Deng Simiao Sent from Vice Driver Temple
Intelligent Car Reference | Public Account AI4Auto
Tesla’s AI chief and Musk’s right-hand man, Andrej Karpathy, has reportedly taken a rare vacation, a long one of four months.
Netizens expressed concern because Doug Field, Tesla’s senior engineer, didn’t return after only taking a month-long leave.
If Karpathy leaves, there is no doubt that Tesla’s Autopilot and FSD will be greatly affected.
Perhaps for fear of this, Musk took to Twitter to announce the news and clarified the truth – Karpathy is indeed just on vacation.
Karpathy himself also said that during this time, he is just taking a good rest and looking forward to returning to work soon.
Who is Andrej Karpathy?
Andrej Karpathy was born in Czechoslovakia in 1986 and moved to Toronto, Canada with his family at the age of 15.
Karpathy’s academic background is very rich. He has pursued his Bachelor’s, Master’s, and Ph.D. degrees in two different countries and three different universities. He graduated with a Bachelor’s degree in computer science and physics from the University of Toronto in 2009, a Master’s degree in computer science from the University of British Columbia in 2011, and a Ph.D. in computer science from Stanford University in 2016.
After completing his master’s degree, Karpathy interned at Google, where he worked mainly on unsupervised deep learning, supervised deep learning, computer vision, and other related areas.
During his Ph.D. at Stanford University, Karpathy studied under Professor Li Fei-Fei from the Stanford Vision and Learning Lab. Later, Karpathy also served as a lecturer for the university’s deep learning course. This course was the first in Stanford’s history and was designed and taught by him personally. The course, titled “CS231n: Convolutional Neural Networks for Visual Recognition,” has attracted an increasing number of students, from 150 in 2015 to 750 in 2017.
In January 2016, just before he was due to graduate, Karpathy joined OpenAI, which was initiated and established by Musk, and became one of the company’s founding members. Initially, he was mainly responsible for recruiting employees, but later served as a research scientist in the company, responsible for deep learning and deep reinforcement learning of the generative model.
It is worth noting that OpenAI is a nonprofit artificial intelligence research organization aimed at promoting and developing artificial intelligence for the benefit of humanity.At the beginning of its establishment, Tesla aimed to emulate DeepMind, and later achieved explosive AI results such as GPT-3.
After Joining Tesla
After working at OpenAI for a year and a half, Karpathy joined Tesla as the head of artificial intelligence, deep learning, and computer vision, responsible for leading Tesla’s computer vision team for Autopilot.
AI vision is crucial for Tesla’s full self-driving (FSD) function. This vision system processes and aggregates video data from 8 cameras to create high-definition 3D maps that reflect the speed and direction of surrounding objects.
So, what does this team do specifically?
They focus on data collection and neural network training, and firstly, they need to collect data, such as internal data labeling and custom labeling interfaces. Then, they conduct neural network training, and finally, they make it work by segmenting, detecting, and conducting 3D/depth predictions, and arrange production and deployment of chips.
Karpathy said, “Autopilot has improved driving safety and convenience to some extent, but our team’s goal is to accelerate the development of FSD and deploy it to our own millions of vehicles.”
In short, Karpathy is like Musk’s right-hand man, influencing the future and direction of Tesla’s autopilot and various AI products.
However, Tesla’s Autopilot is not perfect, with at least 12 significant traffic accidents related to Autopilot being confirmed by the US National Highway Traffic Safety Administration. Last Saturday, Tesla’s FSD Beta was launched in Canada, marking the expansion of FSD functions to areas outside the United States.
It’s worth mentioning that Musk also announced plans to launch its “Optimus Prime” humanoid robot, Tesla Bot, next year in 2023.
So why was Karpathy’s vacation garnering so much attention in such a critical period?
Karpathy explained that he took the break for two reasons: to take a break, relax, and travel, and to spend some time on self-improvement.On Twitter, he wrote, “After working at Tesla for almost five years, I want to take a break and go on vacation for a while, maybe travel and focus on improving myself, like enhancing my technical skills and training neural networks. Although I really miss Tesla’s robots and the Dojo supercomputer cluster, and I can’t wait to see them again!”
Regarding the latest news about Kappas, he said on Twitter, “I’m currently taking a ‘digital nomad’ trip. Packing my bags, heading east, saying hi to friends, reading papers, and coding along the way. I’m still in the UK, then traveling to Europe and Asia, and finally returning to the San Francisco Bay Area.”
In short, “I will return to Tesla.”
What are Kappas’ goals after vacation?
After four months of vacation, Kappas has two important projects to complete.
One is the deployment of the Dojo supercomputer, which will be put into use next year.
This computer is expected to set a world record, replacing Japan’s “Fugaku” as the world’s fastest supercomputer.
This supercomputer can store a large amount of video data, which can be used to train the neural networks for Autopilot and FSD, thereby accelerating the development process.
The other goal is to update the hardware of FSD and release version 4.0. The computing power of this version can reach up to four times that of the original 3.0 version and can be installed on Tesla’s Cybertruck in the future.
The 2.5 version was released in 2017, and the 3.0 version was released in 2019. It has been three years since the last version update.
When it comes to this 4.0 version, Musk said he hopes it can drive safely, even better than humans. But at this stage, it is still being improved, and he hopes to develop more powerful computers in the future.
–End–
This article is a translation by ChatGPT of a Chinese report from 42HOW. If you have any questions about it, please email bd@42how.com.