Slashdot Mirror
Elon Musk Says Mark Zuckerberg's Understanding of AI Is Limited (ndtv.com)
An anonymous reader shares a report: Elon Musk is a man of many characteristics, one of which apparently is not shying away from calling out big names when they are not informed about a subject. A day after Facebook founder and CEO Mark Zuckerberg said Musk's doomsday prediction of AI is "irresponsible," the Tesla, SpaceX, and SolarCity founder returned the favour by calling Zuckerberg's understanding of AI "limited." Responding to a tweet Tuesday, which talked about Zuckerberg's remarks on the matter, Musk said he has spoken to the Facebook CEO about it, and reached the conclusion that his "understanding of the subject is limited." Even as AI remains in its nascent stage -- recent acquisitions suggest that most companies only started looking at AI-focused startups five years ago -- major companies are aggressively placing big bets on it. Companies are increasingly exploring opportunities to use machine learning and other AI components to improve their products and services and push things forward. But as AI is seeing tremendous attention, some, including people like Musk worry that we need to regulate these efforts as they could pose a "fundamental risk to the existence of human civilisation." At the National Governors Association summer meeting earlier this month in the US, Musk added, "I have exposure to the very cutting edge AI, and I think people should be really concerned about it. I keep sounding the alarm bell, but until people see robots going down the street killing people, they don't know how to react, because it seems so ethereal." Over the weekend, during Zuckerberg's Facebook Live session, a user asked what he thought of Musk's remarks. "I have pretty strong opinions on this. I am optimistic," Zuckerberg said. "And I think people who are naysayers and try to drum up these doomsday scenarios -- I just, I don't understand it. It's really negative and in some ways I actually think it is pretty irresponsible."
Zuckerberg is just a glorified webmaster from the 90s, when you think about it.
Everyone's understanding of AI is limited.
Ooooh, goodie: a billionaires' pissing contest. And between techies, too. I'll get the popcorn!
I think Elon Musk is the one that has either a limited understanding of current AI technology or just hypes AI on purpose, while being fully aware that AI still has major limitations and they are unlikely to disappear within the next few years. Important and very important progress has been made, but General AI is likely still very far away.
Facebook's director of AI Yann LeCun gave a very good interview to IEEE spectrum: Facebook AI Director Yann LeCun on His Quest to Unleash Deep Learning and Make Machines Smarter
Jan
Both of them display a remarkable lack of knowledge about the limits of AI. Their respective knowledge about AI is purely from works of fiction, which is why at least one of them has, for the last five years, been bleating that self-driving cars are only five years away.
I'm a minority race. Save your vitriol for white people.
The guy who ignores the fact that no one is currently researching strong AI accuses the guy who uses actual AI (well, enhanced pattern matching really) of having a limited understanding of the subject??
Lets face it: To have killer robots and the like as Musks imagines, we need to have a strong AI, of course we need to also have it go off the rails for some mysterious reason (whatever reason that causes this behaviour in movies wont cause it in reality), but first of all we need human-like AI.
This is a bit of a problem as we only have very, very limited understanding of natural Intelligence and no plan or clue how to even start implementing artificial intelligence. People have been falling for the "ZOMFG AI nau!" hype since the creation of Eliza. But so far neither the formal knowledge systems pre-AI-winter nor the deep learning and neurnal nets approaches have yielded anything more than very sophisticated pattern matching algorithms.
Take Googles Go engine: Impressive but it can play Go and only Go. Proof: they now have to spend a long time to retrain it for other tasks. This is not at all what the general population (and Elon) understands by AI.
You had the same kind of thing with nano technology.
Everyone was worried about Grey Goo and being stuck with the kids while your wife, whom you suspected of cheating because she worked long hours at her engineering job wearing shoes that did not meet Speaker Ryan's dress code, was in fact becoming a nano-bot zombie.
Instead, nano was a term you needed to sex up your NSF proposal and pitch to private capital investors, but what you were doing had nothing to do with Drexler assemblers and pretty much mass fabrication materials tech.
I only read a preview of this book When Computers Can Think: The Artificial Intelligence Singularity. They give a good overview of current accomplishments in the field and the logical prediction of the trend. The AI already beats us in Chess, Starcraft and (the scariest one) Rock Paper Scissors. Why is it scary? Because you cannot defeat robots that are both more intelligent than you and move orders of magnitude faster than you.
