Ascertain’s Post

What we think of as modern AI is really a set of different tools all based, effectively, on the concept of artificial neural networks. Over the last decade there’s been a progressive unveiling of different formulations of neural networks: CNNs, RL, transformers and LLMs... These formulations have also progressively unlocked new capabilities—for example, CNNs opened the world of AI-based image processing up to us in the radiology community around 10 years ago and has been transforming the field ever since, well before LLMs and gen AI hit the scene. In short: every once in a while, AI evolves and gets new abilities. While it's hard to predict what's coming next, state space models are shaping out as a promising next step in the evolution of this technology. They can enable lower cost, more accurate AI agents--we explore this further in our new blog post on Towards AI. We’re exploring state space models at Ascertain because we think it is a great fit for agent-led automation tasks in healthcare and beyond. Big thanks to Vineet Tiruvadi, MD, PhD and Marek S. for the incredible work in describing the fundamentals of this tech in our new blog post.