A software engineer used a recurrent neural network to predict what might happen in the next Game of Thrones book, but the resulting chapters were awkward and demonstrated that AI cannot yet duplicate skilled writing. The engineer admits that the technology to train a perfect text generator does not yet exist.
It will be a while before fans get their next fix for Game of Thrones. The seventh season of the wildly popular fantasy series just wrapped up on HBO, and it will be at least a year (maybe two) before the eighth and final season airs. And The Winds of Winter, George RR Martin’s next installment in the A Song of Ice and Fire novel series still has no confirmed release date. Meanwhile, a software engineer has trained a computer program to digest all things Westeros and come up with what might happen next. The resulting five chapters were awkwardly worded, bordering on gibberish. While the twists and turns are possible given the fantasy epic’s track record of gasp-worthy moments, it also demonstrates that AI still can’t duplicate what a skilled writer can produce.
Fantasy meets technology:
Software engineer Zach Toutt used a recurrent neural network (RNN) to produce the sequel, which predicts that “Varys poisons Daenerys” and that “Jaime kills Cersei.”
A recurrent neural network is a computer program that processes sequences of data, such as text, similar to how the human brain does.
“(AI) makes a lot of mistakes because the technology to train a perfect text generator that can remember complex textures across millions of words doesn’t exist yet,” Thoutt said.
Protect your devices with Threat Protection by NordVPN
