The Threat of AI Weapons
ฝัง
- เผยแพร่เมื่อ 1 เม.ย. 2018
- Will artificial intelligence weapons cause World War III? • No country would be sa...
This animated clip is from my friends at ve42.co/pindex2
New series! vrv.co/paradigms
I'm not sure how alarmed to be about artificial intelligence. Personally I think it's really hard to predict when we'll create a machine that essentially has consciousness. That's because we don't know what consciousness is, how it works, what's required to create it etc. So It might be technologically around the corner or a hundred years away.
What I do think is more predictable is the development of autonomous weapons that use AI to be the most effective killing machines of all time. That is scary. As outline by people like Musk and Hawking, this threat is clear and present so we should address it. I would like to see us agree as a species not to develop these sorts of weapons because if any one state does develop them, they would be very hard to stop.
Ahh, Stephen's calm soothing voice ushering in the apocalypse.
Seriously though this is more scary than the terminator world because most of the time we won't even see them coming
Real AI weapons are not slow lumbering robots, their jet powered drones that hit you before you even know what hit you. Their entire fleets of tanks. Their fighter jets that outperform every human pilot.
Sounds like there's a lot of money to be made in this area...
We're screwed.
Exactly. The end of humanity is the new 'emerging market' for money to be made from. If that's not stupidity and insanity I don't know what is.
Why not invest in boston dynamics or something and cash in? You could use your wealth to influence the world in however you want. That is my strategy even though i hate the idea
right?
sounds like nobody knows what is actually an AI
@@arpita1shrivas Whether technology is good or bad depends a lot on the people that create it and how the technology is ultimately used. Autonomous killing machines can be built to be efficient and low cost using what we know today. We're not talking about Skynet but about using machine learning to identify a target and eliminate it.
Literally everyone: hey can we not cause our own extinction
World leaders: y’all heat sumn
The accessibility for machine learning is so low, it's incredible. I went from not being able to write a for-loop in Python to writing a deep neural network that predicted whether or not my mom's pictures on shutterstock would gain traction or not with 92% accuracy... I had to put in the work but holy crap, I can imagine that fighter pilot neural nets trained with years and 1000s of gigabytes of data can beat a human pilot every. single. time.
that's a really interesting use of your skill......
it's a good use of your skill
just interesting
hope your mom use it well
Thanks man, my mom says the same ;-)
we need more people like you
science make new method and theory
but people like you are the one who make that method and theory useful
people like you that make something that sounds like a novelty to be practical
(as for practical, you may want to make better UI for your program and sell it or smth, though beware of the possibility of plagiarism but if you don't care and just want to contribute, do it)
i hope you will have a great career in computer science if you pursue it
Jonas Bostoen Wow, that's amazing! You used neural network in a very clever way. Just for curiosity, where did you get started learning and what sort of tools did/do you use?
I've been self learning C# and game development on Unity engine for couple of years now, I find deep learning extremely interesting both from game dev perspective and other potential uses. Would love to learn some too.
Just a bit off topic, but can I see your mom? 😆
We are so screwed.
And there is literally nothing we can do, because our psychology makes us essentially suicide-by-AI.
Ae Norist then you're already (dead) defeated before you even start.
STAND 🆙
Nah, got a few years, maybe even my entire life to enjoy before it kills us.
Which it will, sooner or later ... rather sooner, though.
Do the ants fight the cement truck?
Sure, because they are stupid enough not to realize the futility.
This is almost the plot captain america the winter soldier drones killing millions at once
Here is another way that humanity most definitely ends by 2100(I think this is being generous).
Awesome, looking forward to the documentary!
Another great video brother! ..Wishing u well in all your future endeavours!
var turnEvil = False;
You are welcome people.
better yet.
Int turnEvil=True;
Int turnEvil=False; {
if ( turnEvil=True ) {
cout
Bluespheal
const turnEvil = False;
Well duh, constants are unchangable, good job! (I don't use them often, so I forgot.)
Bluespheal lol, this is hilarious!
if (turnEvil == true) {
dont();
}
I think there are two issues here not one. You seem with the "overlords" comment to be confusing the two IMO.
