Andrej Karpathy, the former head of AI at Tesla and a researcher at OpenAI, has announced the launch of Eureka Labs, a groundbreaking education platform designed around artificial intelligence. Generally, in tech circles, this means it will be developed with AI as an essential element from the outset. Despite its ambitious goals, Eureka Labs is commencing with a more conventional approach to education.
Based in San Francisco, Eureka Labs was officially registered as an LLC in Delaware on June 21. The startup aims to harness advancements in generative AI to create AI teaching assistants that will help students navigate their coursework.
Karpathy envisions AI assistants that collaborate with human teachers, aiming to enable "anyone to learn anything." While educators will still develop the course content, these AI assistants are designed to provide valuable support. Although Eureka Labs has not yet constructed or tested these AI solutions for classroom use, research from Georgia State University has indicated that AI teaching assistants have positively influenced student grades.
In his announcement on X, Karpathy hinted at a future where these AI assistants could be modeled after real individuals, similar to Meta's unique celebrity chatbots or Character AI’s character-based chatbots. However, details regarding Eureka Labs, such as whether it is simply offering a MOOC with chatbot functionality or a program intended for high school applications, remain sparse on both his post and the minimalist Eureka website.
In response to inquiries for additional details, Karpathy has not yet provided any information to media.
Accompanying his announcement, Karpathy showcased what seems to be an AI-generated image depicting a futuristic school featuring a spaceship-like structure, solar panels throughout (even on the floors), and an image of a smiling girl with... is that three hands?
While Eureka Labs aims to develop AI teaching assistants, Karpathy revealed that the initial product will be an AI course titled LLM101n. This undergraduate class is designed to guide students in training their own AI models. According to Karpathy, this "mini-me" AI will resemble the assistants Eureka Labs ultimately intends to create. Course materials will be available online, and the startup plans to host both digital and in-person groups working through the curriculum together.
However, the link provided for this AI course leads to a GitHub repository that suggests a different focus than expected. Instead of offering instructions on building an AI assistant, the link takes users to a guide for creating a "Storyteller AI Large Language Model (LLM)."
The GitHub page states, "Hand in hand, you’ll be able [to] create, refine, and illustrate little stories with the AI," indicating that the class intends to teach students how to build everything from the ground up—creating a functioning web app similar to ChatGPT using Python, C, and CUDA, with few prior computer science requirements.
Regardless of which course Eureka Labs launches first, neither appears to be finalized. A note on the GitHub page indicates that course development is ongoing, with no clear timeline provided.
It remains uncertain whether Karpathy is self-funding Eureka Labs or has secured investor backing, and details about the startup’s business model are also unclear. There are no public records of any investments associated with Eureka Labs, and the LLC filing in California has Karpathy as the sole signatory, with no word on possible collaborations with other prominent AI leaders.
Karpathy expressed on X that Eureka Labs represents the culmination of his two-decade passion for AI and education. He previously taught deep learning for computer vision at Stanford University until 2015 when he co-founded OpenAI. His journey continued at Tesla, where he led the AI team responsible for the company’s Autopilot system—an advanced driver assistance technology relying on cameras to process environmental data for tasks like cruise control and automatic steering.
After his departure from Tesla in 2022, Karpathy returned to OpenAI, leading a small team focused on ChatGPT until stepping down in February. Throughout these transitions, he has emphasized that there was no conflict or issues prompting his departures.
Throughout his career at Tesla and OpenAI, Karpathy has remained dedicated to education. He currently leads an online course called Neural Networks: Zero to Hero, teaching students how to build neural networks from scratch using code. Additionally, Karpathy hosts a YouTube channel where he shares lectures on LLMs and AI.
Ranking
