This startup is helping companies train AI with an old but buzzy technique. Read the pitch deck it used to raise $7.5 million. (www.businessinsider.com)

AgileRL, a London-based startup, has successfully raised $7.5 million to enhance its AI training software platform, Arena, which leverages reinforcement learning (RL) techniques that date back to the 1950s. The platform allows engineers and data scientists to efficiently input their AI models, run simulations, fine-tune them, and monitor their performance, facilitating a streamlined training process. With a significant shift in corporate focus from RL to transformer-based models following the launch of ChatGPT, AgileRL's co-founder Param Kumar emphasizes the necessity of integrating RL to refine specific tasks that transformers alone cannot optimize. This funding round, led by Fusion Fund and supported by other notable investors, is poised to accelerate AgileRL's growth as it aims to establish an office in San Francisco and expand its team. The platform has seen over 300,000 downloads, attracting clients like Airbus, IBM, and JPMorgan, highlighting its potential to unify and expedite AI development processes. As the AI/ML community recognizes the limitations of transformer models, AgileRL’s approach offers a complementary strategy, reinvigorating interest in reinforcement learning for complex task optimization within AI systems.