Demystifying the state of machine learning: why are machines still stupid? … or are they?

Presented at ToorCamp 2018, June 23, 2018, 1:30 p.m. (20 minutes)

We grossly underestimate just how much multidimensional data we process as a human, for years, that contributes to conscious as well as subconscious, automated learning, in addition to contextual and background knowledge human have gathered evolutionarily. The disconnect between this realization and what we think should be enough for a machine to “learn” in the same way as a human but within much faster timeframe and less data creates unrealistic expectations from ML and machines that causes an inaccurate understanding that machines are inherently stupid.


  • Ren
    Ren is an infrastructure and ops engineer at Arista Networks, busy being a jack-of-all-trades debugging Linux servers and kubernetes clusters at datacenters, all the way to building all sorts of automation beasts that control server fleets and services. Her passion is to make things run faster and faster with even less keystrokes by writing friendly, intelligent automation so that everyone else can have more time to do the things machines can’t do. Like enjoying a nice cup of coffee.


Similar Presentations: