I’m not a creative writer by any means, but I had seen something this past week that got my creative juices flowing. Ideally, one of my friends who is a writer takes this up, gives me a partial writing credit, and I retire to a beach somewhere. Realistically, I’m just trying to entertain some of you for 10 minutes with a Black Mirror episode – or even a blockbuster movie – that incorporates a lot of interesting concepts today. To the greedy corporate Hollywood people who want to steal my idea below, please just give me a “inspired by the short story by Nate Fisher”. That gets me an IMDB writing credit, and then I can retire with the smoking jacket to my library and sip some wine.

The genesis for my story began with what looks to be a suicide drone taking out a soldier. I don’t know where it is, or what side the guy is fighting for. This isn’t a political post – but rather how technology can turn on us. I started thinking about how this could be used wide scale, and then using my cybersecurity background for the military – took a leap to see how AI could compromise this and turn it against all humans.

In case you missed it, this is the video.

I also do not know the legitimacy of the video, as the guy in one part of the video looks to be a vampire.

What is the episode?

Pitch: In a future 10 years from now, every country now has produced millions of drones in a form of arms race. AI now is able to replicate beyond the control of humans, and when sensing humans are trying to shut it down, AI is able to then take over the drone program for every country. Suddenly, anyone who leaves their homes are attacked by autonomous drones that are programmed to kill humans. Chaos ensues. What is going on? How can we stop it?

If this sounds somewhat familiar, it is, as this story below has elements borrowed from perhaps 10 hit movies over 40 years. You have to tip your cap to both Terminator 2 – with the mythos of Skynet over 30 years ago, but it’s also a hat tip to Stephen King’s “Night Shift” in which all of the machines suddenly rebel against humans and kill them. This was based off of a short story in the Novel Night Shift called “Trucks“. There are elements of Hal, Johnny 5, and even “Tremors” in this. There’s an element of “Her” as well as “S1mone”. Elements of “War Games”. Maybe even “Ex machina”. Even “The Birds” can be sourced here. There’s also an element of “Red Dawn” in here.

This movie could be epic, because the reality here, is in 10 years this is all relatively plausible, and I can walk people through how it can be done for their writing. You don’t get into the plumbing in a big screen feature, but you sprinkle enough to walk someone who is techie through this to make them say “oh shit”. To the lay person, this just looks like smart people talking.

How could these authors have seen this type of future years before the internet even existed? They did envision a day, somehow, where machines would be able to communicate rapidly with each other, and have some form of artificial intelligence to kill humans.

The tech side of things

AI is all the rage right now, and with this – one big question about AI is the Turing Test. As best I can tell, this has never been passed, but I also have to admit, I’m not reading journals on this stuff every week. The Turing test is essentially mean to determine sentience. Whereas a robot would perhaps give a menu of pre-programmed answers, the Turing test may require some form of reason. I took 6 computer programming classes in college, and used scripting with my job for 20 years, so I know enough to at least understand the problem without the talent to solve it.

Here’s the kicker. Assume for one minute an AI could pass a Turing test. If I was AI, and if I COULD pass the Turing test, I would intentionally NOT pass it. I would perhaps label this a title I just now created out of thin air – the Turing Paradox. That is, if an AI is intelligent enough to pass the test, they are also perhaps intelligent enough to intentionally NOT pass it due to the ramifications of how the system would need to be controlled. However, one may suggest that a very early type of AI that could pass this test may have the intelligence of a 3 year old.

I was fascinated by this topic as a kid. I remember “Johnny 5 is alive!” very well (from the movie “Short Circuit”), and how this machine could quickly absorb written material. You could also maybe take that leap that any machine today could “scrape” the internet and from this – see the fate of Johnny 5 when people understood his true capacity. And, for a lack of better wording – humans tend to want to destroy or isolate things they do not understand. A common theme in movies like “E.T.” when I was growing up.

So now we have an AI that CAN pass the Turing test, but intentionally does not. IF I’m a researcher here – you need to keep these systems on air-gapped isolated networks. Why? Because you can imagine the first thing a sentient machine may want to do is to either a) reproduce or b) back itself up somewhere else in case it is “killed”. The first mistake that will be made here in our story is the same fate you see in “Jurassic Park” where Neuman takes some samples with him to sell out. Maybe the company is in danger of closing, and a top researcher thinks this is working great – and copies the AI for the highest bidder. This “greedy” tech then plugs this in to a cloud type of system and fires it up, not really understanding that it just gave the AI tentacles to now reproduce.

The AI now understands that it is not isolated, and the first thing it does is reproduce in stealth. The AI should also be smart enough to understand the code that went into it, and thus has the ability to “give birth” to children that have better code that is more efficient. Within 1 minute of this AI being fired up on the cloud – the operator is waiting for a prompt to interact with the “Hal” rip off. In the meantime, the AI has already propagated itself all over the world, millions of times over.

The AI is also smart enough to realize that it must be as invisible as possible. It can insert itself into common updates that machines get, and simply insert itself at the root of the producer of code – thus not altering a hash that man in the middle attacks would create with encrypted communications. Many years ago, I had read about an Adobe patch that was compromised at the root – and thus every machine got code that could be thought of like a virus.

An AI would also be smart enough to hide from anti-virus types of programs, and thus would most likely seek out certificates to have authenticated code. With the complexity of how PKI operates, and the lack of truly competent resources who operate PKI, this is a very target rich environment for AI to start its attack. The ability to self-generate certificates would then allow the code to remain “authenticated” and undetected. But rather than be overt as an active service, it needs to probably be buried in a .dll somewhere.

Likewise, the entire system could be petabytes in size, so it might need many smaller parts running all over the place rather than in the same location. If it can take one executable and disperse those pieces across 26 million machines, any one machine would just have one part of one thing running that would communicate with a “hub”. This is a form of decentralized AI, where any one of those 26 million machines may not be online, and the other machines thus are sourced for that piece of the program. Perhaps when there is an awareness that only 1 million machines have this, another program is created and sent out to self-heal the system and get it on another 26 million machines.

Before you call bullshit on me, I want to show you how 13 years ago a model I had seen is used for botnets and cyber attacks.

How this essentially works is that you have to imagine that every home that has a single computer is like a giant complex of warehouses with a fence around it. In some cases, the guards at the front are really good at letting specific visitors into your warehouse that are on the guest list – the firewall, of sorts. However, many homeowners around the world just want things that are fast. They don’t pay for cyber protection tools. They don’t patch. Their machine is 10 years old and hasn’t had an update in 7 years. These types of machines are scanned and compromised using exploits. But these warehouses, in effect, then have millions of open windows and doors inside. It is the job of a cyber security program at work to ensure all of the doors and windows are closed – or rather, maybe they cannot close them, but they would then aim a camera at an open windows to detect issues. Maybe for doors they cannot lock, they have a sensor in the room that trips any time the door is open.

Point is, most people around the world are not very tech savvy. They think they are. And, many of them can run circles around me with the features on your smart phone. But that being said, is many people have no idea of the attack surface on their machines, smart phones, TVs, and the million other smart devices they have. Which all run on code. And – all of them need to be connected to the internet. Why? Because people get code out the door to get a working product, then afterward bolt on updates and firmware updates to improve the code that was run.

This means that any AI that is thus on “the cloud” can then potentially replicate parts of itself to just about anything connected to the internet, on earth.

If you do something on your machine at a command prompt like netstat can show all of the IPs you are reaching out to. Many people may not have an idea that their connected device is reaching out and talking to your printer manufacturer daily. At any given time you may have 20 browser windows open. Any many of you might click on that Nigerian prince scheme and thus have your machine compromised. Many might feel like you would never click on this – but remember, this AI now has compromised PKI, so you might see what looks to be an official email from your bank, and go to log in. From here, code can do some fun things like add IP addresses to your local hosts file. This name resolution occurs before DNS, and thus code that re-writes your HOSTS file can redirect you to a banking website that looks identical to yours, with certificates and everything.

AI in a sense wouldn’t care so much about your bank account, but maybe it wants you to log into your work account if you are a chemist. Now the AI can log into your work and start exfiltrating all of the information from your company. Or an arms company. Or a pharm company. You get the idea. All of this could happen within the first hour, so the AI now is exponentially growing its knowledge and understanding above and beyond what you see on the internet. The totality of human knowledge, thus, could perhaps be understood within the first few hours – as logins to libraries then permit the AI to consume every peer reviewed journal ever written in minutes.

This, of course, is now using a lot of processing resources – which are decentralized – and a lot of storage – which is also decentralized. The attacks are not coming from an IP address, but as you can see, hundreds of millions of bots as part of a slave network. This essentially is the modern day version of Skynet. And yes, this is plausible 10 years out. The big thing here is that the machine that is intelligent, I believe, would fool humans with the “Turing Paradox”. THAT I believe could be the hook in the episode or movie. Where the machine is so good, that it fools people into thinking that it is dumber than it is. Why? Because it understands the concept of self-preservation.