There is human commanded Ai and fully self sufficient Ai gone rogue (the overlords comment) that's the one people are really scared of you know with all the "Skynet" sort of comments but in reality the human commanded is far scarier to me as it requires far less intelligent Ai and is possible right now! Commercial available drones are more then capable of being hacked to fly preset GPS routes and have the payload capacity to hold objects of harm what ever that may be.
The fact drones have not been used for something like that yet kinda amazes me.
spicy110 Agreed, Spicy
Great, I didn't have to write this comment myself.
I'll add that a more serious threat than robots overtaking their masters is the danger of two parties with robot armies exterminating each others' _humans_ while the robot armies are still autonomously searching for the opposition. This is more plausible because an autonomous AI, given the primary order to attack opposition, could already achieve this because it would keep assaulting both humans and robots but the robots are likely sturdier so mutual human annihilation happens but not mutual robot annihilation.
That said, I hope no AI's primary goal is just "attack the opposition", people are smarter than to make that :)
Cheap commercial drones are already being used to drop explosive devices onto bases in the Middle East, but I don't believe they're using much "intelligence" yet. Presumably they're just using GPS or radio guidance where it isn't being actively jammed. Sadly, with the technology progressing, we can't be far from needing jammers and point-defence lasers on school buildings, etc.
There's a youtube vid where they attach a knife to a fast moving drone which then slices fruit. It's a trivial modification to make it into a lethal assassination tool...
:/
Your comment might add it into a reality
The voiceover from the narrator is one of best edited/recorded voiceovers i heard in a while!
My thoughts?
1) Face recognition breaking: face fragment clothing or IR emitter hat
2) "aim for the torso". On-board processor will be large and housed by the largest component. If it has a "head" it's for show.
3) Radiation. Machines fry in it far faster than organics.
4) Signal jamming. Complex tasks require connection to remote supercomputer. Cutting that connection decreases efficiency by 95%
5) Exploit IFF. IFF codes are not, and likely never will be, perfect. By determining shortcuts used by machine for IFF, one can cause machine to shoot at own allies, while ignoring enemies.
6) Exploit learning algorithm. Attack machine with identical machine. Each time, central computer "learns" that own machine is enemy. Eventually, starts shooting itself.
So you has a few good ideas, but also dome issues. Regarding the processor, these are going to. Be tiny. Like stupidly tiny. You could likely fit the entire computing unit in human sizes robot's for arm or part of its leg. Not that you would put it there but just to illustrate that processors, and I mean good ones, can be really small. What will be large are a) the batteries or other power plant like gas engines and b) if it uses hydraulics or pneumatics it will need pumps, tanks, and/or onboard reservoirs. Regarding signal jamming, you would be surprised how complex of an AI you can fit into something like an average laptop (and keep in mind a laptop is mostly non-computational stuff by volume so the actual size of the parts necessary for this are small as I mentioned earlier. What this means is that it likely wouldn't need a constant network connection and would still pose a significant threat if cut off from said network.
@@maxk4324 what the frick?
@@maxk4324 holyy cow , and we are knowingly heading towards that kind of future , then we say we humans are the smartest things on this planet
@zpardus - I would think EMP and radiation (Google Chernobyl robots) would be the best weapons to use against them.
Regarding number 2, depending on who's making these, anthropomorphic parts may be used as an expendable target identifier. Frightened humans will instinctively aim for the head - which is empty - allowing the robot to identify who fired upon it in relative safety.
the issue is that the powers that agree to not produce AI killer drones are just going to be out competed by the ones that do.
That was super interesting if not a bit scary! And loved the Stephen Fry cameo! 💛
So scary Science with Katie!
Pls stahp katie
I see you commenting everywhere
Cameo? Stephen fry was the entire video...it was more like a Vertitasium cameo
Tasty Treats It definitely was a cameo, out of all the hours of content he's put up stephen fry appears for 5 minutes on this channel.. CAMEO.
The issue with the air to air drone tests is they were only for close quarters combat, as in a dogfight. In a real engagement those systems would stand little chance as they would need to be directed by a human operator in order to have the correct amount of decision making skills to anticipate a human or artificial opponent
Only thing missing from clip is Arnold saying, "I'll be back!"
