At BOTSchool, we are always eager to explore new technologies and innovative solutions related to Artificial Intelligence. As providers and enthusiasts about this theme, we remain up-to-date about what is new in the market and how it can be relevant to our activity or society as a whole.

The emergence of ChatGPT has been a significant milestone in the field of Conversational AI and in human-machine interaction. However, it’s essential to clarify the concepts and understand the difference between our work as a Conversational AI Platform and what ChatGPT does.

To achieve this, we engage with both our team and ChatGPT to provide a comprehensive view of this topic.

Let’s start by answering the main question: “What is ChatGPT?”.

Nobody better to answer the question than the ChatGPT itself, this way:

“ChatGPT is a large language model developed by OpenAI, trained on a dataset of conversational text. It is capable of understanding and generating human-like text and can be used for a variety of natural language processing tasks such as text completion, question answering, and language translation. It’s part of the GPT (Generative Pre-training Transformer) models’ family.”

The answer given by ChatGPT doesn’t reflect the level of impact and curiosity among people related to this technology in the past few months since its launch. Its ability to answer any question and generate text allows users to interact and seek help in daily tasks, making it a valuable resource.

But, how ChatGPT knows the answer to so many things?

Contrary to what it may seem, ChatGPT was trained on millions of data points from different sources such as websites, books, articles, and other materials. Through this training, it learns to recognize patterns and relationships between words and phrases, and that way forming text and complete answers.

As ChatGPT explains: “Think of like teaching a child to speak. Just as a child from the language they hear around them, a language model like ChatGPT learns from the text it reads. The more text it reads and the more it is trained, the better it becomes at understanding language and generating responses.”

It was exactly after some years of training the GPT family that it was possible to create a chat for the public. In truth, the story behind this innovation doesn’t start last November.

Let’s take a quick look at the timeline of this technology to understand the complexity behind the ultimate product that is ChatGPT.

The story of GPT family began in 2018 when researchers from OpenAI released GPT (Generative Pre-Trained Transformer) or GPT-1, a language model that was capable of completing tasks such as common-sense reasoning, reading comprehension, understanding sentences, and reasoning through ideas. It contained 117 million parameters and was pre-trained on a large amount of internet content.

In February 2019, ChatGPT-2 was released, more powerful, and containing 1.5 billion parameters. This was the moment when people start to understand the potentiality of this technology and gained attention for its ability to generate coherent and seemingly human-like text. However, this ability leads to concerns about the creation and propaganda of fake news and misinformation.

OpenAI decided not to release the full version of GPT-2 to the public, and only made a smaller version available to researchers and developers.

In June 2020, GPT-3 was released with 175 billion parameters, making it the largest language model to date. It could understand different aspects of language and word relationships, having been pre-trained on an even larger corpus of text than its predecessors. To be more precise and translate this in numbers, the system was fed with 300 billion words.

In November 2022 was revealed ChatGPT that is a fine-tuned version of GPT-3 that has been optimized for conversational interactions. ChatGPT has been used to a wide range of applications since there.

On March 14th, GPT-4 was launched with even more features, the most impressive of which is the ability to analyze images including photos, charts, tables, among other. This new model is more creative and collaborative. According to tests conducted by OpenAI, this version is 40% more likely to produce factual answers when compared to the previous ones. However, access to this version is limited to premium users.

As any new technology, ChatGPT presents limitations and challenges, particularly related to ethical concerns. The model is trained using information from the internet and is therefore susceptible to incorporating incorrect information, stereotypes, and personal biases.

As a result, the Chat may provide misinformation when questioned about certain topics. It becomes important to keep this in mind when interacting with ChatGPT and have the discernment to analyze other sources, if necessary, rather than accepting the information provided as absolute truth.

Efforts are being made to continuously train the model to improve the accuracy of its responses. As a Conversational AI, it is expected to evolve over time, just as other Conversational AI Platforms as BOTSchool, are constantly incorporating new technologies to enhance the accuracy of information provided.

ChatGPT is a Conversational AI, it makes part of this family, however, it can’t be called a Conversational AI Platform, even if are both related to natural language processing and machine learning, they have different purposes.

ChatGPT is a game-changing in the field of natural language. Can be used in a variety of cases, essentially in language processing tasks. The capacity to generate answers in a natural language, very equivalent to what a human could do, raises conversational tools to a new level.

However, it cannot be seen as a substitute for Conversational AI Platforms. Conversational AI Platforms is a term that refers to the use of natural language processing and machine learning techniques to create interactive experiences and human-like conversations between humans and machines, in a variety of use cases related to customer service, customer support, e-commerce, business operations, and personal assistant services among others, that you can explore with our case studies.

An enterprise conversational virtual assistant needs to know answers to specific queries and execute predetermined tasks in a consistent and predictable manner. This requires a level of customization that may not be achievable with a general-purpose language model like ChatGPT.

Our team’s expert in Conversational AI and Artificial Intelligence provided some precious insights about the emergence of ChatGPT and its potential, as well as the consequences from now on to the AI market.

Jorge Sousa, Head of Conversational AI, stated the following:

“We are witnessing a technological revolution in the field of language processing. The emergence of LLM models has opened a range of opportunities, unleashing the true potential of AI in our lives. An example of this is the GPT-3, a Generative Pre-Trained model, and what was done by OpenAI with ChatGPT, a general applicability virtual assistant aimed at the common user.

In my opinion, the biggest evolution that ChatGPT has brought is the public perception of what AI, and in particular, the latest generation LLD models, are capable of.

On the BOTSchool side, we are focused on bringing the evolution to companies, enabling them to have access to these evolutions, reducing the adoption barrier, and allowing them to address from small problems to even the most complex situations. What we do know is that models will continue to evolve, but at the end of the day, we intend to make these evolutions to everyone”.

Sara Furão, a software engineer with an interest in Artificial Intelligence has the following opinion:

“ChatGPT has generated global awareness of the applicability of AI. Due to the type of product generated it managed to reach a lot of people, some of whom did not have knowledge of AI beforehand. This visibility has increased the interest in this kind of AI system.”

About the importance of the technology:

“These systems are quite relevant because they make an agile communication bridge between customers and a company’s services. BOTSchool has already been working in this area. There is already clear evidence of the success of this system, so the future of AI will be very promising and with exponential growth. The model will continue to move forward, with more cognitive capabilities, more intelligence, and more automation in the variety of activity, allowing the automation of more tasks and in more sectors.”

“BOTSchool will continue to improve this technology and achieve more solutions for companies, always ensuring that the main goal is to contribute to a better society by helping to solve and simplify difficulties that people may have in their day-to-day lives.”

While ChatGPT represents a significant milestone in natural language processing and machine learning, it cannot fully replace Conversational AI Platforms. The evolution of GPT until ChatGPT was a huge step forward in improving the capabilities in generating human-like answers in various domains. However, Conversational AI Platforms offer a more comprehensive and customizable solution for businesses to create conversational tools tailored to their specific needs.

Our experts acknowledge the potential of ChatGPT to advance the technology behind Conversational AI Platforms, and we aim to continue to develop our platform so that it offers the best features. The democratization of conversational AI and all its benefits is one of the pillars of BOTSchool and will continue to be.

Therefore, Conversational AI Platforms and ChatGPT serve different purposes and have different strengths. ChatGPT will not replace the use of Conversational AI Platforms, as they are essential to providing innovative solutions for companies to create and customize conversational agents and experiences. Instead, ChatGPT will complement them and improve their overall performance, and make them better and more innovative.