Let’s talk about the future of AI – it is not just about the intelligence provided, but the ability act and interact with this information in a transformative way.
There are 4 main components to this trend in AI:
AI - the first and most foundational is the massive progress we’re seeing on the AI front in the form of Generative AI as well as the ability to scale custom models. Generative AI is great at language interpretation and images; however, organizations need specific models for their analytical and predictive needs. There has been a lot of progress in the past few years on AutoML, the ability to have machines analyze and pick the best model and hyperparameter tuning for your scenario. There is also a concept of many-models that in addition to selecting the best model type, allows you to build a lot of models at scale so that you can have tailored models for your scenario.
Discovery – we’re looking for relationships that we didn’t know existed, this is commonly referred to as Data Mining. By sitting on top of a Data Lake, you will have a vast amount of data that you have access to, making sense of this, understanding correlations within your data – this is referred to as descriptive analytics. It is also about building prediction models and looking to the future, using past information to inform future actions…
Action – discovering a concept, relationship, or building out a predictive model is a base requirement, but inefficient to enact change – you must do something. I commonly refer to analytics without action as ‘navel gazing’, just analyzing your data, but not doing anything with it. This is where we need to take relationships and predictions and move them towards prescriptive actions and recommendations. Building in the ‘next best action’ as a result of your analysis and predictions is a critical step to improvement.
Network – as we build out our solution, we are seeing a large benefit with regards to a larger collective data set that many organizations can leverage. The technologies for these solutions exist, however building out the predictive models requires a vast set of data that a network can share. We have examples of smaller health systems building robust models when they’re able to leverage a larger pool of data and train their data against that larger data set. We are also seeing a similar property with the use of AI Agents. Agents are specific role-based AI solutions that perform one function very well, when paired with other agents in a network, the result is much better than trying to build one big model. We’re seeing the growth of a network of reuseable AI Agents in our solution set.
And that’s how we came up with AIDAN – Artificial Intelligence Discovery and Action Network