Firstly, let’s set the mood. I need you to watch this video.
Go on. Stop what you’re doing and press play. I can wait…
Done? How good was that, right? RIGHT?! Mmmhmm, I knew you’d like it.
Now, onto ALPHA…
I’ll set the scene.
Imagine it’s the mid-eighties. Your name is Dr Myles Dyson and you’ve just invented the neural-net processor. You see your invention as a massive success, a gift to humanity, a major stepping stone across the treacherous waters toward world peace.
… and then Sarah Connor shoots you.
That’s Cyberdyne. This is Psibernetix. My bad. I’ll start again.
University of Cincinnati doctoral graduate Nick Ernest may not have built the neural-net processor (thankfully), but he’s definitely created something on that level. Ernest and his team at Psibernetix have created ALPHA, an AI set to be the ultimate wingman of the sky(net)… which runs on a Raspberry Pi.
Exciting, yes? Let me explain…
ALPHA is an artificial intelligence with the capability to out-manoeuvre even the most seasoned fighter pilot pro, and to prove this, ALPHA was introduced to retired U.S. Air Force pilot Col. Gene Lee in a head-to-head dogfight simulation.
When pitted against Col. Gene Lee, who now works as an instructor and Air Battle Manager for the U.S. Air Force, ALPHA repeatedly shot down the pro, never allowing Lee to get a single shot in.
“I was surprised at how aware and reactive it was. It seemed to be aware of my intentions, and reacting instantly to my changes in flight and missile deployment. It knew how to defeat the shot I was taking. It moved instantly between defensive and offensive actions as needed.”
Before ALPHA, pilots training with simulated missions against AIs would often be able to ‘trick’ the system, understanding the limitations of the technology involved to win over their virtual opponents. However, with ALPHA this was simply not the case, instead leaving Lee exhausted and thoroughly defeated by the simulations.
“I go home feeling washed out. I’m tired, drained, and mentally exhausted. This may be artificial intelligence, but it represents a real challenge.”
Prior to their work alongside Col. Gene Lee, ALPHA was set up against the current AI resources used for training manned and unmanned teams as part of the Air Force research programme. Much like its sessions with Lee, ALPHA outperformed the existing programmes, repeatedly beating the AIs in various situations.
In the long term, ALPHA looks set to continue to advance in the field with additional development options, such as aerodynamic and sensor models, in the works. The aim is for ALPHA to work as an AI wingman for existing pilots. With current pilots hitting speeds of 1,500 miles per hour at altitudes thousands of feet in the air, ALPHA can provide response times that beat their human counterparts by miles; this would allow for Unmanned Combat Aerial Vehicles (UCAVs) to defend pilots against hostile attack in the skies, while learning from enemy action.
This ability to run ALPHA on such a low-budget PC makes the possibility of using the AI in the field all that more achievable. As confirmed by Ernest himself (we emailed him to check), the AI and its algorithms can react to the simulated flight’s events, and eventually real-life situations, with ease, using the processing power of a $35 computer.
And that, ladies and gentlemen, is incredible.
This blog post was bought to you by the 1980s*. You’re most welcome.
*Yes, we know Terminator 2 was released in 1991. Give us some slack.