Adept, which Luan co-founded in 2022, is focused on building agents that can perform tasks at higher and higher levels of abstraction while maintaining a high reliability standard. The company has developed a series of models and tools, including ACT-1, a large Transformers model optimized for browser interactions, and Fuyu-8B, a multimodal model designed for digital agents. Luan believes that the focus on agents and customer feedback will allow Adept to build AGI faster and more effectively than pure play foundation model companies.
Key takeaways:
- David Luan, CEO and co-founder of Adept, believes that agents are the correct long-term direction for AI, with a definition of AGI oriented around a model that can do anything a human can do on a computer.
- Adept is focused on building agents that can do a higher level of abstraction tasks over time, while maintaining a high reliability standard. This is achieved by learning from users how to get to the next level of abstraction faster.
- Luan argues that the future of AI will be driven by the deep co-design and co-evolution of product and users for feedback and actual technology. He believes that labs equipped with the tools to do this will excel.
- Despite the success of large research labs, Adept is focused on commercial agents. Luan believes that the next couple of Meta Llama models will commoditize regular LLMs and regular multimodal models, making it crucial to focus on agents.