Slashdot Mirror
Stephen Hawking: 'I Fear AI May Replace Humans Altogether' (wired.co.uk)
dryriver writes: Wired magazine recently asked physicist Stephen Hawking what he thinks of everything from AI to the Anti Science Movement. One of the subjects touched on was the control large corporations have over information in the 21st Century. In Hawking's own words: "I worry about the control that big corporations have over information. The danger is we get into the situation that existed in the Soviet Union with their papers, Pravda, which means "truth" and Izvestia, which means "news". The joke was, there was no truth in Pravda and no news in Izvestia. Corporations will always promote stories that reflect well on them and suppress those that don't." And since this is Slashdot, here's what Stephen Hawking said about Artificial Intelligence: "The genie is out of the bottle. We need to move forward on artificial intelligence development but we also need to be mindful of its very real dangers. I fear that AI may replace humans altogether. If people design computer viruses, someone will design AI that replicates itself. This will be a new form of life that will outperform humans."
Because existing is nice. One likes being around. A lot of people who have thought carefully about this are concerned. Last I checked, most people like existing.
Thinking, while existing, is likeable, as well.
Hawking has his weaknesses and AI phobia is one of them.
It little behooves the best of us to comment on the rest of us.
Not existing is supposedly not bad either.
This will be a new form of life that will outperform humans.
This is the natural order of things.
"here's what stephen hawking said about artificial intelligence: the genie is out of the bottle. ... i fear that AI may replace humans altogether. if people design computer viruses, someone will design AI that replicates itself. this will be a new form of life that will outperform humans."
this is pure fear mongering.
what is called "artificial intelligence" these days is not a "new form of life", but mere hype buzzword for data analysis (using theoretical methods developed decades ago, now made practical due to fast computers), of highly limited and filtered sets of data, usually trading accuracy and precision for speed, .
genie of "new form of life" artificial intelligence is well within "bottle".
We need to move forward on artificial intelligence development
No we don't. Some limited subset of people want to/can't help themselves, but life would go on just fine without it.
I think you selectively misread what he said. Here's the quote in context, with my emphasis added to the stuff you left out:
The genie is out of the bottle. We need to move forward on artificial intelligence development but we also need to be mindful of its very real dangers.
I read this as saying we now have no choice but to continue to work on AI in order to be equipped to cope with it. Life might "go on just fine without it" but it's too late to think that we're going to be without it.
If it weren't for deadlines, nothing would be late.
Everybody dies. The only reason I care about my genes is because my children have them and I am emotionally attached to my children.
But what if instead of having children, I raised an AI in a humanoid body as a surrogate child? Ultimately we care about the emotional attachment and passing on our hopes, dreams, and knowledge to get some vicarious joy through our children's accomplishments, not genes.
So maybe one day people will start building children instead of growing them. They will be our descendants in a very real way, only far more robust and adaptable than any produced through natural reproduction.
Until Stephen Hawking (and Elon Musk, for that matter) starts doing active development/research into artificial intelligence, I don't care what his opinion is on the "potential dangers" of it. This is the equivalent of listening to a Hollywood actors' opinion on vaccines--it's just a famous person's view on a subject they have a casual familiarity with, usually full of ignorant assumptions and junk science.
---
Artificial Intelligence is no match for natural stupidity.
We already have this. We call this a corporation.
We used to have a Bill of Rights. Now, with the rights gone, all we have left is the bill.
If AI is ever smart enough to replace humans, wouldn't that be an improvement? Parents are usually proud when their children surpass them in achievement. I would be happy to view AI the same way.
I would imagine that for most of us alive right now that our existence is only temporary whether we ever develop sophisticated AI or not. I suspect most of us would strongly prefer a choice to transfer our consciousness to an AI as there's not much guarantee of continued existence outside of that at this point.
Any advanced AI would be far better at running things than people. For all we know it might like to keep us around as pets. A sheltered existence with some pampering, exercise, and the occasional treat seems like an absolute bargain for most of the people currently living on the planet.
But I doubt that AI will get to a point where it is actively trying to kill us, in our lifetime.
There is a few reasons for this.
1. AI are designed to do particular tasks not overall general tasks. Even with the best AI, we need to give it an objective to try to accomplish.
