ComputeNovelty

ComputeNovelty: Advancing Open-Ended AI Systems used in Product and Service Design: through Responsible Innovation and User-Centered AI Agents

Introduction and Objectives of the Innovative Research

Led by CarefulAI this research aims to develop innovative open-ended AI systems that embody the principles of responsible AI, safety, and user-centered design.

It builds on Prof. Joseph Connor's work in AI enabled product development and UCL Computer Science's work on open-endedness

The research aims to ensure that open ended AI systems outputs are not only innovative but also ethical, robust, and accessible.

​This will be tested in collaboration with leaders in the product centred design domain: UsTwo and the Design Council, sources of unregistered AI IP e.g arXiv and registered IP, and product & service providers that aim to bring AI to market in regulated sectors.

Phase 1: Research and Framework Development

- Research how product and process designers can use open-ended AI systems to more swiftly and effectively embody user needs embodied in an AI agent (i.e. UXAI) and expectations of AI generated use cases that are screened to be innovative by an AI Agent (i.e. NovelAlgo)
- Develop a technical framework for ComputeNovelty to integrate open-ended AI system use, NovelAlgo, PRIDAR and UXAI outputs to deliver business value

Phase 2: Design and Prototyping

In an open-ended system:
- Design user-centered interfaces for ComputeNovelty, NovelAlgo, PRIDAR and UXAI. For example chatbot interfaces so that users can test their value and safety.
- Implement risk assessment and assurance modules to ensure the whole ComputeNovelty system and its elements can be deployed safely.

Phase 3: Testing, Iteration, and Deployment

- Conduct end user user testing and gather feedback to refine developments.
- Deploy the proof-of-concepts across various end user sectors e.g. Healthcare, Finance and government
​
Target Impact and Benefits

Open Ended AI design systems that promote safe, responsible innovation.

Expected Outputs

UXAI: AI agent that will ensure safe AI (from PRIDAR) is designed to meet the user needs of users and is fit for purpose and have a Sona type interface.

NovelAlgo: an AI agent that will take the artifacts from a generative agent trained in open ended design thinking and confirm they are novel prior to them being tested.

ComputeNovelty: a mixture of algorithms that ensure that any open-ended methods e.g. those developed by UCL, or Google are seen to be novel to investors, safe, and meet the needs of users.