[TML] Current USAF fleet was Re: CT Sup 7, Crews room sizes.

[shadow at shadowgard.com]

On 1 May 2008 at 12:27, Tom B wrote:

> Hmmm. A friend of mine in the late 80s was doing his PhD on training
> neural nets to identify air vehicles at different ranges and
> resolution. With enough time and training, they produced some pretty
> good results. I'm betting 20 years has helped a lot. I'm not thinking
> this is a very long term obstacle.
> 
> Keep in mind human pilots often make mistakes (Blackhawk = Hind,
> Scimitar = BMP, PPCLI = Taliban, etc). All you have to do is manage to
> be at least as good. And computers can do a lot with optical image
> recognition now (they don't have to JUST rely on other data like EM
> sigs, radar returns, etc).

No, humans being humans, an error rate that's acceptable for a human 
pilot will be totally unacceptable for a robot pilot.


--
Leonard Erickson (aka shadow)
shadow at shadowgard dot com