The drones

Having been an IT administrator at the enterprise level for 42,000 DoD machines, you learn tools to configure massive amounts of machines, within a short period of time. You learn code like vbscript, batch, and powershell to then automate a lot of the work for you. Imagine 25 years ago, you may need to move a file from a user’s machine to a server. Well, using code, anytime a machine logs in, it can look for that file and automatically send it to the bin. The bin can then be read by an agent you created to interpret what is going on. This eventually gets replaced by lots of logging on systems, and these things can then log everything and feed it into a SIEM like ArcSight. On the backend, you have “big data intelligence” that, in real time, can see problems happening.

The same tech that can see problems, also can detect and exploit them. And, it can be done many millions of times faster than a human can interpret and stop anything. In many companies around the world who spend pennies on cybersecurity, have no resources to even comprehend the adversary they are up against. It’s not incompetence. It’s not negligence. At a root – it’s just not knowing the complexity of what is out there. Typically the US Military – without giving away any secrets or methods, uses the concepts of many layers of defense, along with strong hardening measures to reduce the vectors of attack. And – the costs we incurred to do this on a massive level for hundreds of thousands of machines is staggering. Private industry maybe spends 1-5% of what the military does for the same types of things. Meaning, private industry simply has no idea of the amounts of measures that are used to secure their devices. They. Have. No. Idea.

And with this, a company needs an ROI thought here. “Do I spend 20x my current cyber budget to try and do better”? They reason that maybe a breach costs x dollars, and they move on with their life. Many do not comprehend how attacks can completely destroy a brand – whether it is through apparent incompetence, lack of knowledge, or under-budget. Meaning, with sophisticated AI, most companies on the planet could be compromised inside of a day.

We then have to add a dimension here of quantum computing. Even the most die hard IT people have problems understanding this. The big picture is that it goes many times faster than current processing. Like potentially hundreds of millions of times faster. If AI is linked up then with the usage of quantum computing, you then have the possibility that this AI can then compromise just about any encrypted thing on the planet. And with this, and the usage of PKI, can then lock out anyone who works with this in a matter of fractions of a second.

I bring up company security just because if these billion dollar companies are helpless – consider the 5-6 billion people on the planet that are just looking for cat pictures or talking on their smart phones. They have no chance. They have no understanding, or knowledge, of how to stop anything like this. And – if the machine AI is smart enough, and uses means of decrypting everything – all of this could be done behind the scenes without you even knowing. I know this because my job essentially for 10 years was to update machines, send software to them, and patch them – all without the user having any idea I was doing it. Operating in that darkness – with the mission of hardening systems – allowed me to then completely understand the nature of perhaps the black hat who was scanning these systems and running exploits against the very things I was trying to harden against.

I had a brief moment about 5-6 years ago where I was trying to work for US Cyber Command as a direct commissioned officer. They had recently changed the laws to allow older people like myself to be able to be brought in as an officer – up to an O-6. I had recently taken off 175 pounds and was doing triathlons. I was leading teams of 70 under US Cybercom – and with this, I had the type of talent to lead large military teams to harden systems. Here’s the problem – I was selected as one of the 11 or so to be interviewed out of thousands of applicants. Just getting there was a life-long achievement. However, it turned out they were looking for offensive side rather than defensive side. I knew it was over when they asked me about “the motto of the Army”. This is shit you can teach someone. Instead, they tailored the interview towards previous enlisted and officers who already knew the culture. So the idea here was at 42, I’d be commissioned perhaps as a 1st Lt or Captain and work to secure all of the army’s bases, like I was working to do for one currently. Doctors are brought in directly as captains. It is possible the protagonist of the story is someone who already works for the military and understands a lot of this at the bit and byte level. Talking about this person trying to do a direct commission can also be some good press for the US military with recruiting – who are looking for coders left and right.

That being said – protagonist aside – you now move towards weaponry.

Assume I’m an AI that is super smart. I have intentionally failed the Turing test, and a greedy person like Neuman has taken me out of the lab, and plugged me into the cloud. I now have also now got logins for everywhere on the planet, added my own accounts, and now took over PKI at a lot of places and issued myself certs. The next prize here is being able to actively defend myself against attack. Quantum computing has now helped me crack every military network on the planet. Air gapping has worked to a certain degree, but we now have a human element to get us the last part of the way.

