AI NEWS PAGE DOSSIER
Andrej Karpathy
Andrej Karpathy's file: Hinton's classroom, Stanford vision research, early OpenAI, Tesla Autopilot, a return to OpenAI, and then teaching the internet how the machines work.
Karpathy learned deep learning around Geoffrey Hinton’s Toronto orbit, built one of Stanford’s defining neural-network courses, joined OpenAI at the beginning, ran Tesla’s vision effort, returned to OpenAI for GPT-4 work, then became one of the internet’s most influential explainers of how modern AI actually works.
HOW THE HELL DID ANDREJ KARPATHY GET HERE?
2005
HE FINDS DEEP LEARNING IN HINTON'S ORBIT
Karpathy began his University of Toronto computer-science and physics studies in 2005 and says Geoffrey Hinton’s class and reading groups were where he first got into deep learning.
RECEIPTS: Andrej Karpathyfirst-party
2011
STANFORD: VISION, LANGUAGE AND NEURAL NETS
Karpathy began doctoral work at Stanford on convolutional and recurrent neural networks, computer vision and language, while also interning at Google Brain and DeepMind.
RECEIPTS: Andrej Karpathyfirst-party
2015
HE BUILDS STANFORD'S DEEP-LEARNING CLASS
Karpathy designed and became the primary instructor for Stanford’s CS231n, which grew into one of the university’s most popular deep-learning courses.
RECEIPTS: Andrej Karpathyfirst-party
2015
HE JOINS OPENAI AT THE BEGINNING
Karpathy became a research scientist and founding member of OpenAI, remaining there through 2017.
RECEIPTS: Andrej Karpathyfirst-party
2017
TESLA HANDS HIM THE VISION STACK
Karpathy became Tesla’s Director of AI and led the computer-vision team behind Autopilot, including labeling, training and deployment on Tesla’s inference hardware.
RECEIPTS: Andrej Karpathyfirst-party
2023
BACK TO OPENAI FOR GPT-4
Karpathy returned to OpenAI in 2023 and says he built a team working on midtraining and synthetic-data generation.
RECEIPTS: Andrej Karpathyfirst-party
Jul 16, 2024
HE TURNS FROM BUILDING MODELS TO TEACHING EVERYONE ABOUT THEM
Karpathy launched Eureka Labs, an AI-native education project, while expanding his public teaching on large language models and neural networks.
RECEIPTS: Eureka Labsfirst-party