Judah vs. the Machines: Facebook’s dog-breed-recognition robot

On the road to making machines artificially intelligent, everything needs to be taught, even the facial structure of a dog that looks like a mop.

In this episode of Judah vs. the Machines, actor Judah Friedlander sets about his goal of “challenging the world’s top artificial intelligence robots for the sake of humanity” by taking a look at Facebook’s applied machine learning team.

The actor, best known for his portrayal of Frank Rossitano on NBC’s much-loved 30 Rock, descends on the Menlo Park campus at 1 Hacker Way to get a little taste of what life is like working at one of the world’s biggest tech companies. After exhausting all of the campus’s free perks, Friedlander tests the wits of FB’s machine learning team. The AI he’s examining isn’t combating fake news or sensitive content on Facebook Live — it’s computer vision tech doing the all-important task of guessing dog breeds.

To find out what the deal was with Facebook’s dog love, Friedlander had the chance to sit down with Joaquin Candela, who is the director of applied machine learning at Facebook. After discussing some of Facebook’s efforts in promoting their systems’ natural language recognition skills, Friedlander offered his own ideas for the future of communication before taking on the company’s dog-breed-fussing, AI which you can see in the video above.

Facebook’s AI is appropriately not perfect, but does manage to best Friedlander. One of the breeds that befuddles both Judah and the FB AI is the Puli, which is the type of dog Mark Zuckerberg owns (Beast). There’s never really a clear explanation of why Facebook has been training its system to do this, but I guess the better question is why would you not want a “Shazam for dogs”?

See all eight episodes of Judah vs. the Machines here.