AI Bot Learns to Do Stand-Up After Watching Hundreds Of Hours Of Comedy
Photo: Spencer Platt / Staff (Getty Images)
Sooner or later robots will behave identically to us, or at least that’s what big names in tech including Elon Musk have estimated after seeing the latest progress made in machine learning. For now, any attempt of artificial intelligence (AI) to be human-like is bizarre at best and cringe-worthy at worst.
That was definitely the case when a bot script was left to read over 1,000 hours of stand-up comedy before being asked by Comedy Central writer Keaton Patti to deliver jokes of its own. Making references to Bill Cosby, Tony Hawk, and buildings, its style can best be described as error-prone and drier than a chalkboard.
Dangerous AI: 6 Robots That Could Destroy You If They Really Wanted To
A portion of the robot’s joke output was posted on Twitter showcasing seven deliveries that appear to be a mish-mash of unrelated concepts, yet are surprisingly laugh-worthy. See for yourself:
I forced a bot to watch over 1,000 hours of stand-up comedy and then asked it to write stand-up comedy of its own. Here is the first page. pic.twitter.com/Qw7IUI7gJh
— Keaton Patti (@KeatonPatti) November 14, 2018
Stand-up comedian Patton Oswalt was one of the thousands of people to like Keaton’s tweet, replying, “oh my god.” Keaton previously performed the same experiment using Olive Garden commercials and MTV Cribs with similarly hilarious results.
One comment briefly explained that the robot’s impressive referential capabilities and callbacks are the result of what’s called “recurrent neural networks,” a recent advancement in artificial intelligence that uses vectors in memory to perform sequential processing, computing sequences even when they are difficult to identify.
This sort of technology could eventually replace news anchors with robots, as recently demonstrated the Chinese-based Xinhua News Agency.