‘How to bring human reasoning to AI’

“AI is stubborn and perverse.  We are all trying to bend Generative AI to our will. But we end up battling with issues of accuracy, explainability and trust. But now there’s a breakthrough. RAR is a real gamechanger, and I will tell you why.”

Ben Taylor is co-founder and CTO of Rainbird.AI, and a respected member of the All Party Parliamentary Group on AI. He was the expert speaker at our July ‘Innovation Talk’.

“LLMs are great at structuring chaotic data… pretty good at summarising… but truly terrible at reasoning. There is no logical explanation for outputs. No contextual evaluation of problems. And absolutely no logical chain of reasoning.”

“People are attempting to build architectures – such as Retrieval Augmented Thought (RAT), Semantic Rails and Prompt Chaining – to diminish these issues. But they can’t eliminate them.”

“But RAR can. Retrieval Augmented Reasoning uses a symbolic reasoning engine  and a knowledge graph to navigate document sources. It is sector agnostic – capable of creating digital assistants in any domain. And it can incorporate human expertise simply by adding knowledge to the graph – balancing it with weights and certainties. RAR is a total AI gamechanger. And, here I explain why.”