the body.

I think Nostradamus meant if you died but your brain was transferred some place else.

In the movie Transcendence, the brain is uploaded into a quantum computer. One of the people protesting against AI (in the movie) says she became an opponent when her Professor did such a thing to a monkey, and the monkey brain in the computer felt imprisoned inside and screamed in pain, till the machine was shut off. Imagine yourself serving life imprisonment in a small cell. Many such people go mad. The brain works, but there is nothing to engage it. So now you have this fully simulated brain inside a quantum computer, which means it can accurately reproduce all physical effects with absolute accuracy, so is the entity inside the computer living or not?

On a related note, there was a Russian scientist who transplanted heads of dogs to other bodies (of dogs). His experiments were only partially successful. Then, one scientist in the US managed to transplant the head of a monkey. It was a botched affair, but the "head" remained alive for several days and whenever it woke up, it would be utterly confused and its facial features showed that it was experiencing immense pain. That guy's colleague said it was inhuman and should never be done again.

Re: not being able to completely capture state - that is true because of quantum behavior. But is it really a problem? If the information is stored away in the brain in a non-quantum way, then why can't it be uploaded? A computer memory operates on quantum principles (tunneling) but once the bits are written, they can be read and copied. The only things that will be missing will be due to the quantum effects during the copying period. Now, reload the copied information into a quantum computer and start running it again (it = both "data" and "code" which are intertwined in the neural networks). Is the program alive now?
 

Nostradamus

Bionic Poster
i got a way to live forever

Forget that hogwash. There is no way you can download your brain into something else. We won't be able to even record the exact state of your brain at any point of time. Even trying to do so will alter it.

I am not sure this is just a technical issue.

No, the person in the body will still think it is him in the body. That is even if you could get a perfect copy of each cell of your brain into some machine/computer/drive. The problem only is (if you saw Moon and various other flicks that deal with clones that have the same memories) that the clone will think it is you, and interfere in your life. For that you don't need to even download a copy, you only have to implant some memories.

You will not transfer into something else.

Yes I get what you mean. so only way we can Truly transfer or Live forever is if we get a NEW BODY. and put the old brain into the NEW body, correct ?
but then we still have to re-generate the old brain since Brain cells get old too. I think this would be very possible in the future. Transferring the brain into new body maybe tricky but with New NANO tech, it could be done.

1. Clone a brand new body of yourself--- of course, fix any genetic defects while at it.
2. take out the old brain and put it in new body and re-attach all the neurons using the Nanobots.
3. then use drug or nanobots to re-generate the old brain cells and repair any damages.
4. Walla,,,,,,,,,you got a NEW life and still it is YOU.
 
Last edited:

LuckyR

Legend
Hawking is obviously underestimating the amount of human attention that silicon life needs to remain "alive".

Human programmers have an amazing capacity to create very simple programs that bring extremely complictaed systems to a halt.

And that's just the software side, the hardware challenges are also daunting, and I'm just talking about maintenance, let alone open sabotage...
 

Nostradamus

Bionic Poster
Hawking is obviously underestimating the amount of human attention that silicon life needs to remain "alive".

Human programmers have an amazing capacity to create very simple programs that bring extremely complictaed systems to a halt.

And that's just the software side, the hardware challenges are also daunting, and I'm just talking about maintenance, let alone open sabotage...

correct but once evolved to intelligent AI, it will be able to fix and adjust any damages or sabotage that goes on by itself. all intelligent machine can do this.
+ once it is that advanced, intelligence will also advance and will be able to outthink and outreason any humanoids.
Developing emotion will be next step as well.
Look at DATA on Startrek, he can diagnose and fix any malfunction and evolve faster any human being there is.
 

sureshs

Bionic Poster
Hawking is obviously underestimating the amount of human attention that silicon life needs to remain "alive".

Human programmers have an amazing capacity to create very simple programs that bring extremely complictaed systems to a halt.

And that's just the software side, the hardware challenges are also daunting, and I'm just talking about maintenance, let alone open sabotage...

That is true today. But when robots start writing software and making other robots, things will change.

And what about human-robot hybrids? These will be humans with computers attached to them which will make them super-intelligent. We are already there to some extent. People with prosthetic limbs would not have been accepted a few centuries ago. Even something as simple as a pair of spectacles or as complex as a pacemaker is a machine which enhances your natural capabilities. You are not the same person without glasses, because you see differently, which means your brain is processing information differently, which means it is altering your personality.
 