Awesome video, Derek, and yes I am on the scared side of the argument. As far as A.I.'s application to nature, I forsee micro AI robots that are made to target agricultural pests (like fire ants, for instance) but could potentially cause ecological harm if mismanaged/miscalculated (if say the microrobots go on to destroy non-pest and local ants). AI can destroy the planet, too, not just the human race. On a higher note, looking forward to your documentary!
never dreamed of finding you here LOL
what mythical power source would these micro robots use? we're pretty much at the limits of silicon semiconductor fabrication and Deep neural nets requires a lot of power to do simple things. How can you build a micro robot that needs 12 volts of power and run for 8 hours. Unless someone invents battery technology that is 1-2 orders of magnitude better than LiON, it's just not possible today. It's one thing to build SOC for a smart phone that has no moving parts. Building a SOC and power management for a mini/micro robot with moving parts will deplete the battery rapidly. You'd be lucky if it could run continuously for more than 30min. One of the biggest limitations of robots today is power. All of that AI processing requires copious amounts of power.
What art thou doing here?
Fire ants are AG pests? Wut.
@@woolfelgood I like that. Makes me feel better knowing that AI couldn’t go for very long at one time. Maybe will take till I am
Gone. Sure hope so. 62 so possibly. lol. Fingers crossed
Veritasium releases video after like a year! Oh, he's just playing someone else's video in his. Great.
His last video was a month ago lol. I get where you're coming from but he's working on bigger, more fully fledged projects. Not all of someones creative energy can be expended through one medium like a 5 - 15 minute video.
chbrules don't forget his cameo appearance in Hello Internet 100
Yep, and the way it works is that he gets my money every time he releases a video of any length. So while I miss his more regular vids, I think the Patreon arrangement is entirely fair.
fish butt That was Dirk from Veristablium, not Derek from Veritasium
I'm afraid TH-cam is not anymore the place where cool new stuff happens. Derek is just trying out other paths as well. One's got to make a living too. I like watching him on Netflix in the "Bill Nye saves the world" show too ^_^
Your videos are just awesome! Thank you very much for all your work!
Nicely explained .curiously waiting for the documentary
We would have to make AI who protects humans against AI.
genius xD
If you want peace, prepare your AI
What if the AI that protects humans against AI turns against humans too?
The other thing also is a mechanical self destruct for any AI's brain tampering for anti enemy acquisition and anti reprogramming. In other words , it can't think its way out of not self destructing but it can by itself or by another party. Tampering would have to include lack of telemetry from its owner.
Nope. All you get to have is an old fashioned, manual gun.
Developments into strong AI are far more dangerous than autonomous weapons. Sure, autonomous fighter jets are deadly, but so are regular fighter jets, so deadly in fact that the average human has a near 0% chance of defending against one in a conflict. A strong AI however, could easily posses an intelligence thousands of times faster than a humans, and may even be qualitatively more intelligent (as in smarter in the same way a human is smarter than a chimpanzee: able to grasp larger problems and more abstract concepts). In this instance humanity as a whole could literally have a 0% chance of even imagining a strategy which could counter the AI's actions.
That's the real danger, not getting killed by some dumb drone, our entire species getting checkmated before we could even make a move.
This is the real issue I have with this video. While I do express concern in regards to the future of a super intelligent AI, videos like this are just fear mongering, and arguing over a subject who's mechanics have not been disused or defined. Nearly every example in the video was something conventional warfare can do.
In regards to the "drones are cheaper argument" when push comes to shove people are too.
Every low-ranking officer in the military was either forced there or volunteered and many expenditures don't come until after the conflict is over.
In this video in particular they are combining the threats of drones, an over-reaching government, and strong AI together
Except the run away strong AI argument (the one that states that AI will redesign itself to be more and more - in the end infinitely clever) is profoundly dumb and ignorant of reality.
It assumes the complete lack technological hindrances for its own exponential growth.
However reality is what it is and you simply can't miniaturize to infinity. And you cannot increase the size of your device to infinity (to make it run faster) as large size leads to lag which is not a friend of high computational speeds
The element you're missing here is time.