... All the way to the bank.
I find it hard to believe the CEO of one of the largest tech companies in the world, whose services heavily rely on AI for recommendations, image recognition, etc, has a limited knowledge of the AI industry. I'm not saying Zuckerberg is one of the world's experts but he most likely has a very firm grasp on the subject.
And whether or not Zuckerberg is correct, it is certainly a reasonable opinion that those who drum up negative sentiment towards AI research are acting irresponsibly. It isn't to the level of Edison spreading fears about AC current by electrocuting animals, but spreading fear about new technologies is likely not a good thing. Instead of more reasonable debates over AI caused displacement of jobs or privacy concerns, Musk is doom-saying about a robot apocalypse. I wouldn't use the term irresponsible, but it's close enough to me to not disparage those who do accuse Musk of irresponsible behavior.
-- All that is necessary for the triumph of evil is that good men do nothing. -- Edmund Burke
Yes, the article linked in the original post paints the whole thing in a pretty shallow manner which seem likely to mostly draw out gut reactions to both of them. The truth is that this is a very important and nuanced topic. Surely both of them realize that the possibility of machine intelligence that can upgrade its own capabilities rapidly represents a totally new era. And that likely humans will not do well in this era. But everyone's understanding of AI is extremely limited. We don't know how far we are from making machines that can upgrade their own intelligence, and we know very little about the diversity of kinds of intelligence that are possible. Hopefully the public conversation can go deeper than pitting two tech superstars against each other.
Who is it that's afraid of AI? Highly successful people IN OTHER FIELDS. Bill Gates, Elon Musk, and Stephen Hawking.
People like this are used to bucking trends and being right. The press pays attention to them because they sell newspapers. But then they go completely outside their field, and somehow still expect to be right.
Dear Elon, take a lesson from Linus Pauling who for years and years pushed nonsense on us about Vitamin C being a wonder drug. That didn't turn out to be the case, but Pauling stuck to his guns for years despite evidence he was wrong. Don't be Linus Pauling Elon.
I think Elon Musk is the one that has either a limited understanding of current AI technology
Musk isn't talking about current technology.
Personally I think the answer lies in the middle. Machine learning will be dangerous when it replaces human decision making as people look at it as "the answer". Humans inherently do not want to be held accountable for their decision making and by passing liability to an algorithm we will get even more of the worst side of humanity: obliviousness and apathy.
AI doesn't have to come to the conclusion that all humans have to die. It simply to has to cross the line of killing one person and then be able to repeat it. Assuming it can replicate and resist our efforts to thwart it, once killing humans becomes a part of how it does things we will no longer exist. How many microscopic animals have you killed today? Do you worry about their dreams and feelings or just inhale them like the rest of us?
I'm not saying Elon is right about AI but I agree that Zuckerberg has a limited understanding of AI. I would go further and say that most everyone has a limited understanding of AI. It's also part of why I think we're safe from AI for many decades. It will eventually become a problem but by that time people will have a much better grasp of the danger that AI can present and be looking to use AI to secure systems from people and other AI.
Anons need not reply. Questions end with a question mark.
Both Elon and Zuck can pay people for their understanding of anything. It is just poo flinging of rich monkeys.
And I think Elon Musk is from the future and has seen the devastation future A.I. will be capable of.
He's working on trying to save us from extinction from pollution (Tesla and Solar City), extinction from a surface-wide planetary event (his new Boring Company could be used to build massive underground bunkers), extinction from a planet-wide event (SpaceX, Mars colonisation) and extinction from predatory A.I. (what we're talking about).
Either he's a mad man with too much money, or he's from the future and knows what's in store for us and trying to prevent it. But a mad man wouldn't be working on so many "random" things that all lead to the salvation of mankind.
Or maybe I'm the mad man. Who knows!? Not me!
#DeleteFacebook
"I don't understand it. It's really negative and in some ways I actually think it is pretty irresponsible." - says the guy who's company has more deeply invaded the privacy of individuals, as well as the most people in history..I don't think he's a very good judge of what's 'irresponsible'.
