We're a network of independent machine learning researchers who work on projects together in small teams. Team members are typically self-taught technical individuals who are interested in building and delivering new capabilities and practical products. Most, if not all, of our members are deathly allergic to academia.
New ML technologies have the potential to change how we educate ourselves. How do automations and technologies-of-convenience improve or hinder the learning process? Can we take advantage of "hallucinations" of generative ML models?
The development of language models has created a large pool of techniques and strategies that can be applied to other types of sequential data. Can we create new language systems that work with language models in new domains?
For many problem domains, the low availability and high costs of human experts are major barriers to the accumulation of high quality data. How can we build synthetic data generators to amplify these experts?