From Pridictive Analytics to Prescriptive AI in O-RAN RIC * Predictive Analytics vs. Prescriptive Analytics Prescriptive analytics goes beyond predictive analytics by not only forecasting future outcomes but also providing specific actions to achieve the desired outcome. Predictive analytics forecasts potential future outcomes based on historical data, while prescriptive analytics helps in decision-making by providing actionable insights and recommendations. Both types of analytics can be used together to yield more granular and actionable insights. Predictive analytics involves interpreting trends, while prescriptive analytics uses heuristics-based automation and provides specific recommendations. * Predictive Analytics vs. Predictive AI Predictive analytics and predictive AI, while closely related in their goals, differ significantly in their methodologies, complexity, and applications. The key difference lies in the capability of predictive AI to evolve with new data. While predictive analytics offers valuable insights based on existing data, predictive AI takes it a step further by continuously learning, refining its models and adapting to new patterns as they emerge. This makes predictive AI more dynamic and responsive to changing conditions, thereby offering more accurate and timely predictions in complex situations. * Prescriptive Analytics to Prescriptive AI Initially, prescriptive analytics involves using mathematical models and algorithms to suggest the best course of action for a given set of objectives and constraints. This process is heavily reliant on structured data and predefined rules. Typically, the models used in prescriptive analytics are static, meaning they do not learn or evolve from new data without manual intervention. The scope here is generally limited to specific, well-defined problems, and the insights provided are based on existing data without real-time learning. As we move towards prescriptive AI, the system starts incorporating advanced AI and machine learning such as Deep Reinforcement Learning and recently, LLM Agents technologies. These technologies allow the system to learn from new data continuously, adapt its recommendations/actions, and even autonomously improve decision-making processes over time. This not only includes adapting to new scenarios but also predicting the consequences of different actions in dynamic environments. Prescriptive AI in O-RAN RIC embodies a shift from a reactive, network optimization approach to a proactive, learning-based strategy. These AI models can make more nuanced prescriptive decisions. For instance, the prescriptive AI-enabled RIC could dynamically adjust network parameters in real-time by predicting and responding in advance to sudden changes in user demand or network conditions, such as during a large public event or network faults. #ORAN #OpenRAN #PrescriptiveAnalytics #PrescriptiveAI #PredictiveAI #NetworkAnalytics #NetworkAI
Jinsung Choi - Agree generally, but don't discount domain models and deterministic reasoning. All the LLM providers recognize that their is a necessary bridge from brute force deep learning and symbolic AI. 2023 was the year of the LLM, but 2024 will be the year of the graph domain model to ground LLMs and other forms of AI/analytics with real-time operational knowledge for contextualization, safety and correctness.
For Prescriptive AI Three important concepts are gaining traction: Explainable AI (XAI), Interoperable Machine Learning (IML), and Prescriptive AI. While each holds individual value, they also intertwine in fascinating ways, ultimately contributing to the realization of truly **actionable AI**, where insights lead to concrete, effective actions. Explainable AI (XAI): * Sheds light on how AI models arrive at their predictions. * Builds trust and confidence in AI systems, especially in critical domains. * Allows for debugging and identifying biases or errors in models. Interoperable Machine Learning (IML): * Enables seamless communication and collaboration between different AI models and systems. * Promotes shared learning and knowledge transfer across diverse datasets and tasks. * Facilitates the creation of more robust and adaptable AI solutions. Prescriptive AI: * Goes beyond prediction, offering actionable recommendations and suggestions. * Leverages insights from multiple models and data sources to optimize decision-making. * Empowers humans to take informed actions based on AI-driven guidance.
Founder at Cirrus360 Corp
1 年Alex, nice summary. Prescriptive will sit on top of Predictive which will sit on top of Descriptive. So they are not competitors but partners, right? My point being that we need to do the each of the lower layers very well before moving up to the next one.