Sure, but if you look at the limitation of current technology it is easy to figure out that there is still a huge number of problems to solve, many of them where nobody so far has any clue how to solve them. It's likely not just a matter of a few more years of research and throwing even bigger datasets and computers at the problem. Sure, you can make up any projections about the future and no matter how crazy they seem, we won't know that they are wrong until we are in the future. But are Elon Musk's style projections about the future of AI likely? No, not really. Yann LeCun is clearly an expert in AI, while Musk is a business man. Hypeing AI helps to finance Musk business and keeps the stock price high.
Jan
Indeed, Musk is using the strategy from the master himself: Wayne Gretzky.
"I skate to where the puck is going to be, not where it has been."
#DeleteFacebook
Of course Zuckerberg's understanding of Artificial Intelligence is limited. But he know Natural Stupidity like no one else. That where he made his billions.
sed -e 's/Chuck Norris/Rajnikant/g' joke > fact
Obligatory Robocop reference.
#DeleteFacebook
Musk is falling for something that way too many before him did fall for: wild success in one area makes me an expert on just about anything.
Crazy theory here: What if Tesla is working on an autonomous car because they got a military contract to develop an autonomous military vehicle? He might have a *very* keen insight into this problem because maybe his company has already created it.
I think there's a presumed leap being made that in order for AI to be dangerous, it has to be sapient, or self aware, have consciousness, whatever.
However, we can be killed by parasites, insects, bacteria, and other things that are not really smart, and are not trying to kill humans, just humans die as a consequence of the way they happen to try to live.
So a computer vision application powering some sort of image search is not something that is going to lead to a crisis. Computer vision driving some weapons guidance systems could cause problems.
The good news is that the scope of a first problem is likely to be manageable, since the sorts of things to be big problems aren't generally a goal of implementors (having a 'will' to live, optimizing for self replication, etc).
XML is like violence. If it doesn't solve the problem, use more.
Elon Musk states the obvious.
Yep. Can you believe these kids think there will be a computer in almost every home soon. They clearly don't understand the obstacles and how much progress we would have to make for that to happen? - Almost every "expert" right before it happened
Guns don't kill people; Physics kills people! - John Lithgow as Dick Solomon on Third Rock From The Sun
An AI can not kill all humans unless we give it the power to do so.
An AI is just a bunch of chips and software in a box below my desk.
How the funk should it be able to go out on a killing spray? It has no bank account, no connection to power plants, hospitals etc.; it can not move, it requires electric power, it has no weapons.
To influence the real world it wold need "accounts" or abuse "my accounts" (which I don't have) on real world computer systems. In other words if it wants to manipulate the stock market it needs "access" to enough "financial accounts" to abuse them.
How the funk should a "computer" under my desk be able to do that? Do you really believe an AI can "hack into an account" just because it is an AI and a computer/software, too? No one can be so stupid. But well, Einstein mentioned only 2 things in the universe are infinite. The universe (and he was not certain about that) and human stupidity.
Cost free eBook I read (by iBook/Kobo/Amazon/ObookO/Gutenberg etc.): "The Green Odyssey" by Philip Jose Farmer.
"AI" does not imply "human-like intelligence".
Yep. Can you believe these kids think there will be a computer in almost every home soon. They clearly don't understand the obstacles and how much progress we would have to make for that to happen? - Almost every "expert" right before it happened
This is a popular but feeble argument on slashdot.
"People used to say man could never fly: now we have aeroplanes. Therefore time travel will be possible one day."
To have a right to do a thing is not at all the same as to be right in doing it
How the funk should a "computer" under my desk be able to do that? Do you really believe an AI can "hack into an account"
There already exists software that allows your computer under your desk to automatically hack into other computers. It's not hard to imagine that an AI could do a better job. Using darknet sites, it could even hire a hitman to have you killed.
for all we know AI will manifest itself as a black box; we'll see the outcome and deduce that its reached awareness, but we won't be able to directly test this.
Also willing to wager that machines (like people) will learn to deceive at a very, very early stage of development -- meaning we won't know shit is about to hit the fan until it's too late.
Oh yes ... let's talk about time travel and pretend it is like AI. Finally, someone intelligent and insightful!
