Skip to main content Smithsonian Institution

A Neural Network Attempted to Write the Next Game of Thrones Book

Online Media

Catalog Data

Smithsonian Magazine  Search this
Blog posts
Smithsonian staff publications
Blog posts
Published Date:
Wed, 30 Aug 2017 17:15:04 +0000
Blog Post Category:
Smart News
Smart News Ideas & Innovations
<p dir="ltr">Since 2011, fans of George R.R. Martin's epic book series, <em>A Song of Ice and Fire</em>, have been eagerly awaiting the next installment—all while the hit HBO series "Game of Thrones" continues on its own narrative path. But when that wait will end is still unclear. So to help fill the void, a software engineer trained an artificial intelligence bot to write its own version of the forthcoming novel.</p><p dir="ltr"><span id="docs-internal-guid-2e0f0a80-303f-227e-dd55-8f810e827670">Software engineer (and <em>ASOIAF </em>fan) Zack Thoutt was inspired to create the bot after taking a course on artificial intelligence, writes Sam Hill for </span><em><a href="" target="_blank">Motherboard</a></em>. He programmed the bot as a <a href="" target="_blank">artificial neural network</a>, a setup consisting of thousands of different data nodes that can work in tandem to process data.</p><p dir="ltr"><span id="docs-internal-guid-2e0f0a80-303f-227e-dd55-8f810e827670">Unlike a computer that has to be programmed, neural networks can modify their responses over time using databases fed into the system, similar to learning. To train his bot to write the next <em>ASOIAF </em>sequel, Martin fed the neural network all 5,376 pages of the previous five books to give it a sense of the characters, places and writing style, reports Hill. For each chapter the AI generated, Thoutt gave the bot a word count and a so-called "prime word" to kick off the section. Then he set it off to its own devices.</span></p><p dir="ltr"><span id="docs-internal-guid-2e0f0a80-303f-227e-dd55-8f810e827670">Thoutt has posted five short chapters online so far, </span>and the results wouldn't win literary awards. E<span>ach one is largely readable, if awkward, and chock full of nonsensically dramatic quotes and descriptions. For example, this meandering detail takes place in the second chapter: "</span>The dog wandered the stair, to allow the high officers to help you at home. the woods are gowned on bloody yellow and glass. It may be fewer as well as the north."</p><p dir="ltr"><span>Eerily, however, several of the predictions made by the bot for the next book mirror popular fan theories about what will happen to favorite characters, such as whether one will end up riding a dragon or another may get poisoned by a close adviser. It also managed to create an entirely new character called Greenbeard.</span></p><p dir="ltr"><span id="docs-internal-guid-2e0f0a80-303f-227e-dd55-8f810e827670">Martin is known for his lush, descriptive prose (especially </span><a href="" target="_blank">when it comes to food</a>), giving the nerual network a whopping 32,000 unique words to ingest. And though oftentimes larger training sets are better for AI, the complexity of so many of these words complicated the training process, Hill reports. Neural networks generally work better with simple words, Thoutt notes, and fewer imaginary locations.</p><p><span id="docs-internal-guid-2e0f0a80-303f-227e-dd55-8f810e827670">If these AI </span><em>ASOIAF </em>chapters didn't quite fill the void, never fear. The <span style="font-size: 1em;">website <em>Inverse</em> has joined forces with a San Francisco company </span><a href="" rel="noopener" target="_blank">Unanimous A.I.</a> ​<span style="font-size: 1em;">to makes predictions about what will happen in the final season of the GOT TV series, writes John Bonazzo for the </span><em style="font-size: 1em;"><a href="" target="_blank">New York Observer</a></em><span style="font-size: 1em;">. T</span>he team is collecting predictions ​to <span style="font-size: 1em;">draw on the </span><a href="" style="font-size: 1em;" target="_blank">power of a "hive mind"</a><span style="font-size: 1em;"> and make guesses about what will happen next. </span></p><p><span style="font-size: 1em;">How accurate will all these predictions be? We'll all have to stay tuned to find out.</span></p>
Search this
See more posts:
Smithsonian Article Database
Data Source:
Smithsonian Magazine