What is ChatGPT?
ChatGPT is a generative artificial intelligence (GenAI) chatbot which is publicly available on the internet.
Over the past few years, it has become a part of the daily routine of 200 million users around the globe.
It uses the power of a supercomputer to complete given tasks, from writing a poem to helping with home renovations.
What does GPT stand for?
GPT stands for generative pre-trained transformer, which is a type of large language model (LLM) widely used in GenAI chatbots like ChatGPT.
The model is based on a deep learning architecture called the transformer, which was first proposed by researchers at Google in 2017.
OpenAI was the first to apply generative pre-training to the transformer architecture, introducing the GPT-1 model in 2018.
GPTs are pre-trained on large data sets of content, which allows them to generate new content.
When did ChatGPT come out?
The popular chatbot was released on November 30, 2022 using the GPT-3.5 model.
It soon became the fastest-growing consumer software application in history, with 100 million users joining within the first two months.
It is believed that ChatGPT sparked the AI boom, which saw massive investment in the field and the release of other similar models by other companies such as Google Gemini or Microsoft Copilot.
Who created ChatGPT?
The GenAI tool was created by OpenAI, which is a research company based in San Francisco, California.
It was founded in December 2015 by a group of people including billionaires Sam Altman and Elon Musk.
Initially a non-profit, OpenAI was restructured into a capped-profit company in 2019.
Its mission is to “ensure that AI benefits all of humanity”.
While many people contributed to the development of what has become the world’s favourite AI chatbot, some of the key figures involved in its creation include:
- Sam Altman: As the CEO, Altman set the overall vision and strategic direction for the company. He was also key in raising the capital needed to fund the expensive research and development of a LLM.
- John Schulman: As the lead developer, he was instrumental in the creation of the model. His team used human feedback to train the model to be more helpful and accurate.
- Ilya Sutskever: As a co-founder and former chief scientist at OpenAI, he oversaw much of the foundational research that made ChatGPT possible.
- Greg Brockman: As the President and a co-founder, Brockman’s role was more focused on the operational and technical aspects. He oversaw the development and management of the massive computing infrastructure required to train and run the AI models.
Who owns ChatGPT?
As the primary owner and developer of ChatGPT, OpenAI is the organisation responsible for creating and maintaining the chatbot.
The company has a unique ownership structure with both non-profit and for-profit elements.
The original non-profit OpenAI, Inc. serves as the parent organisation while OpenAI LP, a capped-profit subsidiary, handles commercial activities.
Microsoft has made substantial investments in the company, which gives it a large stake in OpenAI’s success and access to its AI technology.
The tech giant currently holds 49% in the for-profit entity, with the remaining 51% held by other stakeholders and the non-profit foundation.
However, operating on a capped-profit model means that the profits investors can receive are limited.
What can ChatGPT do?
While its initial and main function is text generation, ChatGPT is now capable of a wide range of tasks. This includes:
- Text analysis and summarising: Need to quickly get through a large amount of text? Copy and paste it to the chat window and let the magic happen!
- Proof-reading: This is particularly handy when you need to send an important email early in the morning or you feel like your article just needs some final touches.
- Brainstorming: ChatGPT is excellent for overcoming writer’s block and generating a starting point for any project.
- Language translation: From Czech to Chinese, ChatGPT can serve as a great interpreter or even a language teacher.
- Answering questions: Many people have started using ChatGPT to find quick answers to their queries instead of Google.
Besides, the more you use the chatbot, the more it learns about you and can tailor its responses to you.
Can ChatGPT read PDFs?
Since autumn 2023, ChatGPT can access and process Portable Document Format (PDF).
This allows users to upload their PDFs directly to the chat interface.
The AI model can then analyse and summarise the file or extract a specific piece of information from it.
There are certain limitations though, such as file size and length. For larger or more complex documents, the chatbot may not be the best choice.
As the creator of the PDF format, Adobe offer its own AI Assistant built directly into the Adobe Acrobat software, giving it a deep understanding of the document’s structure.
Can ChatGPT generate images?
In October 2023, OpenAI integrated its image generation software DALL·E into ChatGPT, allowing it to respond to users’ requests visually.
This feature became available to free users in August 2024.
Can ChatGPT edit photos?
The ability to directly edit images came to the AI assistant with the rollout of GPT-4o, which began in late March 2025.
The key change with the update was that image generation became a native capability of the core model itself, allowing for a smoother interaction.
This means that users can upload an image and ask ChatGPT to make specific changes, as well as refining previously generated image.
You can for example turn yourself into a Barbie doll or an action figure.
Can ChatGPT transcribe audio?
OpenAI’s Whisper model, built into ChatGPT, offers a highly accurate speech-to-text transcription.
You can either upload an audio file directly into the chat, or use the record mode to take notes from meetings and lectures.
The ChatGPT mobile app also has a voice input feature that allows you to speak your prompts or have a real-life conversation with the AI model.
Can ChatGPT generate videos?
While OpenAI allows users to generate videos, this functionality is currently only available through its advanced, specialised model Sora.
The text-to-video service allows paying users to generate a short video clips based on given description, including the style and characters.
How does ChatGPT work?
As a LLM, ChatGPT’s magic is based on accurate prediction of the most probable word in a sequence of text.
The model was trained on a massive amount of data from the internet, books, and other sources to match patterns in the correct context.
Its neural network is designed to process and understand sequential data, like the words in a sentence using a mechanism called self-attention to weigh the importance of different words in a prompt.
This allows it to grasp the context and relationships between words, no matter how far apart they are.
For example, in the sentence “A man was wearing a hat, and it was red,” the model understand that “it” refers to the hat, not the man, by connecting the words.
Further training was then then needed to ensure that the model’s responses are helpful and harmless.
A fine-tuning process is what makes ChatGPT a conversational chatbot rather than just a text predictor.
When you give it a prompt, the transformer quickly goes through a three-step process:
- Understand your question: It breaks down your words and the conversation history to understand the full context of what you are asking.
- Predict the best answer: It uses its training to predict the most likely and relevant next word, then the next, and so on, building a complete response one piece at a time.
- Refine the response: It checks the generated answer to make sure it is helpful and safe before showing it to you.
This whole process happens within seconds, which would not be possible without a supercomputer infrastructure built and managed by Microsoft.
It is essentially a massive cluster of tens of thousands of powerful graphic processing units (GPUs) interconnected by a high-bandwith network.
This infrastructure serves as the engine for all of OpenAI’s models, from the initial training of GPT-3 to the real-time operation of ChatGPT for hundreds of millions of users.