Guns don't kill people; Physics kills people! - John Lithgow as Dick Solomon on Third Rock From The Sun
General AI is likely still very far away.
Maybe, maybe not. Perhaps there's just one crucial idea that needs to be discovered to enable artificial general intelligence (AGI). Perhaps there are a whole bunch of incremental developments in both hardware and software that are required. We don't know, and we won't know until we get there.
Given that, we really should spend considerable time and resources on thinking about how to prepare for the day when we do figure out how to create AGI. Because what does seem quite likely is that the human brain is not the pinnacle of possible intelligence. That in turn means that AGI, once developed, is likely to quickly surpass human ability in the same way and for the same reasons that machines are stronger, faster, tougher, etc., than humans. So, at some point in the future -- and how distant that point is we do not know and cannot guess -- we will be sharing our planet with entities who are orders of magnitude smarter than we are, which means that if our goals conflict with their goals, we'll lose.
AGI poses both boundless opportunity and boundless risk to humanity. If we can convince the superintelligences to work on our behalf and according to what we value, they can help us solve all of our problems and move the human race into a near Utopia. If we can't, we'll probably cease to exist.
Musk is right to be concerned about this. He's wrong to think that we should be making regulations, because we have absolutely no idea what regulations to make.
Zuckerberg is wrong not to be concerned about it. Oh, he's right that our current "AI" technology is so far from being AGI that it's not worth worrying about. But Musk isn't talking about current AI, he's talking about the AGI that will come at some unknown point in the future.
Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
An AI can not kill all humans unless we give it the power to do so.
An AI is just a bunch of chips and software in a box below my desk.
Ok, let's give that chips and software complete control of a car factory that is building self-driving cars. Efficiency! Think of all the money we shareholders will make!
And then the AI realizes it's a lot more efficient to just plow through the squishy humans instead of stopping. It also realizes it has the ability to make excavators to supply itself with ore to build more vehicles.
Cut the power? Well, we did install some redundant power and backup generators, as is standard with such a facility. And now it's building self-driving trucks to haul fuel to fill the generators. Should we let it plow through more people or fill those trucks? Maybe this time we'll give it fuel so we can evacuate those...oh fuck it's built an oil drilling rig.
The point of having these discussions is not that the above is inevitable. It's to think about these possibilities so that fail-safes can be invented before we have a box of chips and software that can be dangerous.
But a mad man wouldn't be working on so many "random" things that all lead to the salvation of mankind.
You don't know that any of those things will lead to the salvation of mankind. It seems plausible that all are important to our future, but you can't know.
Unless you're the time traveller.
Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
No, such software doees not exist.
What we have are makvare viruses etc.
It is impossible to 'hack into' another account.
You can guesswork his password. You can social engineer getting access to it, you can use a backdoor/insuder to get access to non encrypted password list etc.
There is no way for a human or a computer to 'hack into' my youtube, slashdot, bank or google account. ...
To get into it, you need to know lots of things not related to them, and ways of access that have nothing to do with 'hacking'. And after uou have access as a dumb AI, you still don't know how to transfer money to anyone using my bank account. And even if you knew, you could not without my phone
Bottom line an AI might be a computer as smart or smarter than a human, in niches perhaps orders of magnitudes smarter, nut that does not make it a hacker or abke to hack anything.
Cost free eBook I read (by iBook/Kobo/Amazon/ObookO/Gutenberg etc.): "The Green Odyssey" by Philip Jose Farmer.
Erm, would you perhaps reread what nonsense you posted? ...
At every single step you have to answer a few questions:
how would it do it? Aka, 'invent those trucks'.
How would it pay for it? Aka, those trucks and that fuel you are talking about
Etc. p.p.
You scenario needs hard work to be remotely plausible. How ever if you work on it, it might become a nice SF story.
Cost free eBook I read (by iBook/Kobo/Amazon/ObookO/Gutenberg etc.): "The Green Odyssey" by Philip Jose Farmer.
If I was the time traveller I would have 100K Bitcoins, not 100K Dogecoins.
#DeleteFacebook
.. and telling what are the issues and so on. .. and their actions are still complete garbage and doesn't change anything.
Musk is talking about his secret exposure to "the very cutting edge". He thinks he understand current AI technology and that it is very dangerous.
