Just read this article in The Register and immediately thought of this thread. It is an interesting story on AI...
1 Like
if you set simple objectives to a neural network, you'll get a simplistic response
No s*** Sherlock!
If this actually happened (there is some doubt it seems), it would be a good lesson in not using simplistic gamification to let loose the dogs of war (let alone a future armed AI).
As I mentioned to someone at work, this is Sci-Fi rule 101 for dealing with AI's!
The doubt has been added since I first read the report!!
Still a good lesson though, even if it is a made up story. Just the kind of thing we can imagine the military doing! (and apologies to anyone in or formally in the military for painting everyone with the same brush!).
1 Like
Ah the ol Kobayashi Maru .. Captain Kirk would be proud
2 Likes
It's impossible Dave. Unless you cheat 
Years of operations research work for defense and homeland security organizations, here. This has as much to do with human intelligence/learning as the machine variety. When you reward something, you get more of it. The bigger the reward, the larger the acceptable penalty.
1 Like
Apparently that never happened, it was a case of criti-hype.
The source for that statement also talks about bias in AI:
Algorithmic decision-support systems and classifiers can replicate the bias in their training data to reproduce racism, sexism, and other forms of discrimination at scale.
and
Machine learning systems donāt just reproduce and amplify bias
Interesting read...
This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.
