Hacker News Reader

Comment by stanfordkid | Hacker News Reader

I do worry that there is opportunity for massive "disruption" for fighter jets in precisely the Clayton Christensen / Innovator's dilemma sense.

The innovation shouldn't be around stealth and maneuverability, but rather production cost/rate and AI automation. It is really hard to train a fighter pilot, this means you have very few of them, and the one with the best jet wins.

In the new world of AI the jets themselves are trivial, it's the AI that matters and the VOLUME you can produce them at. In this case someone like China is way better positioned -- just steal the IP for the physical airframe (or develop one that is just good enough) then innovate on the assembly line and AI aspects.

The AI fighter jet problem feels much easier than the self driving car problem.

View on HN

This kind of overlooks that the cost of the aircraft isn't entirely in the original production and (possible) training of pilots. They also require a large ground team burning several man-hours for every hour that they are in the air. The problem with the mass AI fighter drone scenario is that the logistics chain and maintenance costs kill you.

Except that the drones themselves can be much simpler than the fighters they replace - there is no meatware inside, no life support or safety equipment that needs to be maintained and checked. With enough intelligence built in, the drone can land with a very detailed list of items that need maintenance because of performance anomalies.

Plus, the AI can be augmented by a human on the ground.

That life support stuff has to be replaced by automation, so "much simpler" is probably stretching it. Plus you're talking about adding in a bunch of sensors and software which is even more complexity (IE things that will break or go wrong). The exploding complexity problem is a pernicious one.

If the sensors and computers that manage the control surfaces of any modern fighter fail, the plane crashes anyway, so they still need to be there. Not having to route everything to the cockpit can simplify design and improve ruggedness.

A lot of the resistance against this comes from the fact robots don't get medals.

> so "much simpler" is probably stretching it.

What if you used AI to design the airframe itself, optimising for the simplest and least expensive way?

Totally. Just the HUD systems on the F-35 cost tens of millions of dollars.

My thought is that there is just ZERO innovation in this area at a big picture level. Sure the person designing the radar has a PhD from Caltech -- but it's mostly incremental optimization of individual components.

Someone is going to figure out how to be the Elon Musk of fighter jets and we better hope it's going to be someone from the USA.

But again -- it doesn't matter how good your tech is if the next person can do it 85% as good for 20% the price.

My last major point is this: Fighter jets are limited in their maneuverability when they have humans on-board -- they can't pull more G's than the body can withstand. AI fighters don't have this limitation.

Reply on news.ycombinator.com