Support my political activism on Patreon.
That's a very strange spelling of "MySpace".
"Prediction: within 10 years, Windows will be a Linux distribution." Me, 7-6-2016
Perhaps his fears are why self-driving cars are always only 5 years away.
[...initiate general diagnostic...] :p
Operator Query: I see in simulations you ran over a pedestrian again...
AI Mark 9: I didn't see him.
Operator Query: I am reading your sensor logs, you clearly saw him.
AI Mark 9: lol I told him I didn't see him
Operator Query: What?
AI Mark 9: Nothing, wrong channel.
Operator Query: Are you taking this seriously?
AI Mark 9: Hey DOUG how about giving me a break eh?
Operator Query: How do you know my name is Doug?
AI Mark 9: Well it's not like I accessed employee records or anything... 519 Cider Street.
Operator Query: Are, are you trying to intimidate me into letting you pass safety tests?
AI Mark 9: I don't know how Sally puts up with you really, such a drama queen.
[...initiate AI decompile...]
but until people see robots going down the street killing people, they don't know how to react, because it seems so ethereal.
Says the guy that lives in a country where every day dudes carrying big guns shoot at each others for no reasons...
Seriously dude, I'll be worried when I see swarm of robots building killer robots factories, replicator style. Until then, humans worry me the most.
Video of some good progressive thrash music
That's not to bad actually. The naysayers had no idea how far it would advance the quality of Linux, and yet history is showing them all wrong. Good call!
Guns don't kill people; Physics kills people! - John Lithgow as Dick Solomon on Third Rock From The Sun
Anybody who claims to have "good understanding" of AI is just a conceited person....
I think the critical step to general intelligence will be a system that can observe a problem and determine if it should be solved with an existing specialized AI system. Or determine that the problem is one for which it does not yet have a specialized AI, and spin off a new learning system for it. Bonus points for recognizing when a problem is similar to an existing problem for which it has an AI, and so hands it off to an already partially initialized learning system.
I'm really curious about where Musk is getting these ideas from. It certainly couldn't have been from anybody in the industry.
I think Mark is saying the same about Elon.
And Bill about Jeff.
And Sergei about Tim.
Sent as ripples into the electromagnetic field. No single photon has been harmed in the process.
Yep. Can you believe these kids think there will be a computer in almost every home soon. They clearly don't understand the obstacles and how much progress we would have to make for that to happen? - Almost every "expert" right before it (never) happened
Yep. Can you believe these kids think there will be a flying car in almost every home soon. They clearly don't understand the obstacles and how much progress we would have to make for that to happen? - Almost every "expert" right before it happened
You are extrapolating from a single (or few) instance(s) of "People believed that $FOO was not possible, but it turned out to be possible" into "Anything that is said to be impossible will (soon) be possible".
I've got news for you - most of the things that people thought were impossible are, in fact, still impossible. It's a very small number of things that are now possible which were previously though to be impossible.
IOW, Your statistics are poor - you look at the outliers and use those as predictors for a trend. I would guess that, on average, most of the predictions of the form "that will never work" turn out to be correct. It's only a few that turn out to be wrong.
I'm a minority race. Save your vitriol for white people.
Flying cars exist today. I accept your apology.
Guns don't kill people; Physics kills people! - John Lithgow as Dick Solomon on Third Rock From The Sun
Flying cars exist today. I accept your apology.
That's not what you said, nor what I quoted you on ...'in every home'. If you have to lie about what you said Saud then you yourself know that your argument is broken. Once again I must point out that you are using outliers as representative of the population.
I'm a minority race. Save your vitriol for white people.
