Nomi's companion chatbots can now recall details such as which colleagues you don't get along with.

OpenAI's latest o1 model emphasises increased thoughtfulness, but a small, self-funded startup called Nomi AI is building similar technology with a different focus. Unlike ChatGPT, which takes time to think through various types of queries, Nomi narrows its specialisation to AI companions. Nomi's chatbots are designed to respond more thoughtfully to user interactions, retaining memories from past conversations and offering more nuanced emotional responses.

“Our approach follows the same principles as OpenAI, but is tailored to our users’ needs, emphasising memory and emotional intelligence,” Nomi AI CEO Alex Cardinell told TechCrunch. “While OpenAI focuses on the chain of thought, we focus on the chain of introspection and memory.”

OpenAI’s o1 model breaks down complex requests into smaller steps, such as solving a maths problem step-by-step to ensure accuracy. This reduces the chance of generating inaccurate information. On the other hand, Nomi’s LLM, designed specifically for companionship, works a bit differently. For example, if a user shares that they had a bad day at work, Nomi might recall past difficulties with a colleague and ask if that’s the cause of their stress. It can then remind the user of strategies they've successfully used to resolve similar conflicts.

“Nomis remember everything, but a crucial aspect of AI is deciding which memories should be used,” Cardinell added.