Andrej Karpathy, formerly leading AI at Tesla and a researcher at OpenAI, is embarking on a new journey with Eureka Labs, an “AI native” education platform. This phrase often means the platform is built with AI integrated deeply from the start. However, Eureka Labs is initially approaching education in a more conventional manner.
Based in San Francisco, Eureka Labs was registered as an LLC in Delaware on June 21. The company aims to harness recent advancements in generative AI to develop AI teaching assistants that can help students navigate their coursework. According to Karpathy, who announced the news on X, the vision is for these AI assistants to collaborate with human teachers, making it possible for "anyone to learn anything." While teachers will still create the course material, the AI assistants will offer support. As of now, Eureka Labs has not yet tested the effectiveness of these AI assistants in classrooms. Notably, a study by Georgia State University found that AI teaching assistants did help some students improve their grades.
Karpathy suggests a future where these assistants might be modeled after real people, similar to Meta’s celebrity chatbots or Character AI’s character chatbots. However, the details on whether this will be a massive open online course (MOOC) with a chatbot or a more integrated product for high schools remain unclear, as the information available is limited. Karpathy did not respond to TechCrunch's request for more information.
Alongside his announcement, Karpathy shared an AI-generated image of a futuristic school featuring a spaceship-like building, numerous solar panels, and a girl with three hands. Despite these ambitious visuals, Eureka Labs’ first product will be an AI course named LLM101n, designed to help students train their own AI models. This introductory course will be like a smaller version of the AI teaching assistant that Eureka Labs aims to develop. The course materials will be accessible online, and the startup plans to host both digital and physical cohorts for collaborative learning.
Interestingly, the GitHub repository linked to this course hints at a different type of course—one focused on building a "Storyteller AI Large Language Model (LLM)." The course aims to teach students how to create, refine, and illustrate stories with AI, covering everything from basic concepts to a functioning web app similar to ChatGPT, using Python, C, and CUDA, with minimal prerequisites in computer science.
Currently, it's unclear when these courses will be ready, as the GitHub page notes that building the course will take time and does not provide a specific timeline. Additionally, there’s no information on whether Eureka Labs has received any investor backing or if it’s self-funded. The LLC filing with California’s Secretary of State is signed by Karpathy alone, and he has not mentioned any collaboration with other prominent AI figures.
Karpathy, who has a long-standing passion for AI and education, has taught deep learning for computer vision at Stanford University and has played pivotal roles at OpenAI and Tesla. His experience spans co-founding OpenAI, leading Tesla’s AI team for Autopilot, and contributing to ChatGPT’s development. Even after stepping down from his roles at Tesla and OpenAI, Karpathy continues to educate through his online course "Neural Networks: Zero to Hero" and his YouTube channel, where he regularly posts lectures on AI and large language models.
