From the course: Tech Trends

ChatGPT

- This semester, my daughter started a class in college and got a note about the policy for ChatGPT use and that got me thinking that AI and machine learning are already permeating our culture. Now ChatGPT is an online application, which launched in November 2022. It reached one million users in only five days. Now the application lets users have a conversation with a set of technologies known as GPT. GPT stands for generative pre-trained transformer. Its goal is to generate or create new content. In order to do this, the tool is pre-trained with up to 175 billion parameters. To process this data, it uses what developers call a transformer, which is really great at understanding the way people write sentences. ChatGPT was created by OpenAI, a company with a mission to create something called AGIs or artificial general intelligences. An AGI could understand any intellectual task that a human being can do. OpenAI's goal is to create AIs that benefit humanity, not replace it. You may have heard about some of their other products. DALL·E 2, an AI system for creating realistic art, as well as Whisper, a speech recognition, translation and transcription system that approaches human-like synthetic speech. You may be wondering how computers are able to do things like creative tasks, that up to now, only humans could do. Now the answer is that developers have created models and algorithms that attempt to duplicate how humans solve creative problems. Here's an example. I worked at a newspaper and I would watch artists doing portraits. They would print out reference material which included photos of the person from different angles and perhaps styles that inspired them. Then they would translate that inspiration into an original portrait. Now to most people, that process seems magical, but developers recognize that sequence of events as an algorithm, a repeatable set of steps that yields predictable results. As the name implies, ChatGPT is focused on conversations and it's capable of writing essays, scripts, resumes, songs and even more accomplished tasks like writing and debugging code. In the case of GPT, the model is known as autoregressive, which means that it's trying to predict results based on past values. That's why ChatGPT seems human at times. It tries to predict not just what you asked for but also what you probably mean. ChatGPT does have some limitations. It can deliver answers that are wrong and do so authoritatively. It was strained mostly with data based on 2021 and earlier, so it doesn't know a lot about current events. Now this version was also trained using human feedback to fine-tune the model so that it was better than its predecessor. It's considered to be version 3.5 of GPT. It's also able to replicate the mechanics of generating content, but it can't understand whether that content is subjectively good or bad. In the future, I think it's going to speed up the development of tasks by helping humans iterate on ideas. ChatGPT, like other artificial intelligences, is more than just an application. There are APIs available that allow you to create your own products based on the models. ChatGPT itself is just an application based on the GPT API. Now the biggest opportunities in this space are for developers that can understand and use the APIs, entrepreneurs that can leverage the technologies to create new and exciting products and for professionals who understand how to work with AIs and get the most out of what the tools can do. Like I heard someone say, you're probably not going to be replaced by ChatGPT, but you might be replaced by someone who knows how to use it.

Contents