Nyckel will present at the Santa Cruz New Tech on April 5, 2023. Get your tickets!

The New Internet

Machine Learning (“ML”) has evolved greatly over the last 10 years, and today almost any business can benefit from ML in one way or another. The problem is that ML is Still Too Hard for Software Developers. Complexities range from grasping core ML concepts; staying abreast with the literature; bringing up software for annotation, data inspection & training; setting up infrastructure; etc. There are many ML players trying to tackle this complexity. OpenAI, perhaps the most known player, is betting on huge language models (e.g. GPT-4). These are trained to perform almost any task and come equipped with a very general interface.

With a new approach, Nyckel provides "narrow" AI models that are trained for very specific tasks. Contrary to GPT-4 , the Nyckel functions are highly schematized, cheap & light and allow developers to spin up as many as they need for the application. This approach resonates with a lot of developers allowing them to decompose their ML needs into simple digestible pieces.