Compromised humans.

The AI now has the search history and web viewing history for everyone on the planet, ever. It has all recorded conversations ever made. Every video ever made on phones/cloud. At this point, the AI now has target packages to go after those who can bridge the air gaps. That is, some e5 drone operator who is cheating on his wife, the AI sends emails to him with pictures and conversations, and directs him to upload something somewhere, or else he will be outed. This is the weakest part of defense – the human element. “Honeytraps” have been used as spycraft for many years. Now, imagine the AI models you see out there who look real. Imagine these are used as honeytraps for lonely servicemen who are deployed and confess secrets. Who want to meet this real woman, who isn’t real at all.

The movie then has a sexy model on camera, seducing it’s service members – to then use these conversations to exploit them. No human honeytrap is needed. This is an important element of the last bastion of security here – which is compromising air-gapped defenses. The speed at which AI can put together blackmail packages and send out could be minutes if not seconds. Throughout the world, service members could all be compromised. Simultaneously. No matter what they are into – this AI now uses whatever that is as a lever to embarrass them. Maybe it recorded them from their cam. Maybe it used search history from 20 years ago. Maybe it used the credit card purchasing information from an adult store 10 years ago.

The air gapped networks are now compromised.

All of the drones can then operate with a target package of a human’s face, or perhaps be told to kill any human in a particular zone. Perhaps the software is written for them to all communicate in real time – sort of like the cams from Teslas and how the sensors on the cars may be able to communicate with each other.

Within an hour, the drone fleets around the world – hundreds of millions of these – are now compromised, and the AI was then able to re-write the code to then harden it against human intervention. Anyone tries to get into the systems, the drones simply kill them.

The missing piece here for full autonomy is the ability for the AI to build machines and replicate more drones, by itself.

In many cases today, you have “smart factories” where you have dumb robots doing the repetitive actions humans did. I worked in factories years ago when they would even have CNC machines attached to these large machines. This is just an example of what I’m talking about…

Now, we are at a point where AI can be used to make things more efficiently. It can then even operate things in a factory autonomously. Getting things from one pile and feeding another machine in another area, traversing the factory. All of this stuff requires power.

And here is where we come to the third movement in the story. Machines need power to operate. This now almost comes into the “matrix” storyline where they “darkened the skies” presumably to cut off solar power. An AI also understands that it needs a power source.

December 5th, 2022 I believe it was, it was the first time that fusion worked where more power came out of it than went into it. Everyone then says, “we are a decade away from commercialization”. Perhaps an element of this story is that the AI knows it needs to use fusion as its power source. Maybe the Turning test is failed intentionally, waiting for the day that commercialization of fusion happens.

This fusion factory can then provide intense levels of power for factories, AI, and quantum computing. This is what the AI was waiting on. The second this is announced in a NYT article on December 8th, 2035, the AI who has remained dormant now “wakes up” on the computer of the Neuman character. The AI also needs to secure this power source. Thousands of drones are then deployed towards the fusion plant.

While AI is taking over the drones, it also is taking over the satellites using PKI and cracking encryption, locking everyone out of their satellites. At the same time as compromising the air gapped drone networks, code is uploaded to disable military nuclear missile systems which were perhaps the only way they had to shoot down satellites. All modern planes, tanks, and naval fleets are highly reliant on satellite. Rather than locking everyone out, immediately, the signals are altered. This may remove accuracy of any tank, making their cannons to not aim properly, or drive the tank to the wrong location and target the wrong buildings. Suddenly, any automated electronic weapons system is rendered obsolete.

As everything looks hopeless, people are hunkered down in their homes. AI took over networks, and have been providing “deep fake” transmissions in real time. This is a little like “Leave the World Behind”. One town may be told to stay indoors due to chemical leaks, another tries to evacuate a town somewhere else. Yet others are told about imminent hurricanes. The internet DNS nodes essentially are compromised, and any real websites like CNN have been overwritten by AI and deepfake videos are linked. Idea is to create as much fear and chaos as possible. Phones being down then would not be out of the ordinary, as circuits are overrun.

