When we think of conversational AI, we reference an evolving field of AI that equips computer systems with the ability to communicate with humans using natural language (NL) in an interactive manner.

Over the past decade, conversational AI-powered voice- and text-based assistants have become embedded in our daily lives, enhancing user experience across diverse platforms and tackling a wide range of use cases. Revolutionizing the way businesses interact with their customers, they have become critical components of a modern digital strategy.

This section aims to provide a comprehensive overview of the field of conversational AI. It will delve into the history of chatbots and conversational AI, highlighting key milestones that have shaped the progression from rudimentary rule-based systems to sophisticated AI-driven chatbots. We will examine advancements in technology, such as NL understanding (NLU) and the progression from rule-based chatbots...

When chatbots first entered the scene, they were hailed as a groundbreaking evolution in technology, poised to revolutionize the way businesses communicate with their customers.

The early versions, however, were not as impressive as many had hoped. Coupled with the growth of chatbot platforms and the ease with which anyone could create a chatbot, this led to a bad user experience and an overall lackluster performance, which hampered their reputation.

One of the primary failings of early chatbots was their lack of sophistication and nuance in understanding and responding to human language. These rudimentary versions were generally rule-based, relying heavily on pre-programmed scripts. This meant they were only equipped to handle very specific inputs and could easily be thrown off by anything that deviated from their script. As a result, users often found themselves in frustratingly circular conversations, where the chatbot failed to...

Before looking at ChatGPT, it’s important to consider use cases for conversational AI applications and initially consider them in the context of what’s been possible with traditional conversational AI systems and technology.

Conversational AI has emerged as a powerful tool in numerous industries, from customer service to healthcare, banking, insurance, retail, and human resources. By automating routine tasks and enabling enhanced user interactions, conversational AI applications have changed the landscape of many domains. Let’s look at some existing examples of conversational AI.

Customer service

Conversational AI has been hugely successful in the customer support use case, streamlining processes and improving user experience in myriad ways. By adopting AI-powered chatbots and virtual assistants, businesses can handle large volumes of customer interactions seamlessly and efficiently.

Conversational AI offers...

ChatGPT is OpenAI’s chat-specific LLM that uses ML to generate human-like responses to your text inputs.

Within just over a month of launching in November 2022, ChatGPT had over 100 million users. By February 2023, its total website visits had skyrocketed to a billion. This meteoric rise has made ChatGPT the fastest-growing consumer software application in history.

ChatGPT can explain scientific and technical concepts in a desired style or programming language, brainstorm basically anything you can think of asking... and yes, of course, hold complex conversations!

Understanding the large language technology behind ChatGPT

ChatGPT is built on a type of AI called an LLM. Within the domain of LLMs, there are two primary types: basic LLMs and instruction-optimized LLMs. Understanding the difference between these two types is important when examining OpenAI’s ChatGPT.

Exploring basic LLMs

Basic LLMs are programmed to predict...

LLMs offer a plethora of practical uses, with ChatGPT distinguishing itself through its advanced NLP capabilities. In this section, we unpack the array of applications that this technology brings to the table and a summary of the standout industry areas where ChatGPT holds the most potential. While many of these use cases are not necessarily new for conversational AI applications, it’s true to say that ChatGPT can potentially take on more advanced tasks.

Capabilities of ChatGPT

In the following sections, we’ll explore the various applications of ChatGPT, starting with NLU and NL generation (NLG) for chatbots and virtual assistants. Then, we’ll move on to its capabilities in machine translation, summarization, sentiment analysis, and content creation. Additional topics include its use in spam filtering, language tutoring, and software development. Finally, we’ll discuss the extended features available through...

ChatGPT, as with other LLMs, is not fallible and has several significant risks, limitations, and security considerations that are important to be aware of if you are going to use ChatGPT in production conversational AI use cases.

Limitations

There is no doubt that ChatGPT has some incredible capabilities, but let’s take a look at where things can go wrong.

Accuracy and depth

Probably the most well-known limitation and the one that has garnered the most effort to overcome is ChatGPT’s “information cutoff” date. The early GPT-4 models were trained on publicly available internet data, with cutoff dates as far back as 2021. The latest GPT models have shown improvements. According to OpenAI, the newest GPT-4 Turbo (gpt-4-turbo) has a knowledge cutoff of December 2023. However, beyond this point, it is not updated with the latest information or events. This limitation means that conversational AI implementations using ChatGPT...

In conclusion, in this chapter, we’ve seen how conversational AI has made steady progress from rule-based to AI-driven intent-based systems. By looking at the applications of ChatGPT, we’ve seen that as the capabilities of modern conversational AI systems have grown, so have the breadth and depth of use cases for conversational experiences. There are thousands of these systems, carrying out complex tasks across multiple industries.

From our initial focus on ChatGPT, we have seen it’s a powerful tool that has the capabilities to revolutionize the field of conversational AI by taking on more complex tasks and use cases that until recently have been difficult to achieve with modern conversational AI platforms. It offers a wide range of applications in many industries, including customer service, healthcare, finance, law, education, and more.

However, the chapter has also highlighted that it’s important to be aware of its limitations and risks...

To explore more on the topics related to this chapter, you can visit the following links:

• Enterprise conversational AI

  https://www.gartner.com/reviews/market/enterprise-conversational-ai-platforms

• Bridging the Gap: A Survey on Integrating (Human) Feedback for Natural Language Generation

  https://arxiv.org/abs/2305.00955