Symbolica AI, founded by ex-Tesla engineer George Morgan, is taking a different approach to AI research, focusing on structured models rather than the traditional deep learning and generative language models that dominate the field. Morgan’s experience at Tesla, where he worked on Autopilot, made him realize the limitations of current AI methods, which rely heavily on scaling up compute power.
Traditional AI methods, according to Morgan, have limitations in terms of scalability, cost, and efficiency. They require vast amounts of data, compute resources, and time to train, with diminishing returns in performance as scale increases. Symbolica AI aims to address these challenges by developing novel models that achieve better accuracy with lower data requirements, training time, and cost.
Morgan’s perspective aligns with recent discussions in the AI research community about the need for fundamental breakthroughs to advance the field beyond current approaches. Reports from TSMC executives and independent research institutions highlight the growing challenges of scaling AI models and the increasing costs associated with training them.
Symbolica AI’s focus on structured models, which encode the underlying structure of data, offers a promising alternative to conventional deep learning methods. By capturing the inherent relationships within data, structured models can achieve superior performance with less computational overhead, potentially revolutionizing the way AI algorithms are developed and deployed.
As the AI industry faces escalating costs and resource constraints, innovative approaches like Symbolica AI’s structured models could pave the way for more sustainable and efficient AI systems in the future.