I haven't used anything with GPT-3 first hand (chatbot or other use cases) so reading this article was one of my first exposure to it. It somewhat surprises me that a chatbot powered by GPT-3 doesn't seem that much different from the SmarterChild of AIM in the late 90s / early 00s. That was 20 years ago.
Short of any real break-through in AI, I feel that a chatbot just isn't something most consumers want to use.
A few years ago I bought into the chatbot hype for a while (the Facebook Messenger API opening up for building chatbots was a major catalyst for that particular hype cycle; that was when I was into it), and explored a lot in building one and other AI services that helped build one. It was quickly apparent that for the most part it's simply an iteration of the "answer machine navigation tree" that 800 numbers already had, just in text form.
There was a widespread notion at the time that users wanted chatbots for reasons like wanting to speak in a human langauage with a company to get what they need and solve problems as well as having context from chat history. I think the industry has confused the mechanism (the chat) with the intention (speaking with a human). Consumers prefer a chat sure, that's because they don't want to waste time being on hold on a phone (anyone who has called customer service at any big org or the gov't would know); but ultimately their intention is to speak with a human that can solve problems. When problems arise that require speaking to a human, that's usually not something a bot or a program could solve.
Short of any real break-through in AI, I feel that a chatbot just isn't something most consumers want to use.
A few years ago I bought into the chatbot hype for a while (the Facebook Messenger API opening up for building chatbots was a major catalyst for that particular hype cycle; that was when I was into it), and explored a lot in building one and other AI services that helped build one. It was quickly apparent that for the most part it's simply an iteration of the "answer machine navigation tree" that 800 numbers already had, just in text form.
There was a widespread notion at the time that users wanted chatbots for reasons like wanting to speak in a human langauage with a company to get what they need and solve problems as well as having context from chat history. I think the industry has confused the mechanism (the chat) with the intention (speaking with a human). Consumers prefer a chat sure, that's because they don't want to waste time being on hold on a phone (anyone who has called customer service at any big org or the gov't would know); but ultimately their intention is to speak with a human that can solve problems. When problems arise that require speaking to a human, that's usually not something a bot or a program could solve.