What does the release of ChatGPT mean to C-AI Platforms?

The advent of ChatGPT in November 2022 continues to make headlines in the technology world. While ChatGPT is still beta testing, its sheer intelligence and capability indicate that ChatGPT will disrupt the Conversational AI platform(C-AI) market. But before we dive into this topic, let us clarify a few concepts:

• OpenAI vs GPT-3 vs ChatGPT: We often hear people using the three words interchangeably, more so with GPT-3 and ChatGPT.
• Open AI is a research organization that has developed multiple AI solutions such as ChatGPT, GPT-3, DALL-E, Codex, OpenAI Gym, etc. And of course Open AI is not an opensource platform :)
• GPT-3 (Generative Pre-trained Transformer 3)is a large-scale language generation model that can be used to develop commercial applications. The APIs of GPT-3 can be leveraged to build customized applications or solutions using a specific corpus of data. OpenAI offers the GPT-3 APIs in 2 business models.
• Pay-as-you-go or subscription model
• Enterprise License models
• ChatGPT is a specialized version of GPT-3, having advanced capabilities in Natural Language Processing (NLP) and Natural Language Understanding (NLU). One of the key differentiators between GPT-3 and ChatGPT is that ChatGPT is trained on natural language-based conversational data, whereas GPT-3 is trained on a diverse spectrum of internet text. ChatGPT APIs are not yet available for the developer community, it is offered selectively to people based on the purpose of the API usage. Post the availability of the ChatGPT APIs, several intelligent applications can be developed which can be activated via natural language.
• ChatGPT is NOT a chatbot: Many times, ChatGPT gets categorized as a chatbot. Categorizing ChatGPT undermines the true capability and power of ChatGPT. Typically a chatbot handles a list of intents that the developer has developed and trained to perform a set of actions. Chatbots cannot generate content based on the dynamic requests received from the user. In a nutshell, there is a boundary or scope beyond which the chatbot does not work. The unique abilities of ChatGPT such as content generation, deeper understanding of conversations, and context retention makes it clear that ChatGPT can enable technology companies to build hybrid sophisticated conversational applications.
• Natural Language Generation (NLG) and Generative AI are NOT the same: While both NLG and Generative AI are types of AI, they are not the same. Using NLG, human-spoken text or voice-based content can be generated. However, Generative AI is a superior form of AI, where various sets of content, such as images, music, code etc can be artificially created. Both forms of AI typically work well when they receive well-described input from the user.

Impact of ChatGPT on the Conversational AI Platforms

Over the last few years, numerous C-AI low-code and hyper-scale platforms have created a fragmented and cluttered market. However, the market position of the popular C-AI Platforms is likely to get recalibrated now due to the entry of ChatGPT (whenever the APIs are available for commercial use). Platforms must consider integrating and enhancing their product portfolio using ChatGPT

Below are some of the measures that Conversational AI platforms should consider:

Enhance existing Conversational AI products:

• Frequently asked questions (or FAQs) are integral to any Conversational AI Platform. ChatGPT will significantly enhance the FAQ responses from a rules-based static answer to a well-articulated natural language answer to the user
• NLU will be significantly improved with bots able to handle complex multi-intent conversations with the users
• Search Engines for enterprises developed using C-AI platforms will see an improvement in the quality, precision, and articulation of responses.

Transition from text-based bots to Voice: Voice recognition algorithms have matured significantly over the past few years. However, most C-AI platforms have struggled to handle the variety of user utterances in a conversation. Training the bot to understand diverse but valid user utterances, such as acknowledgment, question, comment, small-talk, change of context etc, was challenging. Hence, the selection of voice-based use cases is restricted, and development is time-consuming. Integration of advanced Voice recognition solution with natural language capabilities of chatGPT, will result in text-based AI bots transitioning to voice-enabled AI applications handling natural language conversation with users.

Shift from C-AI Bot to Application Platform: ChatGPT provides an excellent opportunity for Platforms to leap from chatbot Platform to C-AI Application Platform. A combination of the existing NLU solution and ChatGPT’s generative AI and natural language capabilities can result in Platforms capable of supporting the development of sophisticated hybrid conversational applications that can perform complex tasks.

• Report generation: Complex report generation involving multiple data sources, filters, and digital channels will be made more accessible and more scalable using Generative AI.
• Hyper automation: Chatbots b have typically been a last-mile automation solution with high dependency on APIs for developing complex transactions. Integration of chatbots with RPA platforms has improved the automation capability; ChatGPT might just be able to accelerate the hyper-automation through a confluence of technologies using codex and other generative AI capabilities. This can result in true hyper-automation, which has been on the horizon for a long time.
• Conversation mining: Mining of deeper insights and predicting forecasts will be possible from the millions of records of conversational data. Applications developed on conversational mining will enable multiple stakeholders and CXOs with action-oriented insights.

While ChatGPT is promising with many potential opportunities, it also comes with certain drawbacks. Technology companies and C-AI Platforms need to consider them in their ChatGPT journey:

• Since ChatGPT uses unsupervised learning to generate content, there is an inherent risk of generating offensive content with ChatGPT today. While Open AI provides solutions in the form of Moderation API (a tool that can check contents for hate, threat, self-harm, violence etc), C-AI Platforms will need to develop a robust mechanism that can filter and alert about any offensive or biased prompts that can have legal and branding implications.
• ChatGPT does not comply with Explainable AI (or XAI). XAI is a methodology that allows developers or techies to logically understand the outcomes of an AI model. So use cases that are sensitive, especially in the field of medical science, law etc that can lead to severe consequences should be avoided.

Technology companies, both startups and enterprises, have an excellent opportunity to enhance their product portfolio in terms of capability and offering. While ChatGPT continues to mature in the coming months, technology companies must continue to evaluate their product strategy and roadmap.