Bump in the night? AI can help

Bump in the night? AI can help with that.

This story starts with a scary clown. Not pennywise, a real one. I was a young newlywed college student living in Sparks, Nevada just outside Reno. I was running an e-commerce company while going to school, which is really how I learned how to program in LAMP.

Since I was hosting my website and FTP server out of my house (don’t make fun, this was 2003) I wanted fast internet. I upgraded to the latest with a download speed of 3MB, it was screaming fast. Today I have 1000MB up/down and I have a few neighbors who are rumored to have 10Gig. Times have changed.

One night my wife and I were jolted awake by an evil laughing clown. It was so loud, we were both terrified. My wife was pushing me out of bed to go and figure this out and I remember resisting. I was convinced I was going to open my bedroom door and see a laughing toy clown on the floor in the hall, maybe something like this:

That wouldn’t really be scary, but why is it there? Who put it there? My house is my castle and it was just invaded… all of that is really scary. It turned out it was my office computer, some terrible advertisement banner on this shady free music streamer I was using blasted on at 3am.

Fast forward 17 years later, my wife wakes me up again. I guess this night I’ve been snoring logs (more common these days) and I didn’t hear anything. She thinks someone is knocking on our front door at 3am. I slowly get up, go down stairs in the dark, look out the window, nobody is there. While walking up stairs I think, next time AI will do that for me instead.

The promise of a mature AI security is around it’s usefulness. It will be actionable, real-time, accurate, and reliable. In the future when there is a bump in the night I don’t think anyone will even get out of bed, if anything the AI perfected white noise cancelation will make me oblivious to any false alarm.

Castle wall:

When I visited South Africa last year I was really surprised to see the homes in Pretoria (Elon Musk’s home town) with large walls, electric fence, and signs that said “Armed Response”. Like this:

If I installed an electric fence around my property in Utah and posted armed response signs I would be a top priority for the city to figure out how to get me to legally remove these security measures. In South Africa, depending on the region you are in, these measures are completely justified. When it comes to AI I need my castle wall protected, any potential outsider, day or night, rain or snow, has to be identified and assessed for risk. Then if my home is breached (broken window, or front door) AI needs to identify that scenario accurately as well.

False Positives:

In the future the false positives that trick an AI system today will slowly go away. Today things like snow, excessive rain, moths, cats, and raccoons will trigger my outdoor security systems. Glass breakers can be trigger by a champagne bottle being put into an ice bucket. False alarms happen too often today. In the future everything will be actionable, and reliable. I won’t even have to wake up for most of this, but alien invasion, my AI home better wake me up.

Threat Interrogation:

If a threat cross the threshold it may make perfect sense for a dog like robot or flying drone to go and investigate. That probably feels like completely science fiction, but I’d argue in the next 10–20 years that would feel very common. This allows for even the most advanced intrusion to be interrogated. Think you can trick my outdoor cameras? Think you can trick or disable my electricity to bypass my glass breakers? No worries, I have a battery security agent that is off the grid, has better sound echo-location that any living thing, and can detect an unregistered heart beat on a different floor. Of course it will also have full access to ALL of the sensory systems in my home, it sees and senses everything my entire home sees in real-time. Good luck hiding from it when it comes trotting or flying downstairs.

Escalation Options:

The standard audio alarm can get an upgrade where logical pathways can be external flood lights, all lights in the home, fake dog barking, or your home making the intruder think that multiple people are waking up and coming down the stairs. There are plenty of interesting escalation options where force is not even needed to rattle an intruder into fleeing the scene.

Force/Lethal Response Ethics:

What if they don’t leave? I have learned ethics are not global. Something considered unethical in the US could be welcomed in South Africa. Mace response? Sure. Lethal response? Again, it depends on the country. Here is a screenshot of a carjacking system that is potentially lethal and supported by courts in South Africa. If you get car jacked you have a switch that will burn your attacker. In the US we can’t relate to this, but we also haven’t had the number of violent carjacking cases like they have had, so really it isn’t our place to insert our ethical opinion on their reality.

So the allowed response of your home’s AI watchdog will be different based on the country and the perceived risk to the home inhabitants. This is brand new territory for AI ethics, even in the US if your AI systems determines that you are an armed intruder with an unregistered heart ♥️ rhythm (you obviously don’t live here) and you are ignoring commands to drop your lethal weapon I wouldn’t think it would be completely out of the question to have an aggressive response. Ideally a mace cloud or air taser, but we might see systems that are willing to to do more if the intruder is uncooperative and the local courts allow it. I don’t know what this future will look like, but I do know I will have fewer false alarms. Maybe a safer scenario would be a very aggressive defense strategy (every bedroom get’s auto-locked by your home and the doors are strong enough the intruder can’t kick them in), then a live police scanner blaring across your home’s PA system, scaring them into thinking the police are seconds away.

Jump into the comments on technology thoughts around security, what do you think will be allowed with escalation? What about country specific ethics, e.g. South Africa vs US/EU. Would you sleep better knowing every room in your house was being monitored by AI including all external areas of your home? Imagine a bump in the night where your alarm systems announces to you what the perceived threat is.

Ben is a cofounder at Zeff.ai, delivering automated deep learning into production. Ben is a recognized deep-learning expert and keynote speaker.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store