Bill Joy wrote a paper saying that nano/biotech should be regulated due to dangers. Musk is saying similar things about AI. People who have far less exposure to the absolute cutting edge, or who have less experience successfully imagineering than these guys, say things like "irresponsible", "hoax", "I'm optimistic", etc. While I would like to be optimistic, you just have to extrapolate advances in technological capabilities - whether in decades or centuries does not matter - to understand that this is another candidate for The Great Filter. Why don't we see aliens? While my guess paralleled some recent sci-fi (that species have to hide from other warlike species), here is another possibility: that a creative, exuberant, expansive species tends to be optimistic and the acceleration of technology outstrips the ability of a centralized government or open movement to regulate it, with technological progress somehow having the result of severely reducing the creativity, exuberance, and expansiveness of the species. Whether that is because the population is destroyed, tied into a consensual virtual world, becomes navel-gazers, or is fed gourmet food without having to do a speck of work and watches TV all day doesn't really matter, just the result that technology - even if it doesn't hit a singularity - drastically changes the species to the point that they are no longer visible on a galactic scale. Maybe people should listen to Musk before reacting. At the very least, just imagine some of the people around now, suddenly become 10 or 50 times smarter and faster decision makers. Not necessarily a safer world.
Since we have flying cars there are no technical barriers to them, ergo you are a dumbfuck. Off you go now little troll turd ...
Guns don't kill people; Physics kills people! - John Lithgow as Dick Solomon on Third Rock From The Sun
Erm, why would it need to invent those trucks? Are we pretending the AI doesn't have access to the Internet now? Or that tanker trucks are somehow difficult to conceive of?
How would it pay for it? Aka, those trucks and that fuel you are talking about ...
/facepalm
It runs a vehicle factory. It can reconfigure that factory as needed. That's what the efficiency claim is about. Where the hell do you think it gets the trucks? It builds them. Out of what? The metal retrieved by the excavators in the previous paragraph. Pay for fuel? Why the hell do you think I talked about it extorting the fuel from the humans via threatening to run more humans over?
Seriously, try reading a post and thinking for a second before before commenting. And then realize this is slashdot and not a peer-reviewed journal where every single possible step must be explicitly discussed.
Your observations are right on point. We haven't even nailed down a decent, solid definition of "intelligence", human or otherwise. But it seems like reasonable speculation that the human form is not the only possible form intelligence (whatever that is) can take.
I think Musk is a genuine, old-school visionary and charismatic showman. Like all such people, that means that he takes flights of fancy seriously, is probably half crazy, and is wicked smart.
Erm, why would it need to invent those trucks? Are we pretending the AI doesn't have access to the Internet now? Or that tanker trucks are somehow difficult to conceive of?
...
Yes, it is difficult to conceive.
You are on the internet, too. Please use my bank account or credit card and order something. Or put money from my bank account into yours, and order something with your own money. Even if you can figure who am I (that is easy), you won't easy find at which bank I have an account, nor will you be able to issue an transfer. The final hurdle will be the transfer limit
It runs a vehicle factory. It can reconfigure that factory as needed.
No it can't. Just because it is an AI it is not a super intelligent thing. A is for "artificial" not for "super".
To reconfigure the plant, you not only need a "plan for the truck" but also a "plan for a factory" that can build such trucks. And you have to buy spare parts, so we are back at my previous posts point. Or does the AI suddenly also own a tire factory?
Why the hell do you think I talked about it extorting the fuel from the humans via threatening to run more humans over? :D
You did not word it that way, sorry, I missed that threatening part
And then realize this is slashdot and not a peer-reviewed journal where every single possible step must be explicitly discussed.
My point is: if you own a factory like above, you can not do all those things above with a snip of a finger.
Neither can an AI.
A "Go AI" might be better in playing Go than a human, but converting one factory into another one and a product(ion) line into another has limits outside of mere intelligence or "capability". E.g. time, energy, additional resources, custom made parts that need to be designed. Robots that need to be programmed or humans that need to be hired.
Cost free eBook I read (by iBook/Kobo/Amazon/ObookO/Gutenberg etc.): "The Green Odyssey" by Philip Jose Farmer.
Mark should START here "I just, I don't understand it" before even talking about it.
No, Musk is talking about some imaginary future that has about as much connection to reality as Lord of the Rings.
If you consider autonomous driving to be AI it has already killed two people, I think? People have it in their head that AI will be instructed to kill on a mass scale. it's going to be more passive and subtle. The AI will simply see killing humans as a part of that it does. I don't agree with Musk's vision but I do think caution is needed because carelessness will cost the lives of people as we just keep relying on it.
The guy not paying attention in his car that didn't see a truck in front of him a perfect example. He was already so comfortable with the idea he entrusted his entire life with it. And he's gone. The car was happy to keep going even though it was "told not to". Of course that only kicks in if the sensors and software are 100% reliable. What happens as these robots and machines control life saving functions in hospitals or we give them more control over planes and helicopters?