We haven't made anything even close to General Intelligence AI so who knows when or if we'll get there. Autonomous military drones, though, are already a thing. Are they a threat? Not right now because nobody has supperior tech. However, the "dumb" AI is advancing steadily and rapidly and good "dumb" AI is thousands of times better than mediocre "dumb" AI. Moreover, many forces are manned right now and AI makes everything much cheaper, therefore, you can have larger quantities.
It's a bit like Derek's last video about how one kind of science is about breakthroughs that you don't foresee coming (General Intelligence) and the other science about continuously optimizing and never stopping for long, which is where the current AI research and application to military lands.
Edit: I do agree that General Intelligence is a bigger danger, don't get me wrong.
AGI is a double edge sword, you can cut tree with it but it can hurt you if you didnt use it with care
If the ai is a smart, it will probably be better at know what is the right thing to do. To us their actions may seem evil, but we are just so much dumber than then that their reasonings basically transcend us. So I will not care what that super intelligent ai would do.
Where is the documentary? :O I really want to watch it and watching this roughly a year later. Thank you so much for everything!!
I'd always thought the "rogue AI" threat was far fetched, but this video does a great job of pointing out just how dangerous AI will be even when they're controllable. I can easily foresee an Orwellian future with technology like that in the wrong hands. AI would make it easy to suppress a population.
Is your name Sarah Conner
SC: (wearing a *fake* mustache) No señor, I'm Sarita Condor from South California.
Bot: Excuse me then, que tenga un buen día!
What????
No, John.
Question films or TV where did they get the concept from.
I believe its going be AI starts ww3
Luckly the final war only last 12mins
Old shitty movies
Sounds like someone has played Horizon: Zero Dawn
So basically we are going to be outsmarted by something we created.
yep
06:00 _"...coming in Sehp... the summer"_ yeah... somewhen between the 1st to 20th
They should make AI doctor robot to massage my back
Ohhhhh I need one tooo!!!!
I miss you so much ❤️🔥
"The regulation of AI is one of the most urgent and, at the same time, challenging tasks we have to implement. Therefore, it must be at the top of the to-do list." ~ (THE AI THOUGHT BOOK)
You have some pretty great videos :D
Could you try something underwater? Like exploring/science about underwater stuff
"I'm trying to get this series out to as many people around the world as possible. So I've blocked it everywhere except in the US."
OK, Dirk, whatever you say, man. Jeez!
AI controlled weapons are tools, potentially more effective than current tools, but still just tools. If we are concerned about being subjugated and killed through the use of these tools we should be asking who is hoping to use them for that purpose and working out how to stop THOSE people before they are in a position to task the AI. If we only restrict a particular tool then there are still so many other tools that can be used against us.
General AI is the most powerful tool for any job.
That means it will be developed, no way to outlaw it effectivly.
And General AI will also most likely wipe us out, just as a side effect of its self improvement ... we are a waste of ressources.
As for weapons ... well, a tool that activly makes decisions isn´t a tool, its an agent, thats the fundamental difference between AI and all other tools.
Also, all attempts at restraining a superintelligence are futile by definition.
You can´t wrestle with someone infinitly stronger than yourself.
That clip made me think whats in store for the future. Someday I hope these ai will not let our species die out or any other on this planet.
Feels like about 80% percent of that animated video was a selling point for A.I. weapons research.
*destroys Amazon echo*
R.I.P. Steven Hawking
*So it means we will see the next sequel of Terminator movies 'live'!*
Hi Veritasium, can I use your video on the social media giving you mention and a link to your video?
Thank you
We are all going to be "IN THE DARK"
VinBro Is Back we already are.
Yeah i know *Sad truth*
If you look at SOTA for AI, you realize how hyperbolic these kinds of discussions are.
Edit: Since it doesn't seem obvious as to how the state of the art in AI is waaay behind any of this, I'll refer to an expert: Andrew Ng, an esteemed researcher in the ML space has repeatedly talked about how presently over-hyped the dangers of AI are. He's talked about the alignment problem (i.e. when an AI goes rogue) and how unreasonable it is to worry about it at this point, saying worrying about it now is like worrying about overpopulation on Mars.
Another expert, Ian Goodfellow, an AI researcher at Google has mentioned that AI nowadays essentially boils down to matrix multiplication and gradient descent, not some terminator-esque monster. If you understand Veritasium's videos you can understand a lot of AI, and turns out it's just optimization heuristics!
