AI Chat Platform CHAI Hits $1.4B Valuation
AI company CHAI announced it has maintained a 3x annual growth rate, reaching $68 million in annual recurring revenue (ARR). This growth has resulted in a new valuation of $1.4 billion for the company. The announcement also highlighted the company's focus on AI safety amidst its rapid expansion.
- CHAI’s recommendation system is built on a Mixture-of-Experts (MoE) architecture, where a gating model selects the most suitable specialized language model for a given conversation to optimize for user engagement. The company's primary metric for A/B testing these models is user engagement, which they measure by analyzing user screen time on the app. - The company has a strong focus on in-house model development, having moved from open-source models like GPT-J to deploying their own 6B and 13B parameter LLMs. They utilize techniques such as Reinforcement Learning with Human Feedback (RLHF), Proximal Policy Optimization (PPO), and Direct Preference Optimization (DPO) to fine-tune their models for user retention. - To support their large-scale operations, CHAI has built a significant amount of custom infrastructure, including their own GPU orchestration system and custom CUDA kernels for more efficient inference. This became necessary as off-the-shelf solutions could no longer support their user base of over 500,000 daily active users. - Founded in 2021 by William Beauchamp, the Palo Alto-based company was one of the first to launch a consumer AI chat platform. - The platform's growth has been substantial, reaching a $1.4 billion valuation with $68 million in annual recurring revenue. - CHAI has published research papers on topics such as AI safety frameworks and rewarding chatbots for real-world engagement, indicating a focus on both user safety and model performance.
