Not-so-Natural Language and politics

In March 2016, nearing the final footsteps of an already exhausting Presidential election cycle,  Bradley Hayes, a robotics researcher at MIT, launched Deep Drumph – an artificial intelligence (AI) program and associated mock campaign to raise money for GirlsWhoCode.

Hayes was able to demonstrate the absurdity of the discourse of Donald Trump by the ability of an artificial intelligence program to effectively recompose and represent his transcribed diatribes in a mass of AI-crafted Twitter posts. 

Simply, it's a difficult thing to do.

Despite the complexity in crafting natural language (NL) systems, Hayes was amused to discover the linguistic tendencies of some individuals, like Presidential candidate, Trump, are consistent enough to be believably replicated by his program. Perusing the @DeepDrumph Twitter feed – the social channel hosting the outputs of the AI program – one may be struck by the accord between the post content and the man himself.

The true magnificence of this effort is its ability of the AI at hand to parse and regenerate seemingly coherent language that could be representative of a human form. I've worked on projects focused on building NL systems to be algorithmically implemented by AI, and it's incredibly difficult; the ways in which human beings create and externalize emotional and intellectual expressions is little short of fucking amazing. And complex!

You may quote me on that.

So if at the close of the 2016 U.S. Presidential election to find yourself disillusioned and fearful for the future of our political system, you may find joy in the unexpectedly impressive consequences of Trump's predictability unpredictable language patterns and the ongoing development of evermore-accomplished Artificial Intelligence programs.