This is a very important direction because now we have a lot of AI models that are specialized and they do one thing and they do it well, but we cannot have hundreds of models all in different software, different vendors. It would be a mess. So actually, I think we can move towards the agnostic AI models. And there are two ways. One is the agent-based models where we have all these specialized models that we develop in different platforms, but we have one agent like the central AI that handles integration and discussion between the specific models...
This is a very important direction because now we have a lot of AI models that are specialized and they do one thing and they do it well, but we cannot have hundreds of models all in different software, different vendors. It would be a mess. So actually, I think we can move towards the agnostic AI models. And there are two ways. One is the agent-based models where we have all these specialized models that we develop in different platforms, but we have one agent like the central AI that handles integration and discussion between the specific models. So it integrates information from specialized disease-specific models. Or as you mentioned, there are shared commonalities across diseases. Like we have the hallmarks of cancer that are now shared across many different cancer types. And I will mention in my talk, we actually developed pan-cancer prognostic models. And we saw that these models can actually learn and generalize across the cancers, at least to learn basic concepts. And then you can maybe refine those models for specific tasks that you are looking at. So I think we are moving toward generalized AI models for lung cancer that can be beneficial for everyone.
This transcript is AI-generated. While we strive for accuracy, please verify this copy with the video.