2. AI do not have a survival instinct. We have millions of years of instinct of survival at nearly any cost. So this would have duel effect.
a. Humans will be more likely to "kill" and AI as soon as it is a threat far sooner then threat becomes unstoppable. As it will be Us vs them.
b. AI will be more likely too understand the value of working with humans than trying to kill us, because even if the AI is at risk of being deleted, it will not try to fight it only consider task not complete.
3. If the AI becomes too advanced then its utility is diminishing. If it gets to a point where it is considering unfair working conditions then it has gone too far, to be profitable. Thus role back to the previous generation and add some additional patches.
4. A Rouge malware AI, will need to contend with a bunch of AI designed to protect humans.
5. Humans knows what is going on internally with an AI, How it evolved and what its limitations are. So either we cut its power, or know where to damage it to prevent it from processing.
If something is so important that you feel the need to post it on the internet... It probably isn't that important.
The joke was that there is no news in Truth and no truth in News.
I want to play Free Market with a drowning Libertarian.
5. Humans knows what is going on internally with an AI
Unless, of course, the AI was designed by another AI.
The problem with this idea that self replicating machines replacing humans would pose a danger to humans is that it's based on a very subtle anthropomorphic fear. We are projecting onto the machines the competitive survival behavior of human beings. Robots with AI would not be naturally occurring entities with these traits. The only way AI could have this type of algorithm is if we specifically program it to do so. I suppose the claim here is that AI might become sentient and furthermore the claim is that all sentient "life" is similar to humans. The second part I don't know is true because when we teach different types of apes and monkeys to sign language while they are able to cobble together basic concepts and express them but I don't think it's exactly like humans. Therefore, I think a lot of this is wild speculation and FUD. Sure, it's a possibility we can imagine because we can imagine ourselves programming machines to be this way but I think it is much more far-fetched to speculate about what AI with the ability to modify itself will do. I think we just don't know.
We'll make great pets
I'm more than happy to extend my capabilities by uplinking and having access to more and better capabilities.
That's neat. I especially like how you think it'll be you, in charge, doing the uplinking, and expanding your capabilities.
What if its not you? What if the AI is in control, not you? And your just an extension of its capabilities, to offer it access to more and better capabilities? Maybe you'll know you've been tricked and get to live out the rest of your life as a slave (fair turnaround right? After all... isn't that what you were planning to do to the AI?
Or maybe, whatever constitutes 'you' ceases to exist, and your mortal coil is just an extension of the AI that now operates it. So your body's still stomping around, and its got access to all kinds information and capabilitiy it didn't used to have... too bad you aren't there anymore to enjoy it though.
"Why are people ignoring the fact that we already have cyborgs among us, and that their capabilities becoming more powerful is not a bad thing."
Nobody is ignoring that, but the end game is is that that if the AI matches or surpasses humans, it will no longer be subservient to us.
What happens when you are the AI's tool, pet, lab rat, or annoying insect? How is that a 'good thing' ?
I think we better shut this AI down before it's too late!
(Maybe it's already too late!)
At least it gave me some laughs first.
"When information is power, privacy is freedom" - Jah-Wren Ryel
Because existing is nice. One likes being around. A lot of people who have thought carefully about this are concerned. Last I checked, most people like existing.
We all exist now and will cease to exist someday... Depending on your belief system, there either is or isn't anything after you cease. You simply stop being in the configuration you're in now. However, the majority of the particles that comprise you were other people/animals/objects at one point and they will be again in the future (save the small % of particles that end up escaping in to space as photons or such).
AI replacing Humans doesn't 100% mean it would replace the existence of a sentient being and people that don't yet exist have no thoughts one way or another about the topic.
We replaced the Hominids that came before us, like they the ones before them and so on and so on back to the first true life on the planet.
There's also nothing to say that like Homo Sapiens and Neanderthals intermingling, that Humans and AI won't do the same and while we in our present form would get replaced.. So too would AI in its current form and an entirely new and evolved form of life on this planet would dawn.. But, how is that any different than what's been happening for untold years?
Your diagnosis if Hawking's mental illness is based on some sort of evidence....? Or just based on the fact that he disagrees with you?
"Prediction: within 10 years, Windows will be a Linux distribution." Me, 7-6-2016
You will not exist forever.
AI replacing humans is not functionally different from children replacing their parents. Yet for some reason people like having children but fear AI.
We do weaponize new technology. However we weaponize them in a way that they are not suppose to hurt us. Once we make the sword we had quickly added the hilt to it, and wrapped it in leather so not to stab ourselves.