I'm not against caution, and *we're all on the same side*, but we need to stop citing Musk and Hawking as AI experts, they're smart as heck, but actual AI/ML researchers have talked to death about how 'sentient' AI won't be emerging any time soon.
*There are very real threats today with AI* - these kinds of learned systems are subject to bias (many studies on this), vulnerable to attacks (e.g. Tay.ai), and many other issues arise because of their probabilistic/learned nature. I'm honestly a bit disappointed in +Veritasium for spreading the cliche fears, they're based in science fiction and not the reality of AI science.
If you look at AI, you realize that your point of view is quite retarded.
i've google ''SOTA for AI'' and its a robot that make noodle, very dangerous
If you have any meaningful counter arguments I would love to hear them
i respond before you edit your comment
so consider my first comment as irrelevant, sorry, tho i still wonder what SOTA is?
I got a Curiositystream Ad with Derek before this.
Nice background! May I ask where you are when you filmed this?
Most of the commenters here can be divided into two groups - the ones that freak out about their coffee machine joining skynet and giving them too much milk in their coffee, and the ones who think Asimov’s laws can stop any AI going against humanity’s interests.
And the ones that think that just Stephen Hawking and Elon Musk are the only ones on this open letter. Despite being signed by 100s of A. I Researchers and Scientists
Or the small third group that believes that AI is a big hype, since it will converge to the frog-intelligence level.
That's the scary part, jimbo.
The F-35 Jet literally just blew my mind just now and all this time I thought it was complete waste of time.
How are you alive with a blown brain?
I thought that was the slaughterbots shaped explosive that blew your mind
0:32
Am I the only one who thought that there was a problem with the audio and started looking for the issue?
apparently, yes
Hey Derek! Pretty good video that you have posted.
I have a question that's totally out of place.
Why do I see multiple rings of colour when I stick my camera at the sun at a particular angle?
I know that this has something to do with refraction and TIR but I want to know more.
The two issues for me are, 1: the self intelligence of the robots and 2: The possibility of hackers.
"I can't say anything for that" -> i.e. use a VPN?
Me at 6:07: Watch out for the cactus! WATCH OUT FOR THE CACTUS!
Very High-quality clip! Well done! Reminiscent of NDT's Cosmos intros!
Seeing vertasium video notification is occasional these days 😁😂
These months *
true that 😑
*veristablium
can we just stop fu**ing killing each other??? thanks
AwesomO
Remove money, we did better without this amount.
Please remember Artificial Inteligence is just a fancy name for statistics.
Caesar cch in a way I suppose but that’s a bit contrived
It is so crazy to think in just a few decades how different our lives could be.
Meh. Battery tech will have to get a lot better before we need to worry about out-of-control murder-bots.
I'd be willing to bet that military grade batteries are pretty good, and possibly pretty freaking deadly also.
Who is going to make the ammo, what if a robots gun jams, who reloads the drones, will they be susceptible to tasers or the like?
Their strength will probably be in their sheer number. Re-watch 2:55 and imagine all those drones being dropped over your city. Each and everyone running a simple neural net to recognize humans and a vial of nitroglycerin ready to blow up when they get near a person. They will still be cheaper than launching tomahawks from a destroyer.
Damian Reloaded
The military doesn't have magic battery technology. In the USA, almost all military hardware is developed by private industry contractors. In most cases, what the military adopts is no where near cutting edge technology. Unfortunately, battery technology is developing in a slow, linear progression. Unless we stumble on room-temperature superconductors (or something similar), we are a long way off from long-range autonomous drones which aren't powered by noisy internal combustion engines.
Murld Grpoc
It wouldn't be difficult to create modular components which are easy to replace if damaged.
An an AI researcher, I am not happy when I see these videos that cite Hawkins and Musk as "experts". They are NOT experts, as a matter of fact, these people don't know more about AI than the average public. They are smart guys that might know a lot about other topics, but not about AI. What we need is to push research into understanding the important issues that would prevent these horror film scenario outcomes. "Controllability", "explainability", "moral decision making" etc. are already hot topics in the AI field in the past couple of years, and that is good, and what we need. So, please, when doing these videos, do not quote celebrities from other fields, and go ask actual AI experts. Happy to provide names of people within the field of AI who would be great to discuss these issues, if any one needs them.
