I know I bang on about large language models and AI a lot, but it does seem inescapable at the moment. While I gladly use it in my day job as a software engineer, I am wary of using it for artistic purposes. Aside from issues with its quality, the ethics of it bother me. It is taking the livelihoods of people whose aggregated work it has effectively stolen while being trained. Another more pragmatic part of me however wonders whether I couldn't find some use for it despite these objections. Could I use it to create promotional videos for a book, perhaps? I don't have the budget to actually pay a professional, so nobody's missing out on any work.
I am aware that I made the same arguments when file-sharing first took off. It wasn't like I could have afforded all those Metallica albums. It's a morally slippery slope and one which at the time I had merrily slalomed my way down with considerable enthusiasm. My ethics can be sent packing by the flimsiest of temptations, it seems.
So, with that objection cheerfully waived, it didn't take long to get from using AI to create cheesy videos to using it to steal my neighbours' post.
I should give some background information. I live in a block of flats, and there are frequent deliveries of parcels to the building from Amazon et al. There is no concierge, so parcels are often left in the lobby. They are sometimes left outside our flats' doors, which may seem more secure. However parcels delivered to either location are frequently stolen. We don't know who is stealing them. It could be someone tailgating the delivery people. It could be people getting in via the car park, the door to which is often broken. It may not even be one person. This is where my opportunity lay. The best place to hide a needle isn't in a haystack, it's in a needlestack. With parcels already regularly disappearing, a few more would raise no further suspicions.
You may be wondering why I wanted to steal my neighbours' parcels. Firstly, as I already suggested, if I didn't steal them someone else probably would. Secondly, this is a really cool use of AI tech, and having conceived the idea, it would be stupid not to execute it. Sure, some people would have their stuff stolen, but that's the price of progress.
I did have an ethical debate with myself about whether I should reveal how my scheme works, but you can take a guess at how long that lasted. Besides, information wants to be free. It's like the argument over whether 3D-printed guns are a bad idea or not. The conclusion is that other people will eventually work it out anyway, so why not be the first to give it to the public domain?
With the hand-wringing dispensed with, here's what I did:
There's an email group chat for the building. When residents see they have a delivery coming and they're at work or wherever, they often ask the group whether someone can collect their parcel before it's stolen. There follows a narrow window of opportunity, during which agentic AI steps in.
I have an agent set up to monitor the email group and identify collection requests. This agent then makes its best guess as to the time of delivery and passes it, plus the flat's address, to another agent. This agent contacts a preferential list of organisations I won't name, but they specialise in supplying people for odd jobs. Think assembling IKEA furniture or mowing the lawn. They can also make collections.
To cover my tracks, the same agent also arranges another person to whom the collected parcel is handed in a random local area. The agent passes that person's details on to a third agent, which looks up the nearest post office using Google Maps and directs them there. Once at the post office, they post the parcel to a package consolidation company in Oregon.
I have a fourth agent which monitors the package consolidation company. When it reaches the sweet spot for shipping costs by weight and volume, it arranges for it to be delivered to Artie Fisher.
Artie Fisher is of course another agent, but one which I have hidden behind a legal identity and who rents a dilapidated cottage in North Wales. Of course there isn't actually anybody there to receive the parcel, so my second agent returns into play to arrange collection. The collected parcel is taken to a small village post office and finally posted to me.
It's the perfect crime, apart from the running costs. Keeping the AI agents online is expensive enough, but the services it procures are even more expensive. At the time of writing it costs 100x more than the value of stolen goods I receive, and half of those are Graze boxes. Still, such is life at the bleeding edge of technology. Without pioneers like myself, this may never make it to the next level, which I envisage will bring vastly reduced costs by assembling everything into a turnkey platform. Theft as a Service, or TaaS.
Eventually its capabilities could be so generalised that it could steal from anybody, not just your neighbours. Sure, it's not ethical, but it's inarguably technological progress, and you're either for progress or you're a luddite. More specifically, you're a luddite without a hammer to smash the machine, because you ordered the hammer from Amazon and my AI agents stole it.
