The Evolution of Chatbots: From Elisa to ChatGPT (Full Transcript)

Speaker 1: In 1964, researchers at MIT University were working on a computer program, a program so ahead of its time that it would allow seamless communication between humans and machines. In the next two years, they built Elisa, an app that would set the foundation for all the future chatbots. What made Elisa so impressive was the introduction of keyboard-enabled responses. For the first time, users felt they were talking to someone who understood their input.

Speaker 2: Pretend that you have some psychiatric problem. Type it in and see what happens. I'm depressed.

Speaker 1: By 1995, another language processing bot, Alice, came out and was followed by Smarter Child in 2001, setting the stage for all the current generation chatbots. However, a decade later, the introduction of AI chatbots like Amazon Alexa, Google Now, and Siri took the world by storm. But in November 2022, an artificial intelligence firm called OpenAI introduced ChatGPT, an advanced AI-powered chatbot that has taken the world by storm. Let's understand what ChatGPT is. To better understand this chatbot, I went to its website and tried asking a few questions. But now you might think this is not new, and existing chatbots like Alexa, Siri, and Google Assistant can also do this. So what makes ChatGPT so special? While Siri and Alexa can interact with the physical world and complete actions on apps and devices, ChatGPT can't do any of that. Instead, it is designed to provide more detailed and comprehensive answers. For example, writing an essay. Now this is something that other chatbots cannot do. ChatGPT, as it's said, is based on generative pre-trained transformer architecture that is trained on a massive amount of text data from the internet. This is a type of neural network that was introduced in 2017 in a paper titled Attention is All You Need. A neural network is a large network of computers that can fine-tune its output based on the feedback given to it during stages of training. This allows the model to better understand the context and meaning of the input and to generate conversational responses. It can be used to write articles, stories, and even poetry. The capacity of ChatGPT to produce responses like a conversation is one of its key advantages. It uses 570 GB of text data mined from the internet to answer a wide range of queries. Fundamentally, ChatGPT is a language model that can produce text that sound like human speech in a conversational setting. This implies that it can be used to develop chatbots that converse with students naturally and humanely in the academic world. But teachers and academicians are not thrilled about it, especially when it comes to written assignments. They feel that the bot could be used to turn in plagiarized essays that could be hard to detect for invigilators who are pressed for time. Most recently, New York City's education department banned ChatGPT in its public schools. The authorities have forbidden the bot's use in all devices and networks connected to schools. Like any natural language processing model, ChatGPT has limitations related to caliber and volume of the training data. A representative and a diverse dataset was not used to train the model. In that instance, it might not produce reliable answers to all the queries.