Latest UK-focused news and updates.
San Francisco-based startup Goodfire has unveiled a new tool named Silico, designed to enhance the training of AI models, particularly large language models (LLMs). This innovative tool allows researchers and engineers to gain deeper insights into the inner workings of AI models by enabling them to adjust parameters that dictate model behaviour during the training process.
Silico aims to provide developers with a more granular level of control over AI model construction, potentially transforming how AI systems are engineered. By facilitating a clearer understanding of model mechanics, Goodfire hopes to make the process of training AI more akin to traditional software engineering practices.
The introduction of Silico comes at a time when the demand for interpretable AI technologies is rising, especially in the UK, where regulatory scrutiny and ethical considerations surrounding AI are increasingly prominent. The ability to debug and fine-tune AI models could help address concerns about transparency and accountability in AI applications.
As the UK continues to invest in AI research and development, tools like Silico may play a crucial role in shaping the future of AI technologies, ensuring they are built responsibly and effectively.
Source: www.technologyreview.com – https://www.technologyreview.com/2026/04/30/1136721/this-startups-new-mechanistic-interpretability-tool-lets-you-debug-llms/