Show HN: TOD,a universal Decision model based on Task Oriented Design (parseclab.ai)

A new AI model named TOD (Task Oriented Design) has been introduced, aiming to revolutionize decision-making processes in AI applications by integrating both text and visual data. Built on a 12-billion-parameter foundation, TOD is specifically designed to read messages and images holistically, facilitating accurate decision-making in customer support services. It processes up to 48,000 tokens of context and utilizes a dual-step mechanism to rank and select the most appropriate actions without generating text, ensuring users only pay for the input used rather than output generated. Significantly, TOD enhances decision accuracy while decreasing operational costs, offering a flat rate of $0.30 per 1,000 decisions, which is 25% lower than previous models like JEV. The model has demonstrated competitive performance, only slightly lower than an untrained baseline in accuracy, and it effectively integrates image understanding into conversational contexts. With plans for future updates, including the ability to recover more complex decision-making scenarios, TOD represents a substantial advancement in the ability of AI systems to navigate multifaceted inputs efficiently and reliably.