Sara Hooker, previously the VP of AI Research at Cohere, has launched a new startup called Adaption Labs, which aims to develop AI systems capable of continuous adaptation and learning from real-world experiences. This move comes as many in the AI community express concerns that the current trend of scaling large language models (LLMs) may be reaching its limits, with diminishing returns on performance. Hooker believes that the focus on scaling has led to inefficient methods that fail to produce truly intelligent systems. Adaption Labs seeks to challenge this notion by proving that AI can learn from its environment more efficiently. The startup is in the process of raising funds and aims to broaden access to AI research globally, continuing Hooker's commitment to diversity in the field. If successful, this approach could significantly alter the landscape of AI development, moving away from the costly scaling of models towards more adaptive and efficient learning methods.