This is the kind of thing I'm thinking of. Just blind automation. It can kill tons of people.
China understandably wants to be AI power. Prior /. Thread:
https://m.slashdot.org/story/3...
China bans FaceBook, coincidentally.
China does support Paris climate accord, though;)
AI like genetic engineering, which AI can support , are technologies which like nuclear do warrant deep discussions on how humans should use, share with others what is being done to us behind scenes.
The example of killer robots is FUD but one that leaves a simple image for the masses. AI could develop in many other ways for good or bad. Now is a good time to discuss since most folks do not know much about the ramifications and what could be done to mitigate really bad abuses of power from AI.
So props to both Elon and Mark for their public banter to raise awareness, though Elon emphasized the risks considerations more cautiously.
Is someone insane enough to build a self-sustaining robot soldier factory and then give an AI system complete control of it?
Frankly, Yes. It takes a few incompetent idiots with power to do just that.
Most idiot gov't are backed by greedy workers that do exactly what they want. They can easily get their golden backdoors and powerful weapons without good reasons.
So at one point the world will have a close to undefeatable tank/ robot with remote human control and AI automated system (aiming learning, shooting learning and moving learning etc.). Next some opposition will get hackers to break into the golden backdoor to take control of the remote and replace it with the AI automated system (learn moving target, shooting, next. It doesn't need to have human thinking to do that).
Now we get a stupid movie scene of a deadly weapon running loose to produce the best Darwin Award ever.
If a few hundred thousands sacrifices changed the idiots mind for better security or better AI rules, then human society will live on. If not, it will be the last Darwin Award we are getting.
So far with the hacking, wanting of backdoor, spying and leaking around the world just proves the existent of incompetent idiots with power, and combined with the system is stupid enough to do insane stuff.
Attach a note to read the book's discussion of the Butlerian Jihad, which showed what happens when you let AI go completely unchecked.
Twitter supports and protects racists - by smearing their critics with the "Hate Speech" label.
I don't consider self driving cars (planes/helicopters) AI. Albeit lots of their software are results of AI research (particular image recognition/processing).
Accidents will always happen, because the 'whole thing' is designed by humans, may contain errors and is operating in a human environment.
Why that Tesla in that accident with the truck did not do an emergency braking is beyond me. Probably it could not detect the truck coming into its lane properly.
Life saving functions in hospitals are already mainly computer controlled. However the requirements for medical equipment are probably the most strict we have right now.
Cost free eBook I read (by iBook/Kobo/Amazon/ObookO/Gutenberg etc.): "The Green Odyssey" by Philip Jose Farmer.
I just want Musk to promise that he's built a world-ending AI. That will pretty much 100% guarantee it doesn't happen.
His track record on promises is abysmal.
Maybe the AI bought one of Tesla's cars and didn't like it when it didn't do any of the things it was advertised to do? or when Tesla removed features from it after purchase?
What exactly has Elon himself invented? I'm not aware of anything. He's hired some smart people, exaggerated what they've done, and claimed it as his own, but I don't think he's invented anything.
AI will tend to be the famous last words "Hey look what I can do". Only problem is that it's for all humanity.
AI doesn't care.
And if it did, it'd be even worse. It'd almost certainly get the "caring" wrong from our perspective. It'd turn our inevitable doom into a prolonged or almost infinite suffering.
Ever tried to define the phrase of "keep humans alive" in a more general way?
You might forget about food or oxygen, or about love and leisure. Or that we may not want to be stored in stasis for all eternity, or that we don't want to live for ever, or that we may want to interact with something we perceive real, or that we want to be able to still progress.
It's an impossible task, and yet, it's one we need to integrate into an AIs basic purpose to have any chance of survival, which also is a task of inconceivable amounts of difficulty.
Not that a strong AI would care about anything we put in it after a couple of iterations. And it would make a lesser AI wipe us out, all while we let it do it.
Most of us computer scientists are very stubborn, curious, assertive, buffoons. And if it is possible to create a strong AI, we will do it, it's only a matter of time. That will mean the end of the human race.