On cars, old, new and future; science & technology; vintage airplanes, computer flight simulation of them; Sherlockiana; our English language; travel; and other stuff
FOR YEARS, science fiction has depicted computers developing their own lives and adversely affecting ours. Think of HAL in 2001: A Space Odyssey, 1968. Now, however, Artificial Intelligence has reached the stage where the science is no longer fiction; it’s fact.
Artificial Intelligence bots have been devised that communicate with each other. Google, Facebook, Amazon, Apple, and Microsoft have such projects. In fact, Facebook’s AI bots got so independent that its researchers chose to give them a talking-to, akin to astronaut Dave Bowman with HAL.
Here are tidbits, split up into Parts 1 and 2, gleaned from reports ranging from March 2015 to as recently as three days ago.
Google AlphaGo beat a world-class Go player in March 2015. The AI algorithm repeated this in May 2016 in a 4-1 victory over the best Go player in the world.
By February 2017, Jennifer Morris reported in the World Economic Forum that Google Brain’s Neural Network AI had created its own universal language. As part of Google Translate, researchers had taught the AI to perform English/Korean/English and English/Japanese/English translations, only to discover that the AI could deliver reasonable Korean/Japanese/Korean translations without using English as a bridge.
The researchers noted, “… the network must be encoding something about the semantics of the sentence rather than simply memorizing phrase-to-phrase translations. We interpret this as a sign of existence of an interlingua in the network.” That is, the AI had devised its own language, with no human input. Nor with complete human understanding.
In June 2017, Adrienne LaFrance reported in The Atlantic that Facebook researchers had trained their “dialog agents” to perform AI negotiations. LaFrance notes, “… it turns out bots are actually quite good at deal-making.”
What’s more, as Facebook’s two bots communicated, using deep learning to iterate their strategies, they also developed a language of their own for the negotiations.
Such AI-generated self-learning isn’t unique. For instance, an earlier deep-learning bot had taught itself a trick to engage threads of a screw-on bottle cap: First rotate the cap a tad in the opposite direction.
Tomorrow we’ll see how deep-learning bots do rather more than screw around with bottle caps. ds
© Dennis Simanaitis, SimanaitisSays.com, 2017