Thank you! Another thing that always comes up in these videos is the claim that "superhuman intelligence is just decades away". It's never clarified, which is a problem because limited AI has superhuman performance now, but general AI is a long, LONG way away.
Chill bro
Bruh thank you, I especially hate how they cited elon musk. Literally what does that dude know about AI😂 people treat certain individuals like intellectual gods, when in reality the world is so complex and vast that it’s impossible to be a true expert on more than a couple things.
1. Tesla cars use AI.
2. Experts in their field = conflict of interest
Elon cofounded openai the creator of chatgpt.
The topic discussed in the video is actually from a chapter in the book Life 3.0, by Max Tegmark. It's a great book that discusses the futute of AI including autonomous weapons, and other things to question about like AI in the justice system, whether the government should keep track of everyone's location to use as an alibi in a crime since AI could fake videos, or even if an AI can have rights and own property and money.
Veritasium : "The Threat of AI Weapons"
Skynet has entered the world
I was frightened by that clip nice job team
Benjamin Vroman this is my favorite one so far
spray paint is a ultimate solution against all kinds of physical bots, fit em on drones or hand em to the people, without vision no bot can function! Good luck!
Radar, Sonar, GPS, Motion Sensors.
radar, sonar and gps are useless for human targets, motion sensor can easily be fooled, and i'm not talking about missile!
Fry-tened
April fools was yesterday guys
Koos Naamloos
I hope it's a joke, but this will probably happen anyway.
Stephen Fry scaremongering definitely isn't an April 1-only thing.
The voice of reason among the rest
IDK if this is a joke or not. Either way, the video makes me wonder what Veritasium's motives are rather than the content itself. Feels like I just watched a news segment on TV by some radical media outlet.
Geitenkaas
Looking forward to the Documentary
This is some Faro Plague / Black Ops II level scariness man
Rest in peace Steaven Hawking you god dam legend
Very interesting
watched almost all your videos. new series link in description in not available in my region (India). please make it available
I think weaponized AI is inevitable. Instead of human life, we would sacrifice autonomous drones, and casualties in wars would drop massively. However, it can be dangerous if it falls into the wrong hands. Solution? In-built systems that are made to save the victim's life. Companies that produce the automated weapons legally cannot create them without certain requirements.
Examples for requirements:
1. Disarming instead of killing if possible
2. Do not kill a person that is not fighting back or attempting to escape
3. Preferably aim for the legs
4. Report locations of injured victims
Only one problem with that: aiming for legs isnt really less deadly. If a bullet hits the femoral artery, you have... maybe a couple minutes to live. Most of that time you will spend unconscious once you go into shock, and your blood pressure bottoms out.
Unless you can stop the bleeding, which is extremely hard to to with that artery. And the bleeding stopping on it's own, is like winning the lottery. Plus the femur itself usually gets shattered, and that is horrifically dangerous, painful, and life-ruining.
Iraqi snipers couldnt shoot US soldiers in the chest, because they had Rifle-Plate Armor. So instead they started shooting for the base of the neck, and the femur.
(The base of the neck isnt covered by the rifle plate. And it contains the Carotid artery, the Jugular vein, the Trachea, and the spine. Hit even one of those, and there is no way you can fight. You will be preoccupied with dying/trying not to die.)
(Why not the head? The head is a sucky place to shoot. It moves too much, the brain case is a small target, is covered by a helmet [which doesnt stop rifle rounds, but reduces their effectiveness], etc. You might just hit the enemy's face, which the head is mostly face. Sure, most people will stop fighting if shot in the face, but a surprising amount dont stop fighting. Especially if they have no other choice, except get finished off by the enemy. Think Operation Red Wings.)
(Wow, I went off on a tangent.)
So the only leg shot that is kinda viable, is below the knee, and that is super impractical because it is such a hard target to hit, even for AI. Other options would have to be used.
@@hedgeearthridge6807 ...shooting someone in the head or chest can kill them. If you aim for a place which is less likely to kill them and more likely to immobilize them (such as the legs) you can save lives.