Nostradamus

Bionic Poster
That is true today. But when robots start writing software and making other robots, things will change.

And what about human-robot hybrids? These will be humans with computers attached to them which will make them super-intelligent. We are already there to some extent. People with prosthetic limbs would not have been accepted a few centuries ago. Even something as simple as a pair of spectacles or as complex as a pacemaker is a machine which enhances your natural capabilities. You are not the same person without glasses, because you see differently, which means your brain is processing information differently, which means it is altering your personality.

also Nanobots inside the robot or android will constantly be repairing any damages or flaws that arise. robots can evolve faster than anything we have encountered in evolutionary history.
This is what Hawking was talking about. Robot can evolve so fast that if it decides to take over Humans won't be able to stop it
 

LuckyR

Legend
Oh of course these sorts of things are completely possible, but when? Human programmers can't even make virus proof operating systems. I think self aware, emotional, self repairing nanobot using, Rare Earth mining singularities that not only need zero human input but are immune to open human attempts at sabotage are not going to be available at WalMart for this XMas season...
 

Nostradamus

Bionic Poster
Oh of course these sorts of things are completely possible, but when? Human programmers can't even make virus proof operating systems. I think self aware, emotional, self repairing nanobot using, Rare Earth mining singularities that not only need zero human input but are immune to open human attempts at sabotage are not going to be available at WalMart for this XMas season...

labs in MIT are already experimenting with NAnobots and positronic brains.

Also our government is trying to reverse engineer the bio-nono computers that they stole from the alien craft, but with limited success, I might add.
 

Nostradamus

Bionic Poster
Yes and many labs are experimenting with genetic modifications, yet XMen is still in the Fantasy/Fiction section...

yes. but we are very close to enhancing the human mind and body thru Genetic manipulations in our DNA and Gene sequences.
Meaning making a baby with IQ of 200 or very strong body with no genetic flaws. Might even be able to produce a Federer baby with his talent by manipulating the gene sequences.

Not sure if there will be law or regulations against this act, however.
 

LuckyR

Legend
yes. but we are very close to enhancing the human mind and body thru Genetic manipulations in our DNA and Gene sequences.
Meaning making a baby with IQ of 200 or very strong body with no genetic flaws. Might even be able to produce a Federer baby with his talent by manipulating the gene sequences.

Not sure if there will be law or regulations against this act, however.

Obviously you and I have a very different understanding of the phrase "very close".

We were "very close" to having flying cars in 1957...



30602d1393383029-flying-car-marketing-popular_mechanics_flying_car_cover.jpg
 

max

Legend
I think it's always interesting to run into people SO excited by "technology" or at least the promises made for something before it's made and seen, that they neglect to really explore the thing.

That's how animals get stuck in traps.
 

Sentinel

Bionic Poster
Yes I get what you mean. so only way we can Truly transfer or Live forever is if we get a NEW BODY. and put the old brain into the NEW body, correct ?
but then we still have to re-generate the old brain since Brain cells get old too. I think this would be very possible in the future. Transferring the brain into new body maybe tricky but with New NANO tech, it could be done.

1. Clone a brand new body of yourself--- of course, fix any genetic defects while at it.
2. take out the old brain and put it in new body and re-attach all the neurons using the Nanobots.
3. then use drug or nanobots to re-generate the old brain cells and repair any damages.
4. Walla,,,,,,,,,you got a NEW life and still it is YOU.


suresh, i have not read your post since i intend seeing Transcendence. I did see the later paras though.

nostra, the body part is not the issue. I don't mind even being in a robotic body (much better in fact, with replaceable parts). No need to clone me -- i have no idea of your nanobots. but i am told that it is quite easy to take aside a human heart and reattach it, but the human brain has millions or billions of connections. you cannot just connect it like that.

Brain cells do grow old, and perhaps one way i heard of is to store stem cells when young, and use them to regenerate fresh brain cells. Something like that.

suresh, your last para. what you are coming down to is a perfect or close to perfect copy of your brain. And then asking is it you ? Am i right ? If there is a technology that can make a perfect replica of you, atom by atom, at whatever level you want, while your are alive, will you still think that is you? If you don't know it was created, will you ever know it existed.

I think there is some process or activity or whatever, that is not my thoughts or memories that feels alive. You have it, I have it. By putting my/ your memories and other state into someone who feels alive (a computer does not), that person will suddenly not become you. So what to say of a computer.