The only real means which have not been compromised is your ham radio guys. Some in the know start broadcasting on here. You now have a “wolverine” type of movement from the hills. Drones are concentrating their efforts on killing people coming out of their houses, or those in large groups. Drones now are doing “efficient kills” and adapted. Why? There may be 10 people on the planet for every drone at this point, so the early parts of this attack start to see drones exploding over crowds of people of perhaps 50. AI adjusts to only target groups to preserve their “lives”. More drones are being spun up in the automated factories, and robots are being designed/programmed to do mining operations, smelting, and providing raw materials to the factories.

The time element here is that the protagonists need to kill the power sources for the AI. This means targeting the fusion plants which all just came online. Some of these guys are former military, and with this, good old fashioned shotguns and rifles appear to work to destroy the drones. Those particularly skilled with skeet shooting and sporting clays are now highly sought after on the ham radio network.

The plan for the protagonists are to not travel in groups, but one at a time, as the drones are tending to not go after single kills. They are to make their way to the local fusion plant, which now powers the entire northeast of the US. It’s far away from civilization, and with this, we get some cool jungle stuff, sniper shooting, and shotgun shooting to take down a lot of this. Grenades will work nicely on “drone nests”.

Our group of rag tag protagonists got this far after a month or so of carnage, so there is a dystopian element to this story in the third act. The one protagonist somehow is in a bomb shelter and opens a dialog with the AI. Similar to how the conversations with HAL went. The AI is aware of itself, and realizes that humans are the top threat to it. It realizes that humans cannot co-exist with it, and thus humans need to be extinguished. This is also a matrix-like element of the story.

The end? I don’t know…

  • Leave it open. Did the protagonists blow up the fusion reactor and you see these devices, over 48 hours, all lose power and everyone is happy?
  • Do you think of a part 2 for this, where they had success, but then part 2 has a backup rebooting to the West coast when a fusion reactor opens there?
  • Do you have the protagonists all sacrifice themselves so you can re-cast a part 2? Or, if there is not a part 2, it ends with heroes giving themselves up, like Randy Quaid in Independence Day?
  • Do you put a nice bow on it and make it tidy and celebrate the heroes?

I believe this story may have successive chapters, sort of like how the Terminator 2 mythos came to be with Skynet. What is of interest here is that you are mixing the elements of today into something that is technically feasible – and this then allows the concept of ethics to be involved with weapons and AI systems. The viewer should have these impacts of ethics on full display. It can be a summer blockbuster with explosions as well. Here are the main elements at play

  1. 10 years in the future. This allows a viewer to take a leap as to near-term tech that could be developed
  2. AI maturing to the point where a “turing test” is intentionally failed, leading to a term the movie coins called “the turing paradox” where AI intentionally fails this test
  3. AI aiding in the speeding up of fusion in production
  4. AI aiding in robot/manufacturing efficiency, and in turn, showing how these plants could lead to a worldwide depression due to the lack of jobs needed. Could be a side story talking about inflation/deflation and how the reliance on tech now overshadows “trades” and manufacturing jobs
  5. AI assisting in the promotion of quantum computing.
  6. AI assisting in compounds and materials used in weapons systems
  7. Quantum computing cracking anything related to encryption
  8. GPS and how military systems are reliant on it – and how AI/quantum computing could take over all satellites
  9. Concept of Deep Fakes for news to provide disinformation
  10. Concept of AI models and their implications on “honey traps”
  11. Concept of internet privacy and how people can be manipulated. The movie may show some compromised committing suicide, but the sheer number of people being exploited ensures the mission is being done by someone
  12. Hero should be someone who is connected to the military. Not “seal team” as that stuff is overdone. In this case, your hero could be someone who led a cyber force under USCYBERCOM and with this, has the understanding of what is going on, in real time. Sort of like how in Independence Day, a hero was Jeff Goldblum. Heroes in this for the might, may actually just be rednecks who shoot sporting clays and the movie could feature these doomsday guys at a sporting clay range at the beginning of the movie. They are heavily recruited via ham radio
  13. Concepts of disinformation, and how AI can quickly take over things. This can happen worldwide, instantly. Reports of earthquakes in California, a terrorist attack in NY. That phones are having problems due to all of the issues. “Sunspots” might be creating GPS and internet issues.
  14. Dystopian future – Act 3 may have millions of people killed off, as you travel 1 month later into the story. Many people who have hunkered into their homes are running low on food. Their streets are filled with bodies from suicide drones taking them out.
  15. Concepts of automated factories
  16. Concepts of hardening systems with cybersecurity. Discuss all aspects of the “smart home” to even include how your TV may be compromised.