AI startup Adaptive has recently secured $20 million in an initial investment round, with major contributions from influential entities such as Index Ventures and ICONIQ Capital. This funding will be directed towards the advancement of Adaptive’s trailblazing technology that customizes AI according to diverse business requirements.
The innovation at the core of Adaptive’s operations is reinforcement learning from human feedback (RLHF), a tool that can enhance large language models (LLMs), primarily used in chatbots. However, the adoption of this method often bumps into significant roadblocks, such as high associated costs and potential incompatibility with commercial LLM applications. Even so, the transformative potential RLHF holds for LLMs cannot be overlooked.
Adaptive is stepping up to tackle these challenges by creating models that can glean valuable insights through employees’ or customers’ interactions with the software. According to CEO Julien Launay, user interaction data proves to be more educational than evaluator feedback, as it provides clearer glimpses into actual user experiences and potential issues often missed by internal evaluators. By converting raw data into valuable insights, Adaptive hopes to fine-tune user experiences, thus allowing the software to cater more directly to concrete needs and habits.
Future plans for the company include developing solutions that track user interactions with LLM feedback, thereby improving training and adjustment processes. They also aim to construct a platform that utilizes reinforcement learning algorithms for model modification. Alongside this, they plan to develop cutting-edge AI designs that will augment machine interactions, enabling intelligent decision-making.
Furthermore, Adaptive promotes reinforcement learning from AI feedback (RLAIF), a strategy in which an AI model critiques another AI model during training sessions, offering potential reductions in costs and improvements in data quality. All this suggests Adaptive’s proficiency and adaptability in implementing advanced learning techniques.
Despite the fierceness of competition in the market, Adaptive still stands out. It offers a unique platform that collates preference data from users, thoroughly evaluating different LLMs’ performance. However, it does not support businesses in adjusting proprietary models developed by third parties due to various access barriers. Ultimately, Adaptive’s unique proposition in the market allows it to cement a distinctive position amidst competitors.
