Lloyds Banking Group has partnered with British AI start-up UnlikelyAI to test a new artificial intelligence platform aimed at improving customer experience and innovation with explainable and regulation-compliant outputs.
The trial, which is taking place in the bank’s Innovation Sandbox, will evaluate how UnlikelyAI’s technology, developed using a so-called “neurosymbolic” approach, can be used across the Group to support customers and staff with more reliable AI assistance.
Founded by William Tunstall-Pedoe, one of the creators behind Amazon’s Alexa, UnlikelyAI aims to tackle the issue of large language models’ tendency to produce convincing but incorrect responses known as “hallucinations” with the neurosymbolic approach.
By combining neural networks that learn from data with symbolic reasoning based on rules and logic, the company aims to provide more trustworthy and transparent AI systems.
“This new partnership with one of the biggest, oldest, and most trusted British banks is testament to the demand for novel AI solutions in regulated industries,” said Tunstall-Pedoe, CEO of UnlikelyAI.
“We built UnlikelyAI to solve the fundamental problems that have prevented LLMs from being deployed safely in customer-facing financial services.”
Lloyds is exploring the use of the platform as a complement to its existing AI technologies, with the goal of enabling faster, smarter and more reliable services, while meeting the Group’s standards on accuracy and compliance.
“We’re harnessing the power of AI to revolutionise banking,” said Ranil Boteju, chief data and analytics officer at Lloyds Banking Group. “UnlikelyAI’s neurosymbolic approach will allow us to test innovative new features for colleagues and customers that are not only useful, fast and intelligent, but also transparent and reliable.”
The proof-of-concept trial may eventually inform wider deployment of neurosymbolic AI across the Group’s operations, though no timeline has been disclosed.
