The development of AI agents has been a topic of significant interest in recent years, with many experts predicting a rapid advancement in the field. However, according to Andrej Karpathy, co-founder of OpenAI, we are still about 10 years away from achieving true AI agents. This move reflects broader industry trends, where the focus is shifting from narrow, specialized AI models to more general, autonomous agents.
Karpathy’s statement is significant, given his experience in leading Tesla’s self-driving efforts from 2017 to 2022. He notes that while large language models (LLMs) have made huge progress, there is still a lot of “grunt work” to be done before we can achieve artificial general intelligence (AGI). As Karpathy puts it, “I still feel there’s so much work to be done” - but he adds that “they’re going to get better, and it’s going to be wonderful.”
The challenges that lie ahead are numerous, including the need for multimodal learning, continual learning, and the ability to interact with the physical world. Karpathy believes that it will take about a decade to overcome these challenges, citing his own intuition and experience in the field. He also notes that the diffusion of such a system will still take even more time, due to technological, societal, and legal constraints.
Karpathy’s predictions are not just based on his technical expertise, but also on his understanding of the broader societal implications of AI. He notes that as AI becomes more autonomous, there will be a gradual loss of control and understanding of what’s happening. This is why he is working on Eureka Labs, an education company that aims to provide people with the skills and knowledge needed to work with AI.
One of the key projects at Eureka Labs is nanochat, a full-stack implementation of an LLM like ChatGPT in a single, clean, and minimal codebase. Karpathy hopes that this project will help people understand AI better and provide a “ramp to knowledge” for those who want to learn. As he says, “What I cannot create, I do not understand” - a quote from the Nobel-winning physicist Richard Feynman that reflects his approach to education and AI.
In conclusion, Karpathy’s insights provide a nuanced view of the current state of AI agents and the challenges that lie ahead. While there is significant progress being made, we are still far from achieving true AI agents. However, with the right approach to education and research, we can overcome these challenges and create a future where AI enhances human capabilities, rather than replacing them.
Source: https://thenewstack.io/openai-co-founder-ai-agents-are-still-10-years-away