Bob Mankoff, the cartoon editor for the New Yorker has an unusual problem: The magazine has this contest where funny captions are sent to them and this is a huge workload for his assistants since they get 5000 entries each week. This led to his assistants burning out and he needed a solution.
Mankoff went ahead and collaborated with Microsoft to develop an AI project with the aim that it would teach a computer what is funny. Apparently, the idea was formulated about a year ago where one Dafna Shahaf, a researcher at Microsoft wondered if she could teach a computer to assess funny captions after attending a speech by Mankoff.
Shahaf later on used the New Yorker’s database to create a system for the AI to understand jokes. When the AI system was used, it was able on average identify 55.8% of the jokes the editors were able to identify. “On average, we saved about 50% percent of his workload. It could also save Mankoff the time it takes to hire new assistants” Shahaf was quoted as saying.
The AI system is currently limited with the New Yorker database but that has not diminished Microsoft’s ambitions for it. They actually plan to train computers come up with their own jokes, which could mean they’ll make Cortana a stand up assistant, maybe? We’ll have to wait to see this happening in the future.
If you want to see the actual findings of the research, you can click here. These findings apparently will be presented on August 13th at a data conference in Australia