Insight of the Week: ChatGPT in the Contact Center?

by Kerry Robinson

ChatGPT, the research chatbot released for public use by OpenAI has been causing huge waves of interest and amazement in the power of conversational AI over the last couple of weeks.?

In last week’s email I interviewed ChatGPT about Conversational AI. This week, I’m interviewing Waterfield Tech’s VP of AI Technology, Yves Normandin, about ChatGPT.

?I started by asking Yves, what is ChatGPT?

ChatGPT is a large language model (LLM) from OpenAI. Several companies, including Google, Meta, Cohere, Stability AI and others are constantly competing with one another at creating ever more complex and powerful LLMs. In short, an LLM is a deep learning model, trained on large quantities of text, that simply predicts the next word in a text given the previous words. Using the model to generate one word after another ends up producing sentences, paragraphs and even complete texts that are surprisingly well constructed and relevant. And, frankly (and even scarily), nobody fully understands how and why this works.

How big is big?

A model like ChatGPT has over 170 billion parameters and was trained on over 10 GB of text!

Wow, so, it knows anything about everything then?

Unfortunately not! The model is strictly based on its trained parameters and the previous words in the text or conversation. It doesn’t – and can’t – access any outside information or database to make its predictions. So in a sense, it’s a giant black box that only knows about the data it was trained on. So ChatGPT knows nothing about anything that happened after 2021.

You mentioned other companies are working on large language models. What makes ChatGPT so special?

What makes it look revolutionary – and why everybody is talking about it now – is that it was optimized for conversations. Contrary to previous LLMs, ChatGPT remembers information exchanged in previous interactions so you can ask follow-up questions, challenge incorrect answers, etc. The result is absolutely mind-blowing! Another reason why everybody is talking about it is that OpenAI has made it open for anybody to try and people have been sharing the absolutely incredible conversations they had with it. You should try it:?https://chat.openai.com/, it’s a lot of fun!

So can we ditch everything we’ve built and just connect IVR and chatbot technology to ChatGPT?

Not yet.?As amazing as this technology is, it’s not quite clear yet how it can effectively be used to help create effective Virtual Agents. The technology still has important limitations, for example:

• The model only knows the data it was trained on, so it cannot easily leverage external information like customer data.
• We have little control over what the model generates. It will often produce plausible-sounding but incorrect or nonsensical answers. This is often called “hallucinations”. For customer service, that’s just not acceptable.

Thanks Yves!

So as interesting, and fun, as ChatGPT is to play with, it’s clearly not a solution for customer services. Yet.

In fact, the CEO of OpenAI said as much in a tweet this week:

Yves and I will, of course, be following the developments with ChatGPT and other AI research closely so we can help you deploy better virtual agents and make the most of conversational AI technology in your business.

Kerry Robinson?is an Oxford physicist with a Master's in Artificial Intelligence. Kerry is a technologist, scientist, and lover of data with over 20 years of experience in conversational AI. He combines business, customer experience, and technical expertise to deliver IVR, voice, and chatbot strategy and keep Waterfield Tech buzzing.

Subscribe to Kerry's Weekly AI Insights