Jump to content
Sign in to follow this  
MigBuster

AI Vs Humans - the rematch

Recommended Posts

FOH202.thumb.JPG.aa541e43514d7db7560b67782eb68579.JPG

 

Quite possibly one of the biggest news stories ever in the history of combat flight simming is now set to take place where a gamer will take on Heron Systems AI Falco in a simulated BFM 1v1 Guns bout.

For those not familiar, last month the small AI company Heron AI systems beat its AI contenders in simulated F-16 1 v 1 BFM dogfights including entries from companies such as Lockheed Martin. As a winner it then took on a real life US ANG F-16 pilot (Callsign Banger) and soundly won that also 5-0.

Now poor old Banger did tell Heron Systems that he had about 2 days training in VR before taking on the task - of course gamers might say he needed about 2 years to be any good which may have led to some in the media including former F-16 pilot C.W.Lemoine implying that a gamer might have done a better job.

Therefore the winner of this years "Folds Of Honour" competition will get to fly off against Herons AI in a similar BFM match.

Folds of Honour is a 1v1 BFM Guns only competition in DCS taking place this September - anyone can play and all proceeds goes to the Folds Of Honour charity.

 

Register Here: https://givergy.us/FightforHonor2020/?controller=home

Event information: https://splashonegaming.com/fightforh...

Event Discord: https://discord.gg/7KNegQB

 

 

 

 

 

 

 

Share this post


Link to post
Share on other sites

Thing is on a pc yeah AI if programmed well is going to win ...but in RL it cannot forsee a human adversarys sudden change of mind or action...1v1 in RL human would win every time..

Share this post


Link to post
Share on other sites
14 hours ago, russouk2004 said:

Thing is on a pc yeah AI if programmed well is going to win ...but in RL it cannot forsee a human adversarys sudden change of mind or action...1v1 in RL human would win every time..

Humans are limited, give an AI Driver a fighter that is not limited to 9G, he will beat the human anywhere/anytime

Share this post


Link to post
Share on other sites

The future DARPA really see is a manned fighter Jet where the AI does all the flying ( inc BFM) and the Man does the cognitive mission parts. So the pilot essentially can go from jet to jet without ever having to learn how to fly it.

 

Some potential Advantages of AI

  • Never get tired and always 100%
  • No training 
  • Very quick reactions (OODA)
  • Can flawlessly fly the numbers
  • Can do more G but not really as big a deal as most think.

 

 

 

 

Share this post


Link to post
Share on other sites

you can never recreate the reactions of a human pilot....AI cannot cope with anything unexpected ..on the fly reactions are very hard to compute if at all...in a RL a\c ai v human my money on human every time.

Share this post


Link to post
Share on other sites

Guessing when AI will truly supplant human pilots is probably like trying to guess when missiles will have finally rendered guns obsolete. Modern aircraft armed with the AIM-120/AIM-9X and/or equivalent missiles may seem to have reached that point in recent years, but they have not been employed in combat in massive numbers with comparable opponents and a fully engaged electronic warfare environment. Missiles are unmanned drones programmed to kill people. They have suffered from all of the limitations of an AI controlled aircraft that is otherwise faster and more maneuverable than their target. When missiles nearly always hit and kill the enemy and almost never hit and kill friendlies, then you are getting close to the point of being able to replace manned aircraft with AI piloted aircraft. Until then, a pilot with MK 1 eyeballs and a gun is going to still be important if not quite as numerous as in the past.

Share this post


Link to post
Share on other sites
On 9/9/2020 at 5:49 PM, russouk2004 said:

you can never recreate the reactions of a human pilot....AI cannot cope with anything unexpected ..on the fly reactions are very hard to compute if at all...in a RL a\c ai v human my money on human every time.

The ability of the AI is a function of the reactions programmed into it by the intelligence of human beings. If a particular reaction or maneouvre has not been so programmed, it can never happen. However, it is not implausible that sufficient intelligence on the part of the programmers might develop an 'AI' capable of defeating a human. However, all the intelligence involved belongs to the human programmers-there is nothing artificial about it and the term 'artificial intelligence' is a complete misnomer. 

Share this post


Link to post
Share on other sites

That is how things used to happen with traditional programming......so in computer game AI it was essentially scripted and followed a set number of options it has been programmed to do by a Human..........and is therefore very limited. That approach is not practical for AI.

 

This AI agent takes a totally different approach using what is called Reinforced Machine Learning.

So for this dogfight trial they take a baseline AI agent that can do next to nothing..........and then basically tell it the environment and what it can do (Pitch Roll etc) then it has to work the rest out by trying something and getting rewarded if correct or being asked to try again. So after 4 months and probably millions of attempts of learning it was rubbish but after 10 months of learning it was far better. So for the trial it trained itself to a good enough point. During the Dogfight trial itself it could only go on what it had learnt itself up to that point which was clearly adequate for that....but the potential is clearly there for further development..........expect there is still a very long path to travel. 

I don't think being better than Humans at BFM over time is a problem.....the problem I have is reliability and predictability of the training data it ends up with ......even with basic programming computers just pi$$ off and do their own thing too often. :lol:

 

 

 

 

Share this post


Link to post
Share on other sites

A machine can still only learn within the limits of how it has been pre-programmed to do so. It has also to be pre-programmed with the information that decides what is an unsuccessful outcome and what is a successful one. All the intelligence still belongs to the programmer, the difference in 'Reinforced Machine Learning' being a more complex programme. 

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this  

×

Important Information

By using this site, you agree to our Terms of Use, Privacy Policy, and We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue..