SemiKong is a new LLM trained by Aitomatic and its “AI Alliance” partners and is the world’s first large-scale language model created specifically to serve the needs of the semiconductor industry. SemiKong aims to become part of the workflow of semiconductor design companies, acting as a digital expert in the field and helping bring new chips to the world significantly faster.
According to Aitomatic, the company responsible for developing SemiKong LLM, the semiconductor industry desperately needs to gather information from experts. Several companies are suffering from large expertise gaps as many aging professionals retire and take their knowledge with them. A dedicated LLM trained to industry needs seemed like a surefire way to give new engineers the information they need to stay competitive.
Based on Meta’s Llama 3.1 LLM platform, SemiKong recently released its 70B variant. Aitomatic and other partners in the new AI Alliance (including Meta, AMD, and IBM) collaborated on LLM, and Aitomatic’s DXA system became the backbone of the SemiKong deployment.
DXA (Domain-Expert Agent) is Aitomatic’s way of connecting small LLM agents with a central collection of SemiKong 70Bs. DXA can be tailored to your company’s needs with training on entry from your company’s technology library or expert engineers. Once trained, DXA is utilized in the core SemiKong deployment environment to automate development tasks and provide chatbot-style communication with engineers and employees.
LLM with the current 70B format and smaller SemiKong-based DXA agents far exceeds the usefulness of general purpose AI models in the semiconductor space. SemiKong touts a 20-30% reduction in time-to-market for new chip designs and a 20% improvement in first-time good manufacturing scores. It also claims to reduce the learning curve for new engineers by up to 50%, which is the main claim supported by Meta.
If you are interested in implementing SemiKong’s 70B model, you can download it from their website. SemiKong is one of the first projects to emerge from collaboration within the nascent AI Alliance, announced in December 2023. One of many new corporate alliances that appear to be built to counter Nvidia’s dominance in the technology industry, the AI Alliance includes major companies such as IBM and AMD. Research institutions such as Yale University and the University of Tokyo.