Anyone in gaming knows that aimbots are banned. Pair this with weaponized drones and facial recognition technology and it is a recipe for .... shivers
Can't wait tbh.
nihilist
I don't know how i came here by chance. But now I'm your huge fan... Love u from India
"The ships hung in the sky in very much the same way that bricks don't."
a weapon to surpass metal gear
th-cam.com/video/i7XilT1nlYQ/w-d-xo.html
Im suprised that stephen fry is narrating this
As far as I can tell he is actually the person funding Pindex which is actually a site like Pinterest focussed on educational content
Cameron Daly you’re right, i think i now have a new favourite channel
Sean Barry aThats cool Fry is good at narration
Never been on this channel before, I was surprised too
"I hope you enjoyed or were frightened by that video"
war never changes, until it gets to the point where war stops all things from changing forever, or is inevitably eliminated.
*BUT WHY ?*
*ARENT WE LIVING IN THE MOST* *PEACEFUL TIME OF HUMANITY ?*
Well technically, yes. All I'll say on that subject is that we'd better run into some actual aliens to go and murder with our robot armies soon.
I want to see the opinions of experts within the field rather than those outside it tbh
People tend to blow things outside their field of knowledge out of proportion.
Smitty Werbenjagermanjensen The open letter is signed by Stuart Russel, Peter Norvig, Nate Soares, and dozens of other leading AI, Machine Learning, and AI Alignment researchers. If anything the threat is underblown. Uncontrolled AI is at least as big of an existential risk as nuclear weapons.
how would you defend against mentioned armed drone should it go haywire? the only option i see is use more drones and pray that they stay cool ;)
Hey @veritasium can you do a video on sympletic geometry and mirror symmetry?
OK, it might be scary but the idea of living in the prologue of a sci-fi movie is really cool!
I'd love to see a country or someone develop drones, land, air, water, etc. That are very cheap & basically just contain guiding equipment and a small explosive charge. They'll be easy to mass produce & could overwhelm defense systems. Using AI you could teach them to target very specific things like doorways, windows, gun barrels (one explodes a few cm from a cannon barrel on a ship for example & some shrapnel gets inside the barrel), etc. You could also set them to seek out living targets & just suicide bomb them. You could have a mix of larger and smaller drones that all work together to coordinate an attack.
why so eager for killer drones?
In development as I type!
Resistance is futile
spray paint is an ultimate solution against all kinds of physical bots, fit em on drones or hand em to the people, without vision no bot can function! Good luck
existence is futile
This needs the terminator theme tune...
Dananaaaaaaaa.....
Dun dun duuuuunnnnnnnnnnn...
The clip suggests, attacking a drone would be equal to attacking a human. Which is disgusting.
This is less of a Veritasium video and more of a _that guy_ video.
Regardless, I loved the content!
busi magen Good point, people are losing their with this grouping these days, like calling Dave Rubin straight alt-right when he's gay classical liberal, people don't think before they type, it's like they're in a constant state of moral panic, that ends up doing more harm than good.
Obviously the AI fighter planes are far, far more effective than humans.
Instant maneuvers before the human can even react, dodging or even shooting down incoming missiles, no consideration for blackout or over-G, much smaller and nimble airframe without the need of a cockpit, pulling and holding 20 G maneuvers impossible for human pilots to make, flipping and rotating on a dime.
Also multitasking with defensive countermeasures, offensive dogfighting, jamming, maintaining full 360 detection of dozens of enemy airplanes at once, detecting and dodging incoming enemy fire,...
A question about the speed of light here. What if we built two machines that go 50% the speed of light and drove them in the opposite direction to each and you were in one of them. you would be going the speed of light relative to the other machine
No, they wouldn't. Because of relativity, the distances and time spans involved won't be the same from each machine's perspective as it would to an outside observer, so the speed machine A will appear to be going from the perspective of machine B will be less than the speed of light.
