
I thought I’d take a moment to share how Ami, our conversational AI platform, uses GPT (Generative Pre-trained Transformer) models that have recently been featured in the press.
GPT is a language model which conversationally generates informative, human-like, written content for us to read.
Ami works with content from multiple sources including websites, API’s, live data, and statistical models such as GPT. Ami’s understanding of customer objectives facilitates the ability to predict the most effective and relevant advice, whilst Ami’s AI determines when to use such information.
GPT generated content is already being deployed in Ami’s AI platform. The generated content helps Ami articulate answers from other content sources by providing rewritten material to add variation to Ami’s language. Ami’s AI layer then applies machine learning techniques to find the most effective and customer friendly advice.
There are issues of bias and control surrounding the use of generative models. However, Ami makes decisions which are driven by prior success in fulfilling a customer’s objective. Therefore, when a client is “mature” Ami is confident that the answers and outcomes given to a customer are accurate and effective. Ami can apply the same confidence logic before handling the Generative AI content which consequently mitigates the risk of bias and control.