Artificial Intelligence is one of the world’s fastest-growing fields in science and technology. Some of the earliest recorded theoretical work on AI dates back to the mid-20^th Century, by British pioneer Alan Turing and who proposed considering the question, “Can machines think?” 

The development of AI as a discipline answers this question, as machines are trained using large data sets to perform cognitive tasks typically associated with human behaviour, including decision-making, visual perception and speech recognition. AI can be classified into ‘general’ and ‘narrow’ AI, taking into account the difference between mimicking or replicating human intelligence (general) and performing specific tasks intelligently (narrow). 

What about Machine Learning (ML)?

Although often used interchangeably, ML is a subset of AI and is the process of extracting insights and learning from datasets. For ML to be accurate, datasets need to be correctly constructed, transformed into the appropriate structure and consisting of good quality, representative data of the prediction problem they are applied to. In a real-world context, both AI and ML are being used for predictive tasks from fraud detection through to medical analytics. In a more specific context relating to data quality, these techniques can also be used to improve the quality of data when applied to tasks such as data accuracy, consistency, and completeness of data along with overarching data management processes themselves.

How is Datactics using ML?

One real-world use case for ML can be seen in Datactics’ Entity Resolution (ER). ER is a central part of the KYC/AML process for financial services, producing a reliable golden record of a client or entity that an institution is onboarding and/or maintaining. This is important for tasks such as risk scoring through to regulatory compliance, and is something which AI/ML can assist with by improving consistency and reducing time around manual processes.

AI offers significant benefits for Datactics clients, who are partnering with us to drive business value through explainable AI (XAI) use cases.

So what is XAI?

XAI refers to a partially or completely supervised application of AI techniques. In XAI models, every aspect of prediction, automation and modelling of AI is fully explainable; put simply, users are able to explain why a model has behaved in a specific manner. This offers many benefits over so-called ‘black-box’ implementations of AI, where it’s unclear how the AI has reached a decision, or whether it is expected and consistent with either the data or planned outcome.