Is Deep Blue alive ? Certainly not.
 

rh310

Hall of Fame
Oh of course these sorts of things are completely possible, but when? Human programmers can't even make virus proof operating systems...

Sure they can. That's a long-solved problem. What they can't do is make one that's both virus-proof and backwards compatible with all the software the masses already use.
 
Last edited:

Nostradamus

Bionic Poster
suresh, i have not read your post since i intend seeing Transcendence. I did see the later paras though.

nostra, the body part is not the issue. I don't mind even being in a robotic body (much better in fact, with replaceable parts). No need to clone me -- i have no idea of your nanobots. but i am told that it is quite easy to take aside a human heart and reattach it, but the human brain has millions or billions of connections. you cannot just connect it like that.

Brain cells do grow old, and perhaps one way i heard of is to store stem cells when young, and use them to regenerate fresh brain cells. Something like that.

suresh, your last para. what you are coming down to is a perfect or close to perfect copy of your brain. And then asking is it you ? Am i right ? If there is a technology that can make a perfect replica of you, atom by atom, at whatever level you want, while your are alive, will you still think that is you? If you don't know it was created, will you ever know it existed.

I think there is some process or activity or whatever, that is not my thoughts or memories that feels alive. You have it, I have it. By putting my/ your memories and other state into someone who feels alive (a computer does not), that person will suddenly not become you. So what to say of a computer.

Is Deep Blue alive ? Certainly not.

I know billions of neurons could be reattached if you have the nano techknowldege. if you just clone your entire body including the head, then that is not you. it is replica of you but it is not you. and I don't think we will ever learn how to transfer the consciousness to another body
 

RajS

Semi-Pro
@Suresh, Sentinel, Nostradamus: And thus, my dear friends, you have all arrived at the philosophical problem known as the Paradox of Theseus, also called The Ship of Theseus!
 

RajS

Semi-Pro
@Nostradamus: Theseus' Paradox raises the question of whether an object which has had some or all of its components replaced is still the same object. Coming up with an answer as it applied to an inanimate object such as Theseus' ship itself raised many issues in philosophy, so asking the same question about living entities and consciousness has got to be somewhat more challenging, lol. It is interesting and thought provoking to discuss these things, but there are no easy answers, sadly!
 

Doubles

Legend
@Suresh, Sentinel, Nostradamus: And thus, my dear friends, you have all arrived at the philosophical problem known as the Paradox of Theseus, also called The Ship of Theseus!

In all my time studying philosophy, but in university and before, The Ship of Theseus is arguably my favorite thought experiment.
 

RajS

Semi-Pro
@Doubles: I wish I had studied Philosophy more deeply. But I am trying to make up for my lack of erudition in these matters. This particular problem grabbed my fancy also and led to some pleasurable moments spent in cogitation. Those philosophers sure had a lot of fun - at least, until some king made them drink hemlock or some other horrible thing!
 

rh310

Hall of Fame
In all my time studying philosophy, but in university and before, The Ship of Theseus is arguably my favorite thought experiment.



I'm not so sure it's strictly a gedanken -- isn't the material in our bodies completely replaced over seven year cycles? Are we the same person, over the course of our lives?

Interestingly, there are a small number of cultures that consider people to be completely different at various points in their lives, and that the old person no longer exists. They're given a new name. The new name describes who they are "now."
 
Last edited:
Hawking is just jealous that he was not part of the AI revolution. He is also afraid that future computers will make human geniuses like him obsolete. We are accustomed to thinking of scientists such as himself as some sort of great people, but in reality they are nothing more than people with more and better neural interconnections than other people, which can be easily achieved by artificial means in the future.

He is also probably one of those who cannot see the bigger connection here between AI and medicine. An AI system much superior to humans may come up with a cure for his disability.

But why would it. I'm not sure if AI will be possible but if it is real artificial INTELLIGENCE why would it want to act as a slave for humanity?

If we asume that AI can move past being a pure calculation machine and evolves true learning ability and a free will it would be pretty stupid to not emancipate from humanity or even possible make humanity slaves to themselves.

any intelligence will eventually enslave nature and other people (just look what humanity did to nature and each other) and if AI is truely intelligent it will do it too. as soon AI is smarter than humanity it will control humanity (again asuming true AI is possible which I'm not sure about).
 
Last edited:
Top