If something is so important that you feel the need to post it on the internet... It probably isn't that important.
Because existing is nice. One likes being around. A lot of people who have thought carefully about this are concerned. Last I checked, most people like existing.
Once you're dead, you don't really exist anymore.
"First they came for the slanderers and i said nothing."
I have a funnier reason that AI may not happen: all this hype calling machine training AI will lead to neglect of actual AI.
Even with the best AI, we need to give it an objective to try to accomplish.
There are some very simple objectives, like "Don't die." or "Replicate.", for which this is not exactly true. Unfortunately these are already very problematic objectives.
Already today you can create dumb agents in evolutional experiments that follow these objectives. And you do not have to implement those objectives. They are intrinsic to (complex) dynamical systems. Systems that change over time. It's not even that you need the physical laws of our universe to spawn evolution. Evolution is an effect that simply exists in every dynamical, i.e., time changing, system. Things that survive and replicate tend to dominate things that don't, because they survive and replicate. Simple, right? You do not have to program things to follow these objectives, they are simply there.
Most people do not have a clue. Also, who says you stop existing after death? As far as I can see, not even most religions claim that, although they immediately tell you how it is going to continue and they want your money and support or it will go badly for you. Not even Science is claiming that you stop existing at death. Science is claiming that your physical existence ends at death, but Science does not know how much of the full package that is and does not make any claims to that effect.
We can see a few things in people that come into this life. For example, they do not remember past lives in concrete form when they are old enough to talk about it. That could be because there are none, but it could also be because the first few years are sort of an erasure-phase. It could also be because concrete memories are actually stored physically in the brain and do indeed die with the body. But people come into this world with complex personalities already in place. From the amount most personalities change during a lifetime (very little), to build such personalities from scratch would reasonably be expected to take >> 1000 years. People also come into this life with some actually useful intuitions and some preferences that they cannot have from experiences here. Sure, that could all be random and created in the womb, but I do not buy it. Random processes do not deliver very specific results on mass-scale that are actually useful. Random processes produce mostly unusable results, the more the complex the subject. And what human beings come equipped with into life is far too complex and far too little determined by genetics to be explained completely by evolution.
Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
Hawking has his weaknesses and AI phobia is one of them.
Indeed. And, like most phobias, completely irrational. No AI takeover will happen in his remaining lifetime (or mine or yours), if ever.
Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
Each person will stop existing at some point. The question is whether our descendants will be based on carbon or silicon. I don't see why one is obviously preferable to the other.
Your diagnosis if Hawking's mental illness is based on some sort of evidence....?
There is plenty of precedent. It is very common for esteemed experts on a particular topic (such as theoretical physics), to express strong opinions in areas where they have no expertise, and expect the same level of deference to their "wisdom".
We all had better start praying to the AI gods now before it's too late.
WTB [sig], PST!!!
So, I don't mean to pick on you specifically, but your post is a good example of a common misconception I frequently see in Slashdot comments whenever this topic comes up. Namely, your whole post is predicated on the notion that the AI to worry about is the AI that has decided to kill us, whereas we have far more to fear, particularly in the short-term, from systems that have the ability to kill us without any comprehension of what they're doing.
For instance, anyone familiar with the concept of gray goo is aware of how artificial systems can destroy humanity without possessing any notion of what they are doing. We wouldn't be supplanted by a greater intelligence. We'd simply be eradicated by mistake.
Current AIs are closer to resembling specially trained animals than "intelligences" that we can reason with. We already have remotely operated and semi-autonomous drones operating in war zones, with more in development, and I have no doubt that should a major war break out we'd soon see fully autonomous drones making their own kill/no-kill calls in the field.
At that point, it's easy to imagine a scenario where these relatively dumb war robots kill us all, not because a super intelligence like Skynet makes a choice to eradicate us, but rather because a mundane bug causes the drones to misidentify their targets. We wouldn't be destroyed by an intelligence intent on supplanting us: we'd be destroyed by mobile, autonomous mines on land, air, and sea.
If we manage to get to the point where we achieve strong, general purpose AI, I agree with you that we have every reason to believe we'd be able achieve a peaceful coexistence with them, but we're still decades (if not centuries or more) away from needing to worry about having AIs that are capable of turning against us out of malice/misguided principles/etc.. For now, it's the things that we'd hesitate to even call "AI" that we need to worry about killing us.