What is ChatGPT?


In the world of AI, new developments are on the rise. That includes one of the newest and most well-known advancements — ChatGPT. 

This AI chatbot is a powerful tool from OpenAI that allows you to input questions and get conversational answers back. Not only that — this bot remembers your ongoing dialogue and uses previous conversation points to inform future responses.

One common question is: Where does it get the information? ChatGPT generates its answers from the volumes of information readily available on the Internet. The tool is making waves because it’s free to use, much more interactive than previous chatbots, and is capable of following along with the conversation. It has even been shown to write poems and songs, copywrite emails, or generate blog content. 

It’s important to note that the developers at OpenAI encourage users to be careful when using the tool. After all, it is a machine gathering information from a digital space, and it could output erroneous information sometimes.

So, while most people aren’t expecting this tool to fully take over human content writing, it does seem to have a lot to offer. Let’s take a closer look at the ChatGPT chatbot and how it works.

The imitation game

Chatbots and AI have been around for years as researchers and companies have sought a better way for customers to get information and answers. A host of computer scientists attempted the Turing Test in 1950. Also known as the Imitation Game, it was proposed by Alan Turing to try and determine whether humans can tell if they are conversing with a human or computer. 

There’s been plenty of advancements since then — particularly in generative AI where the machines can deliver outputs on their own. Within just a few days of launching, ChatGPT had more than a million users trying it out. So, what makes it different than early computers from 1950?

How is ChatGPT different?

ChatGPT is a chat interface that allows regular humans to interact with a family of innovative generative AI models including GPT3.5, GPT4, Dall-e (text-to-image), and more — all without needing to know how to write code. The ease of use, generative AI capabilities, speed, and responsiveness are what help it stand out. But development teams can lean on OpenAI and ChatGPT in other areas of AI, too, and can train the models through these methods.

The GPT genus of large language models does the following:


It ingests and does unsupervised training on large amounts of language data (books, articles, websites, etc.). "Unsupervised" means that the AI's learning is self-directed. It identifies universal statistical patterns in human language on its own. It also means it doesn’t require human direction or intervention. Still, the model relies on data to source information.


This is where humans do come in. With "supervised training," humans can continue to improve upon the model with human-labeled data sets. That way, the AI gets better at generating responses that are coherent for humans.


This is centered around the learning that takes place from human feedback. This is the refining step where a team of users interact with a pre-production version of the model and rate the quality of the generated output as part of an extended training phase. Every other chatbot variant that came before the generative language era is simply doing a fancy VLookup between a chatter's input and some response that was pre-written by another human. But generative AI models work differently. 

Through these processes, researchers and developers can customize the models so it’s specific to their own use cases, and can make the outputs more accurate too.

The downside of ChatGPT

You’ve probably already guessed the one main downside: the chatbot might be wrong. If you and five friends ask ChatGPT the same question, you may get six different answers. Sometimes it’s right, sometimes it might be flawed. 

Plus, it’s more difficult to vet the information given by ChatGPT, as opposed to a search engine like Google, which includes a title, URL, and plenty of other identifying information for you to dig deeper. ChatGPT could pull information from anywhere on the web and you may not be able to find the source or verify if it’s even reputable. 

Some people might assume that a tool like ChatGPT will soon run Google Search out of business. However, they still two serve two different purposes. Google gives you multiple options for information that it feels is most relevant —- allowing you to surf, explore, compare, and research further. ChatGPT on the other hand delivers one superior answer that it believes is the best and most direct response. And for simple and straightforward commands, this works out great. It may just depend on what you’re trying to accomplish as to which one is best for you.  

Save time and money with our AI-powered virtual receptionists 

Instead of a bot-only (like ChatGPT) or human-only solution (like your customer service reps), why not get the best of both worlds? Our AI-powered live virtual receptionists are available to answer calls 24/7. We can serve as your answering service as well as handle your lead intake, appointment scheduling, and live website chat support —- so you can enjoy the speed and precision of AI along with the added layer of the human touch that’s needed for complex or sensitive issues.

To learn more, schedule a consultation or reach out to hello@smith.ai.

Business Education
Written by Samir Sampat

Samir Sampat is a Marketing Manager with Smith.ai. He has experience working with businesses of all sizes focusing on marketing, communications, and business development.

Take the faster path to growth.
Get Smith.ai today.

Affordable plans for every budget.

Sign up for our newsletter

By signing up, you agree to receive news & updates from Smith.ai.
Thank you!
Your submission has been received.
Yikes! Something went wrong while submitting the form.

Take the faster path to growth.
Get Smith.ai today.

Affordable plans for every budget.