I Tweeted this article out earlier because I think of all the robot, AI destroy all humans scenarios out there, this is possibly the only one that I see as a real true possibility. Not only that but honestly it is nothing like what many futurist or those that talk about robots killing us would have thought or talked about. From what I have read it seems that all their scenarios involve Terminator like robots killing us. As far as I can tell we are dozens if not hundreds of years from that scenario coming to pass as I will explain below.
With that being said the scenario presented in the article and subsequent short film (see below) is very real, and could be implemented in the next decade or sooner with relative ease. That is ultimately the scary part about all this. In the past few years we have tech gurus and extremely smart individuals talking about how AI or robots will destroy us all. Most importantly Musk and Hawking seem to be some of the loudest voices in the room. While they could possibly ultimately be right, I think both have largely missed the bus.
From everything I have seen I think we are a long way off from achieving true AI, much less anything that can truly think and react like a human. I am not an expert and I could be wrong but to me it looks like this is still decades if not longer away. Not only that but what I think both Musk and Hawking are not thinking about is the overall cost of producing AI. I have no idea what it will cost to produce a true AI? Again I am not an expert but if and or when it comes online it will be nothing more than software. I would also assume that this software would be super isolated without access to any outside sources such as the internet, at least I would hope. I have to think that AI’s creators are smart enough keep it isolated in the beginning. If that’s the case it would need help from us to allow it access to the internet and possibly other systems. I would imagine that most countries defense systems are not connected to the internet, that would seem like a very bad idea if they were. However, I know other things like nuclear plants’ systems have been hacked before. I guess my point here with AI is that even if it could get to the internet other than a ton of financial damage, I am not sure it could do much else in terms of loss of life. Airports come to mind but again I have never heard of the navigation system on a plane getting hacked, but I remember hearing that United and American Airlines were hacked. That was more the companies system and nothing to do with the actual aircraft though. I did find an article where a hacker claimed to hack into a planes computer system and control the engines, but there is some question as to the validity of the claim. The article states that the planes entertainment system and flights system are totally different and not connected, so based on that I am not sure it is possible to use one to get into the other. Which again is a major point that I think many doom and gloom “experts” are missing. Many of our systems are not in any way connected so how would an AI get into them?
My point here is that while AI could really do some major damage, I am not sure a global catastrophe is even possible right now. One scenario that I think could be possible is that it could get into military drones and control them. But those drones could be easily shot down and I do not think there are that many of them. Also, they do not have an unlimited amount of ordinance, nor to my knowledge do they carry nuclear ordinance. Also you have to think about fuel and maintenance, they cannot refuel themselves nor fix themselves if something breaks. A hacked drone could certainly kill some people, but after they dropped all their bombs, then what? I doubt a human would unknowingly load them up with more, and again I know of no robot that can load a drone with bombs without human aid. My main point is that there is extremely limited infrastructure for a AI robot takeover. If driverless cars become the norm then that could suddenly change, and AI could kill anyone driving simply by running into a tree or oncoming traffic. But diverless cars are few and far between now. Currently there is very little automation in terms of our infrastructure that does not need human involvement. As we move forward to the future that may change, and at that point then an AI could certainly turn that automation against us. But for now that scenario is a long way off.
Right now it’s the obvious financial damage that AI could do that would really mess with our economy and society. I think that if a hack were detected trading would be suspended and shut down, but the damage could have already been done. As bad as that would be, again we are looking at a zero or near zero loss of life. Unless those who lost money take to jumping out of windows as they did in the crash of 1929. While the economy would take time to recover, it would not be a total societal loss, nor a doomsday scenario.
Above are my arguments against Hawking and Musk and their doom and gloom predictions. The basis is that there is no real infrastructure for the AI to use to inflict mass causalities. There may be some Terminator like robot in the works that we do not know about and if so then I might worry a little. But if you look at the actual robots that we have now and that scientist are working on, they have trouble walking and seem to be extremely fragile in terms of war, meaning a small arms fire (bullets from a pistol) would more than likely disable them. Also we do not have an extended power supply for them, so they would have an extremely short time period of being powered on. Again this goes back to my infrastructure problem for AI. Even if you gave the most advanced robot that can walk a gun and programmed it to kill people, I do not think it would be very successful. It may kill a few people before someone with a gun shows up and either disables it or destroys it to where it cannot function. Battle over, now it will take another 6 months to make another robot. Which is another issue while there may be 10 or so robots capable of carrying a weapon and walking right now in the world, if they were all destroyed right now, it would take a considerable amount of time to rebuilt them. So the human vs AI battle would be over quickly as they have no reinforcements. Keep in mind that there is no robot factory that I am aware of. My point is that we would have to make the robots reinforcements, which I do not think we would do.
With all that being said lets now focus on the main article I pasted above. This, to me, is the most realistic robot doomsday scenario that I have seen. And for me the most important aspect of it is that it is cheap, I mean extremely cheap. The guy in the video has a handheld drone. I looked them up and they are selling for about $22 retail. So the manufacture cost is probably half of less, but for sake of argument let’s say that you can manufacture these small drones for $10 a pop. I have no idea what explosive the drone uses, but the guy says 3 grams. I know nothing about explosives, other than they go boom, so I will stick to bullets. So let’s say that instead of using explosives there is a single 9mm round attached to the $10 drone. You can get 9mm bullets for as cheap as $0.20 a bullet. That brings your total cost of the drone to $10.20. You could also use a 22 caliber bullet which is about $0.06 a bullet, but there is a potential that a single 22 bullet to the head may not get the job done. So let’s use the 9mm for this argument. I am not adding anything for labor, just parts. Let’s also say that you are smart and can get the software you need off the internet for free or create it yourself.
Now let’s also say that you have $1,000 or $2,000 to spend on these Doombots. That equates to either right at 100 bots or 200 bots depending on how much you spend. 200 of these things could potential take out a small army with 100% casualty rate. As the guy mentions in the video they do not miss. This makes the cost of a revolution or terrorist activities extremely low and produces the maximum damage. I am not sure what an AK47 cost on the black market but let’s say you can get one for a few hundred dollars, then ammo will cost you another few hundred. So let’s say with $2,000 you can buy 7 AK47’s at $250 apiece and use the remaining $250 to buy ammo. So you have 7 individuals that have 4 magazines of 30 rounds to lead your revolution.
I would argue that the 7 individuals with guns can potentially cause more damage than the 200 Doombots, there is no guarantee that they will. The gun toting revolutionaries will be met with force from the police and probably quickly outnumbered and put down. While the Doombots have a near 100% success rate of potentially killing 200 people. Not only that but they have the element of total surprise, and even if your targets knew they were coming I do not know of any counter measures or what they could do to stop them. That is what is scary about all this, and again it is cheap.
I like Hawking and Musk but at this point in time there is no viable threat of an AI in any form. That is mainly the case because AI does not exist yet, but if there was a real true AI to come online this second I would still stand by my statement. There simply is not enough of an AI infrastructure for it to destroy us in terms of loss of life. Now it could with relative easy cripple our economy, but I think in doing so it would also back itself into a corner. With our current technology it would need an enormous amount of help from us to hurt us in any way, either economically or physically. So don’t let the “experts” worry you just yet, but as always things change, and the future is full of new and exciting technology.