girls laughing

AI and Humor

I can’t decide if this is a triumph for analytics and algorithms or if it is one of those gaps that is ripe for human attention.

Arjun Chandrasekaran from Virginia Tech and pals say they’ve trained a machine-learning algorithm to recognize humorous scenes and even to create them. They say their machine can accurately predict when a scene is funny and when it is not, even though it knows nothing of the social context of what it is seeing.

My Take

Apparently, Arjun Chandrasekaran at Virginia Tech and his research team has trained an algorithm to differentiate visual images that are funny from images that are not. The system uses blind analytics – it has no idea why they are funny or not. It just looks at a massive number of visual primitives – associated with those that have been rated as funny or not, and then builds the algorithm. It can then be launched at new images to make predictions.

As with any research study, they used a narrow scope of visual images. It remains to be seen if it could work on the full set we see on Google Images or television. But their results within this narrow scope were significant and promising.

At first, I was hesitant and skeptical. But then I got to thinking about all of the neuroscience findings that show much of the way we think is not too far different from this (as much as our conscious brains try to convince us otherwise). So maybe it is not so farfetched.

I guess we will have to wait and see.

Your Turn

What is your gut instinct on this? Human factors has a long history of deciding what machines can be better at and what humans can be better at.

Is humor about to switch sides?

Image Credit: Ciaran McGuiggan

2 thoughts on “AI and Humor”

  1. The different approaches make me nervous. Humans approach humour in terms of context. Context is what makes something one moment hilarious and the next moment tragic.

    Machines approach these issues differently. Like a person who has zero empathy, but keen observation skills. People like that make me nervous. So machines also make me nervous, for the same reason.
    They may work – often – but I’m concerned about how significant the errors may be when they do occur. I’m not sure increasing observational skills will ever be a substitute for empathy.

  2. I am not sure how broad the AI looks, but it could be considering context in a way. Not as much as a person could of course.

    Your comment made me think of the sociopaths in the movies that train themselves to respond to humor and other social signals because they can’t recognize it naturally. Some Autism-spectrum individuals have to do this as well.

Leave a Reply

Your email address will not be published. Required fields are marked *