I’ve uploaded the slides for the keynote I presented at the 2nd AIOTI workshop on semantic interoperability for digital twins. The main premise is the potential for applying neural networks in combination with conventional information technology to business workflows. Neural networks are used for semantic orchestration and reasoning over unstructured information, whilst conventional IT is used for efficient deterministic processing of structured information. Human oversight is an essential part of the architecture, both for designing and reviewing the operation of AI agents, and also for high risk or high ambiguity cases where human judgement sets precedents that can then be exploited by the agents. That is essential as most organisations rely on tacit knowledge of experienced workers, i.e. knowledge that isn’t written down. AI has the potential for organisations to improve their flexibility and reduce their costs. Rather than being used to layoff staff, AI should be seen as an opportunity to nurture and train the staff that are essential to business operations.
https://github.com/w3c/cogai/blob/master/papers/Raggett-Bridging-the-gap.pdf
Dave Raggett <dsr@w3.org>
