During my talk about AI in crisis decision making at Harvard's National Preparedness Leadership Initiative (NPLI) Advanced Leadership Symposium, I was asked about what I am worried about when it comes to AI.
One of my prepared remarks, one I didn't use, was about my concerns about weaponized AI enabled aerial drones deployed by malicious actors domestically.
While we haven't yet seen this experimentally or IRL, the scenario isn't a difficult leap in technology and capacity from what's possible today. With some additional advancements in drone technology, combined with sophisticated artificial intelligence, we have a very near term possibility of aerial drone swarms with high levels of coordination and targeting capabilities. It could be next year or 5 years from now.
It's a lot to swallow, particularly in an introduction to AI in crisis talk, by leaders who are dealing with very real active shooters, vehicle born attacks, political violence and civil unrest.
So I left it out of my talk.
But we need to be vigilant in dealing with the current threats, while still thinking creatively and planning for future threats.
Mitigating and stopping active shooter and vehicle born attacks is already hard.
Can you imagine if, rather than individuals and small groups of shooters or vehicles, you are dealing with hundreds of drones?
These aren't the commercial drones known for photography or delivery or the large military ones we've come to know.
Instead, these drones are weaponized, AI-enabled, and are designed to perform coordinated attacks. Auto-stabilizing for recoil, very accurate.
Each drone in the swarm could be small, agile, and equipped with computer vision capabilities that can identify and differentiate individuals from crowds, even targeting persons with particular demographics or attributes. These drones are programmed to operate collaboratively, share data in real-time, and adjust tactics on the go. If one drone is taken down, others adjust to fill in. Each drone carries a small payload or projectile weapon - enough to harm an individual but not cause mass destruction.
The traditional anti-aircraft and missile defense systems would be too destructive and risk causing collateral damage in an urban environment.
There's also the challenge of medical readiness, considering the potential for mass casualties from such an attack.
Intelligence, responders and emergency management practitioners would be faced with the daunting task of stopping and responding to this -- and it could happen nearly anywhere.
Do you see why I didn't share this at 8 am to a group waiting on coffee?
And this may never happen, or counter measure technology will be developed alongside threats such as this.
But we should't let a lack of creativity about possible threats, or our current threats, keep us from discussing and mitigating emerging AI threats.
Remember, one of the most important failures leading to 9/11 was one of imagination.
Comentarios