What is ChatGPT? How it works and the company behind it

When discussing AI, it’s hard not to bring up ChatGPT. Stories about the AI software are everywhere – a Google search on ChatGPT news stories returns over 115,000 different results.

With a current user base of over 180 million, many people are using ChatGPT, but what is it, and how did we get here?

ChatGPT is both a website and an app. The AI tool can be accessed for free, though there are paid options available.

Users of ChatGPT can type anything they want into a textbox, and the AI chatbot will generate a response. The New York Times reports that people have been using ChatGPT for everything from workouts, recipes, gift ideas, proofreading, academic research and more.

ChatGPT was developed by the company OpenAI.

OpenAI was created in 2015. One of the founders and donors was  Elon Musk, but he left in 2018.

OpenAI began as an open-source nonprofit company. According to The New York Times after Musk left, Sam Altman — another founder and now CEO — remade the company into a for-profit organization.

‍According to their website, OpenAI researches and creates AI, and their mission “is to ensure that artificial general intelligence benefits all of humanity.”

Before creating ChatGPT, OpenAI had other AI-based projects. Some early OpenAI projects included Gym – an AI learning platform –  and Universe – a software to measure AI intelligence across different applications. Both were released in 2016.

In 2019, OpenAI released GPT-2, and in 2021 they released GPT-3. These are large language models, which help computers understand and communicate in human languages.

Generative pre-trained transformers models – known as GPT – are the base of other AI tools such as OpenAI’s text to image AI tool called DALL-E.

ChatGPT was first launched in November 2022 and, according to Reuters, within two months of its release date, it set the record for the fastest-growing consumer base for an application in history – reaching 100 million users in two months.

For context, it took TikTok nine months, and Instagram over two years to reach that many users. That record has since been broken by Threads, which Forbes reports gained over 100 million users in five days.

In January 2023, The New York Times reported that Microsoft announced that they would be investing billions into OpenAI’s development.  

ChatGPT is a generative AI, meaning that it takes in a lot of data in order to generate a response to a prompt. According to Stephen Wolfram, the CEO of Wolfram Research, when ChatGPT is given a prompt, it tries to come up with the most “reasonable” next word. This reasonable guess is based on the billions of different web pages and other data sources it’s already read.

‍According to Business Insider, ChatGPT is able to learn and mimic human sentence structures to look realistic. And, though it can talk like a human, it doesn’t actually “think” in the way people do – it’s only mimicking what it knows.

As of publication, ChatGPT is free and available through an app and online webpage. Anyone can use ChatGPT 3.5, and paid subscribers can access ChatGPT 4, which has more advanced features.

Like Google, users of ChatGPT can ask questions into a text box and get results, but the results provided by ChatGPT are different from Google. When prompted, ChatGPT generates a response based on its training data, while Google directs users to websites. The information generated by ChatGPT can be accurate, but sometimes it's not correct

When an AI like ChatGPT lies it’s known as a ‘hallucination,’ and instances of AIs lying have become well-documented.

In June, Reuters reported that two New York lawyers were sanctioned by a judge after they submitted a legal brief which cited six fake cases which were generated by ChatGPT. The lawyers did not realize that the briefs were fake when using ChatGPT to research a case.

‍Google is rolling out its own generative AI as a part of its search tool. When users do a Google search, there is an option for Google’s AI to sort and generate a summary of the information.

There have been other concerns raised around the technology as well. This year, the Federal Trade Commission opened an investigation into OpenAI because of its AI’s potential to put personal data and reputations at risk. The Federal Trade Commission is looking into the “products' potential to ‘generate statements about real individuals that are false, misleading, or disparaging.’”

In March 2023, an open letter was published calling on tech companies to pause training AI systems stronger than ChatGPT 4 for at least 6 months. According to the letter, “Powerful AI systems should be developed only once we are confident that their effects will be positive and their risks will be manageable.”

The letter called for AI labs and experts to come together and create a set of shared safety measures for future AI design, to ensure that “systems adhering to them are safe beyond a reasonable doubt.”

Former OpenAI founder Musk was one of the signatories, in addition to many other famous names such as Apple cofounder Steve Wozniak and former presidential candidate Andrew Yang.

This open letter caused other leaders in the tech industry to push back against these claims. In an interview with Reuters, Microsoft  founder and owner Bill Gates said, “I don’t think asking one particular group to pause solves the challenges.”

“Clearly there’s huge benefits to these things … what we need to do is identify the tricky areas.”

ChatGPT continues to receive updates to its software. Previously, versions of ChatGPT were only trained on information up to September 2021. However, in September of this year, Reuters reported that with GPT-4, ChatGPT will be able to browse the internet using Bing.

‍The BBC reported that this means that users will now be able to ask ChatGPT questions around up-to-date information.

On Sept. 25, OpenAI announced that ChatGPT can now “see, hear, and speak.” Users will soon be able to upload photos and have back and forth voice conversations with ChatGPT.