OpenAI’s tool could influence users’ expectations for how they interact with AI solutions and platforms.
To learn more about ChatGPT and its potential industry impact, watch this interview with Amelia CEO Chetan Dube.
More than 1 million people have already experienced the amazing capabilities of OpenAI’s ChatGPT chatbot, utilizing it to answer questions, converse about subjects at length or create stories, poetry, computer code etc., with its generative capabilities. The technology has so perfectly captured the current industry zeitgeist that OpenAI is already working on the next generation. Some predict that ChatGPT will change the way people seek information, answers and insights, which currently is mostly done by “Googling” the vast internet.
The tool’s sudden rise in popularity naturally has tech execs and industry observers theorizing about its impact on current chatbots and more sophisticated Conversational AI solutions (like ours). Plenty has been written about ChatGPT’s overall cost, its “point-in-time” challenges, its ability to work around safeguards for offensive language, and its tendency to sometimes present inaccurate information with unfounded confidence. Most of these are growing pains that will be dealt with over time — and most likely, it will not take long at all. In my opinion, we are witnessing a clear disruption that showcases how natural language understanding, which feeds Conversational AI and virtual agents, is going to be an indispensable part of our near and long-term future.
Our focus at Amelia is enterprise-ready Conversational AI use cases, purpose-built with natural language models and contextual understanding to help customers automate manual and repetitive tasks in customer and employee support. Amelia is regularly cited as the market’s best Conversational AI virtual agent, so how will this tool’s features potentially enhance what our customers can do?
To be clear, Amelia already utilizes generative capabilities of existing transformer models, such as ChatGPT, in our design and use-case creation process, which speeds up development and time to value for our customers. Over time, ChatGPT could likely improve that process even further.
However, from an enterprise perspective, it is critical to understand several things about these new large-scale language models.
- Deterministic vs. Probabilistic: This is perhaps the most important ChatGPT-related issue for large enterprises, which require deterministic (specific) as opposed to probabilistic (potential but not certain) outcomes, i.e., you can’t be right “most of the time” when you’re a bank utilizing virtual agents to transfer funds between accounts or execute a stock trade.
- Answers vs. Resolutions: Most interactions that enterprises have with their customers and employees are not predominantly simple questions and answers, but specific requests that require resolutions through integrations into back-end systems. This is currently not possible with non-deterministic models like those used by ChatGPT.
- Goal-Oriented Dialogue vs. Open-Context Dialogue: When people refer to the conversational capabilities of ChatGPT, they refer to its ability to maintain context, i.e., you get a great answer, but want to ask follow-up questions. This is vastly different from the way enterprise virtual agents are designed — they are goal-oriented in their conversations. Virtual agents ask AI-generated contextual, clarifying, or elaborative questions to get at the heart of what a customer really wants to achieve. ChatGPT currently does not have these abilities.
- Black Box vs. Explainability: This is a variant of the issue with probabilistic outcomes. With the vast amount of data used to train ChatGPT models, gaining insight into why the tool answered a certain way is not possible. As some users have already experienced, a slight tweak to a sentence sometimes provides quite different results from a previous interaction, no matter the similarities. Without this level of explainability, it is nearly impossible to tweak the models and improve the overall efficacy.
These issues aside, ChatGPT is poised to fundamentally change how consumers interact with the Internet, create new content, and explore subjects of interest. It will also significantly raise the bar for how users expect to interact with businesses (via virtual agents) in a more automated and conversational tone. At Amelia, we are thrilled to watch ChatGPT generate this much excitement around AI technologies, and the various tasks they can perform in people’s work and personal lives. It blends perfectly with our mission to utilize automation and Conversational AI so humans can be freed from mundane jobs, and enterprises can create extraordinary customer experiences.