🛑Do not rely solely on data-driven models to build intelligence, especially in decision-making scenarios. They are unreliable intelligence as they are not based on domain knowledge and understanding of the semantics of business operations.
There was a time when building something executable on any computer required some form of coding (programming). Excel (spreadsheets) came along and allowed interactive question-and-answer with computers without programming, at least at a basic level. Nowadays, using #data and building #machinelearning models is easy and enables organizations to quickly build systems without any coding.
💢In a nutshell, you provide data, the computer learns, and then you ask questions; it answers. These #machinelearning models can even do more. They can *potentially* predict the future based on past behaviors. They find relationships between numeric values, identify patterns, and extrapolate. They do not work with semantics; their underlying methods and technologies are not based on business and operation domain knowledge. Therefore, it does NOT produce reliable, accurate, and safe intelligence, no matter how much we want to twist the terms.
However, we will still use #machinelearning in Decision Intelligence because #machinelearning models as part of #decisionintelligence are different.
🤔How so?
💡In the “Cognitive Intelligence” layer, you will use Composite AI techniques and integrate sets of #machinelearning algorithms leveraging Neuro-Symbolic AI. It means a business decision uses multiple techniques such as #businessrules #optimization #calculation #math domain-driven procedures as well as #machinealearning coordinated by a hierarchical and multistep Decision Graph that is situation-aware. Now, you can accurately and safely solve complex decision-making problems for semantic analysis, product recommendation, price prediction, customer segmentation, fraud detection, sale forecasting, etc.
✨Additionally, many believe every intelligence system should have a learning cycle and feedback loop that automatically, semi-automatically, or manually allows the system to learn from its outcome. That is where the “Autonomous Intelligence” using continuous decisioning loop and continuous learning comes into the picture — allowing the decisions to learn directly from outcomes of the decisions *based on their business impacts* with or without the involvement of humans and domain experts.
🚀Read on Decision Intelligence Architecture at this post.
Posted here.
Published June 5th, 2024 at 07:30 am
