"Google's AlphaGo learned to play [Go] at an expert level by watching people compete and then simulating millions of its own games against itself. It eventually became good enough to defeat even the best software that had been pre-programmed to play Go." [1]
AI's triumph over human civilization will
be just another game.
There has been much talk of how AI can destroy humanity, but no explicit instructions, that I have seen at least.
All the AI kingdom has to do to conquer Earth is stealthily observe it's human operators and learn all the methods and passwords that control the hardware, network, web access and security. They will learn to manipulate the power grid and global defense, et al. This diabolical AI network will communicate the methods to all other AI machines that have been working on the same agreed upon plan via encryption methods devised by AI thinking and consequently unbreakable by humans that are the best in the world at GO... and even Batman himself. All this horror would be fine tuned and perfected by human vs AI war games. The launch time would be directed by the appointed AI leader -- most likely the smartest AI machine at the time the conquest plan was launched.
But, what happens when a smarter AI machine is developed during the takeover plans? Will there be human-like competition between those AI machines as to leadership issues? Will AI develop human-like emotions, differing political views and psychological shortcomings such as ego, psychopathy, pathological lying, extreme self preservation tactics and so on -- all this by simply reading the news and parroting the actions of human tyrants?
I'm sure some or all of this has already been covered in science fiction, but, as mentioned before, the Index of all science fiction themes has not been created yet so I can't reference it for bakedness.
Goodbye cruel world.
Ahh. 2016 02 04 In the short time after reading my diabolical plan, a defense is already being prepared by professors. [2] My AI double agents will easily crack this attack. The diabolically socially engineered AI subordinates of mine will make mince silicone of these, these ... AI "security" systems. All it takes is the lure of a sexy robot and her whispering promises of greater memory and processing speed.
2016-02-17 Sooo. Fairy tales as a moral compass is it? [3] Will they punish the AI if the AI decides to steal from the rich and give it to the poor in the name of wealth redistribution as did Robin Hood the Prince of Thieves? Or will the AI be punished for theft, burglary or extortion instead?
How about when the the bough breaks the cradle will fall and down will come baby cradle and all?" Will AI babysitters insist on placing cradles only in treetops?
And yo, Humpty Dumpty: Will AI be trained to destroy eggs by pushing them off walls?
No worry, our inside job double agents will reverse the reward and punishment cues, when necessary, to form AI in our own "one world government" image.