From Semafor: 2024-04-12 14:50:00

Major breakthroughs in AI have eliminated the need for human labeling of data, with deep learning allowing software to teach itself tasks. Large language models like ChatGPT stem from the transformer architecture, enabling neural networks to grow in size and complexity. Companies like Microsoft are investing heavily in AI infrastructure in pursuit of even greater advancements. However, limitations in existing models have led to the need for new approaches, such as utilizing category theory to instill constraints into AI architectures. Symbolica is pioneering this method, aiming to create more reliable and interpretable AI models tailored to specific tasks. This customized approach contrasts with the broad capabilities of generalist models like GPT-4, offering improved performance for specific applications.



Read more at Semafor: Early OpenAI investor now bets on Sam Altman’s firm being wrong