ChatGPT: What's all the buzz about?
Since November of this year, there has been much buzz around a new dialogue language model called ChatGPT developed by OpenAI. ChatGPT is based on the GPT-3.5 family series (trained on text and code) and similar to InstructGPT (aligned with reinforcement learning through human feedback). In a nutshell, ChatGPT is a chatbot that can "answer follow-up questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests." ChatGPT has the ability to respond and converse with scary human-like responses.
?I had the opportunity to get early access to ChatGPT, so of course, I interacted with it. I started by asking "simple" questions like, "Explain quantum computing in simple terms."
Next, I decided to get a little more advanced by asking it to develop some code: "How do I make an HTTP request in Javascript?"
Surprisingly, ChatGPT was able to answer both of my questions accurately and with a degree of detail that I had not expected, with examples. It was fast and efficient and responded in a human-like manner. I even saw other authors' posts about getting ChatGPT to fix their broken code.
?When I started hearing the buzz around ChatGPT, my mind immediately went to customer experience (CX) and what it could do to change the CX experience. Many customers today respond to and interact with chatbots that are less than ideal or human-like, creating a bad customer experience from the beginning. ChatGPT could revolutionize the CX experience by having a near-human-like conversation with the customer that is indistinguishable from a real human. thus increasing response times and decreasing agent interactions, all while increasing the customer experience.
Lastly, ChatGPT could be a huge asset to customer service teams, as it is much faster and more efficient than traditional methods. My thoughts were going toward the QA industry in CX. If ChatGPT is a chatbot, it needs to understand what’s being asked or demanded of it before generating a response or taking action. If it can understand that, then the same model used to understand that conversation certainly can understand the conversation between an agent and a customer. Applying the same models and algorithms to the intra-call or post-call analysis for QA purposes will also be changed, using models such as GPT-3.5, the one used by ChatGPT.
More to come on this as I explore ChatGPT and GPT-3.5's capabilities.
Seeking next senior Operational or Customer Leadership Role | Former CCW Advisory Board Member
1 年Great piece Tommy - you actually inspired me to publish for the first time on the implications this technology may have for the Customer Self-Service arena!