
Andrej Karpathy joined Anthropic’s pre-training team led by Nick Joseph. He will create a new group focused on a recursive approach: using Claude itself to accelerate pre-training research. Pre-training is described as the most expensive, compute-intensive phase in building frontier models, and improving its speed and efficiency could change the economics of large language model development. Karpathy previously co-founded OpenAI in 2015, worked on deep learning research, and later led AI and computer vision efforts at Tesla, including teams behind Autopilot and Full Self-Driving. He later returned to OpenAI briefly, then founded Eureka Labs in 2024. He also expressed continued passion for education and plans to resume it.
"Karpathy is joining Anthropic's pre-training team, led by Nick Joseph, where he will build a brand-new group focused on a strikingly recursive goal: using Claude itself to accelerate pre-training research. Pre-training, the massive compute-intensive phase that gives a frontier model its core knowledge and capabilities, is the single most expensive part of building systems like Claude. Finding ways to make that process faster and more efficient could reshape the economics of the entire AI industry."
"In an X post that racked up 13.6 million views, Karpathy wrote that he believes "the next few years at the frontier of LLMs will be especially formative." He added that he remains "deeply passionate about education" and plans to resume that work in time."
"Karpathy earned his PhD at Stanford under Fei-Fei Li, the computer scientist behind ImageNet, focusing on deep learning and computer vision. He was among the 11 people who co-founded OpenAI in 2015, where he worked on deep learning research before departing in 2017 to join Tesla as its director of AI."
"At Tesla, Karpathy led the computer vision teams behind Full Self-Driving and Autopilot, the programmes that underpin the electric carmaker's ambitions for autonomous vehicles. He left in July 2022, returned to OpenAI for roughly a year, then departed again in 2024 to found Eureka Labs, a start"
Read at TNW | Anthropic
Unable to calculate read time
Collection
[
|
...
]