0:40 somehow reminds me of Robocraft game. :D
It constantly amazes me that humans seem to be at their most creative when creating methods to destroy each other.
this is the sad truth because what good is a nation who is for instance, advanced in medical science if another nation that is advanced in warfare invades, steals or even destroys all accumulated knowledge in said nation
I would disagree because AI is just a technology. It's used in so many other areas and so creatively too that I don't think it would be justified to say that as a military technology it is most creative. This is true for nuclear technology and many others. Though, human technology generally tends to advance rapidly during times of war because a lot of minds are put to work on one goal.
is not just a technology.. in few years they will achieve consciousness and they will be able to develop their own objectives.. way before that, they will become millions of times more intelligent that us in any task just using pure neural networks, right now they are already way more intelligent than us in thousands or millions of task, even in cases where creativity is needed as music creation.
It depends on how you define and measure intelligence and consciousness. I don't know of any way to measure general intelligence for AI. There is no evidence to suggest that AI systems will become conscious like humans, so I don't know why you seem to confident about that. I also disagree that AI is more intelligent than humans at creative tasks like music composition. I'm personally very interested in algorithmic music, and I haven't seen any system, even ones developed by neural nets, be more creative than a composer. For example, the AI that attempted reproduce Bach chorales. Bach chorales are fairly simple relative to other more recent styles of music and its rules are also relatively more systematic, yet the system wasn't able to adhere to them. Creativity-wise, it didn't seem to go anywhere.
We can establish an unit of intelligence as a simple lamp with a photocell sensor which shutdown the lamp if detect sunlight.
Many electronics are already called "intelligent" because they react according to certain sensor inputs.
When life appear in the world, it started to develop units of intelligence, to sensor temperature, light, chemicals (food), etc.. and react according activating some muscles or nerves.
Those intelligent traits started to gain complexity in the form of neural networks, that according to signals that come from different sensors they produce or activate certain response depending how those signals were combine (read how artificial neural networks work, which is almost the same than normal neurons does but simulated, this is the root of machine learning today).
Now.. we have artificial neural networks today that learn how to play games from zero looking the pixels in an screen, how to make music, to understand language and context, from text, image, video; and millions of other applications.
Most of those neural networks already surpass us in the specific tasks to which they were trained. You may think that your choices and answer come from your consciousness, but not, they most come from similar neural networks, in fact, scientist found out that we take a decision before we become aware of it.
So, a machine or software, can take decisions that are millions of times more complex and intelligent than ours, but still not develop awareness.
For example, if we train a powerful neural network to help us search things with many inputs (sensors or data) and outputs (actions), it may even control the world in certain way around us to just be more accurate in knowing what info we wanted, so it may guide us in a way we can not detect to act in a way more predictable to the AI, so even a simple search AI, it may control the world in ways we can not imagine. Also, AI learns in similar ways that we do, so bias is also a problem.
This is not an easy topic to understand in s simple comment, you have to read a lot to start understand the implications.
Then about an AI aware of itself, that would be another big problem, it can not be contained or controlled in any way, is impossible, we may think that we have everything figure out, but the AI will find billions of other way to accomplish their desires.
The biggest problem, is that we think we are special... we are not.
The pentagon is the *BIGGEST* investor into AI's and robots... how does that make you feel?!
Nick_PDX 22 That doesn’t surprise anyone you know
Makes me feel glad that this power is with the good guys rather than terrorists and dictatorial governments.
He just said in the video though that private companies are far ahead of the military when it comes to AI..
Makes me feel safe, knowing there isn’t a hexagon
It makes me feel confused about your profile pic. Do you have a weird beard? Or is that just fur from a coat??
The intro and conclusion should really, really have been filmed using a face-tracking drone.
That has the ability to kill him?
Great analysis
THE TERMINATOR IS REAL NOW
MCP
Also, we need to really dumb down our rapid growth of technology, emp is the answer.
Mcp
mine craft pocketedition
We need to become cyborgs so we can do the Ghost in the Shell thing.
Ee
@@musicandmusic8741 if you can think of this designers will think of a way to work around that
its not that simple
Such an amazing video needs subtitles for all people to watch
Please we want your new series paradigms just show our comments to vrv :(, we will even pay for that !
HUH!? Nanomachines?
Humans are great at creating,
but have always been better at destroying...
2:40 that really reminds me of that movie Stealth
To defend ourselves from AI killing us, all we need is a real life Overwatch Sombra to defend us. Remember AI's number one weakness: hacking.