Since leaving the company in July, Tesla’s former chief of Artificial Intelligence, Andrej Karpathy has since embarked on an unexpected new endeavor — sharing his knowledge on YouTube.
Karpathy recently started a YouTube channel and shared his first few videos on Tuesday, causing many to question the former AI chief’s career moves.
In a tweet on Wednesday, @WholeMarsBlog posted a screenshot of one of Karpathy’s videos, saying “Andrej really left Tesla to become an influencer.”
At the time of the screenshot, Karpathy had just 724 subscribers. At the time of writing less than 24 hours later, Karpathy is already up to more than 8,700 subscribers. Looks like he has already mastered the art of the YouTube thumbnail.
Tesla CEO @Elon Musk later commented on the post, saying, “Got to build those YouTube subscriptions!”
Andrej really left Tesla to become an influencer pic.twitter.com/UFJZ0QSrwD
— Whole Mars Catalog (@WholeMarsBlog) August 16, 2022
“!!!! Ok I recorded a (new!) 2h25m lecture on “The spelled-out intro to neural networks and backpropagation: building micrograd,” said Karpathy on Tuesday afternoon. “This is the culmination of about 8 years of obsessing about the best way to explain neural nets and backprop,” he added.
“If you know Python, have a vague recollection of taking some derivatives in your high school, watch this video and not understand backpropagation and the core of neural nets by the end then I will eat a shoe :D,” said the former Tesla director of AI.
When Karpathy left in July, he called it a “difficult decision to part ways,” noting how much the Autopilot program grew while he was at Tesla.
The news came after Karpathy took a four-month sabbatical from Tesla according to statements from Musk.
Karpathy was also featured on the Robot Brains podcast in March, mostly discussing Tesla’s Full Self-Driving (FSD) beta.
Karpathy’s channel includes the video in question, a deep dive into neural networks and micrograd, as well as three stable diffusion dream AI videos. You can watch his two-and-a-half-hour feature on neural networks below.
