Tesla’s new head of AI and Autopilot Vision comments on his new role

Yesterday, Tesla confirmed that it hired Andrej Karpathy as the new head of AI and Autopilot Vision. It’s a significant change for the young scientist who has been mostly focused on research and academics.

He will now have to deliver real-world applications – something that he commented on following the announcement.

Karpathy obtained his PhD from Stanford in Machine Learning with a focus on Deep Learning for Computer Vision and Natural Language Processing last year.

In a short time, he made a name for himself in the space by teaching a new Stanford class on Convolutional Neural Networks for Visual Recognition that became very popular while he was doing his PhD.

He was working for Elon Musk’s new non-profit AI research firm, OpenAI, since September 2016 and it looks like his research impressed Musk enough to hire him at Tesla to turn his neural net expertise into actual real-world applications.

In a Reddit thread about his move to Tesla, someone brought up the fact that Tesla is not really known for its work in AI research and Karpathy commented:

“At least on the short-medium term, the focus will be much more applied than what I’ve done at OpenAI, and will use techniques more along the lines of ConvNets trained with supervised learning, at scale, and deployed on an embedded system. But on a longer term I certainly hope to remain in the research world to some extent!”

A “ConvNet”, or a convolutional neural network, is a type of neural network inspired by the animal visual cortex and it is used in image recognition applications.

Now that Tesla built its own software to enable the Autopilot features on the second generation hardware in new vehicles, the Autopilot program has become, more than anything else, about improving those features using its own neural network fed by data gathered through its fleet of vehicles.

It sounds like something that falls right into Karpathy’s range of expertise and as we recently reported, Tesla has recently opened the floodgates of Autopilot data gathering after a recent update. The timing seems perfect.