In this overview article of ChatGPT statistics, we’re examining some of the most interesting statistics, facts, and data points related to the world’s most popular generative AI app.

In a relatively short time, OpenAI’s “ChatGPT” solution has effectively transformed the world as we know it. It’s introducing new ways for companies to create content, write code, and even communicate with customers on a massive scale.

The popularity of ChatGPT has even led to a massive increase in companies investing in similar LLMs (Large Language Models) and generative AI solutions. The market for generative AI is now growing at a rate of 32% CAGR, set to reach a value of $126.5 billion by 2031.

So what is ChatGPT, why is it so transformative, and what are the stats business owners and users should know before they dive into the generative AI realm?

Read on for all the top stats on ChatGPT.

What is ChatGPT?

Before we dive into some interesting ChatGPT statistics, it’s worth discussing what ChatGPT actually is. In simple terms, it’s a chatbot, or conversational AI tool, built on large language model (LLM) technology. Created by US-based AI company, OpenAI, the solution is designed to respond in a human, creative manner to human input.

When users ask ChatGPT a question, or ask it to create content on their behalf, it draws on a huge body of data, to create realistic, human-style responses. The solution is built on top of OpenAI’s GPT family of large language models, and has been upgraded through the years.

ChatGPT can write essays and poems in a specific tone of voice, create computer programs and codes, have personalized conversations with users. Plus, OpenAI’s technology is quickly making its way into other software and tools. For instance, Microsoft is embedding ChatGPT tech into Bing search engines, Microsoft Teams, and a range of other tools.

When Was ChatGPT Launched?

Most people became familiar with ChatGPT in 2023, when OpenAI opened the algorithm to the public, and allowed users to start experimenting with the service. However, OpenAI had already been working on the model for five years, starting in 2018.

The GPT technology powering ChatGPT has gone through numerous iterations in this time, including GPT-1, GPT-2, and GPT-3. The first model was trained primarily on language understanding, while the second was given access to Reddit articles and other content. The GPT-3 model most people are familiar with today was officially released in 2020.

This model is also used to power various other generative AI apps, such as DALL-E, which can create images from text, Whisper, for multi-lingual voice-to-text applications, and CLIP, for connecting images and text in content.

illustration of a cat climbing a ladder

How Does ChatGPT Work?

ChatGPT is an extremely sophisticated conversational AI bot or “generative AI” solution, built using large language models. Though it’s fantastic at creating human-style responses to queries, it’s not actually sentient. It works by accessing huge volumes of data to generate responses.

OpenAI trained the initial large language model used to build ChatGPT using reinforcement learning with human feedback. This meant human trainers gave the model conversational data to teach it how people speak, and how to respond to questions.

GPT stands for “Generative Pre-Trained Transformer”, the core technology behind ChatGPT. It’s a form of neural network machine learning model, which leverages millions of data points from the internet. ChatGPT’s training dataset included 300 billion words, and its database is constantly evolving. Using this data, ChatGPT can create intelligent responses to virtually any question or query, though the functionality of the app depends on the version you use.

illustration of a cat climbing a ladder

ChatGPT 3.5 vs ChatGPT 4

As mentioned above, ChatGPT, and the GPT technology has been a work in progress for some time. ChatGPT-4, is currently the most recent version of the software, officially released in March 2023. This version of the software is far more advanced than its predecessor, though it’s only available with a paid subscription to ChatGPT Plus.

ChatGPT-4 is trained on 100 trillion parameters (the same scope as the human brain), meaning its training data could be 571 larger than the parameters used for ChatGPT-3. It can also respond using up to 25,000 words (8 times more than the previous version) and can process image inputs alongside text. It also has the ability to browse the internet for data.

GPT-4 is currently available in two variants (an 8k model and a 32k model), and can handle longer documents, more input, and more personalized conversations than ChatGPT 3.5. Additionally, the newest version of the app is 82% less likely to respond to requests for disallowed content, and 40% more likely to deliver factual, accurate responses.

illustration of a cat climbing a ladder

How Many Users Does ChatGPT Have?

ChatGPT has broken records as the fastest-growing application of all time. The solution earned 1 million users after just 5 days of launch. Additionally, 2 months after its launch, ChatGPT earned more than 100 million monthly active users.

To put this into perspective, it took the rapidly-growing social media platform TikTok 9 months to earn the same number of users. Facebook took almost 5 years to reach this milestone.

According to SimilarWeb, ChatGPT’s website earned around 13 million unique visitors per day in January 2023 (double the daily number in December), and the site’s traffic has grown by around 3.4% per day for the last month. SimilarWeb also found ChatGPT recorded around 1 billion visitors in February, with visitors to the OpenAI website looking at around 6.22 pages per visit.

If we look at the demographics of ChatrGPT’s users, we find that the majority of consumers (11.72%) are located in the US. However, the platform is also popular in India (10.67%), Japan (4.29%), France (3.98%), Indonesia (3.45%), Canada (3.04%).

Here are some other ChatGPT user statistics:

  • ChatGPT is available in 161 countries worldwide: While ChatGPT has users all over the world, the platform is either banned, or unavailable in certain regions, including China, Russia, the Ukraine, Belarus, Venezuela, Afghanistan and Iran (Source: OpenAI)
  • 63.5% of users can’t tell the difference between human or GPT-4 generated content: Tooltester found 63.5% of consumers couldn’t tell the difference between an article created by ChatGPT-4, and a human writer. (Source: Tooltester)
  • 71% of customers would lose trust in a brand for using ChatGPT: The Tooltester survey also found 71.3% of readers would stop trusting a brand if they created content with ChatGPT without being transparent about their actions. (Source: Tooltester)
  • Most ChatGPT users are male: The audience composition for Openai.com and ChatGPT is around 34.32% female and 65.58% male. (Source: SimilarWeb)
  • Many ChatGPT users are younger: Perhaps unsurprisingly, most ChatGPT users fall into the age group of between 18 and 34. (Source: SimilarWeb)
  • 17% of 30-44 year olds have used ChatGPT: YouGov found 17% of 30 to 44-year olds have used ChatGPT since 2023, and 15% of 18-29 year olds have used the tool to generate text. Around 9% of 45-64 year olds use the platform. (Source: YouGov)
  • 46% of Americans have heard of ChatGPT: Around 54% of Americans in the YouGov study said they’ve heard nothing about ChatGPT, while 34% say they’ve heard a little, and 13% say they’ve heard a lot. Men, Americans with higher education levels, and younger adults were more likely to be familiar with the software. (Source: YouGov)
  • Around 9 in 10 visitors to ChatGPT are direct visitors: 87.99% of visitors come to the ChatGPT site directly, searching for terms like “ChatGPT login”, Social traffic accounts for around 4.61% of the company’s visitors, and YouTube is the top channel for social traffic, responsible for around 3 in 5 social visits. (Source: SimilarWeb)
  • Around 9,600 keywords are used to access ChatGPT: The most commonly searched term is “Chat GPT login” with 6.1 million monthly searches followed by “ChatGPT login” with 5.3 million monthly searches. (Source: SimilarWeb)

illustration of a cat climbing a ladder

What are the Costs of ChatGPT?

Experts in the software industry estimate running costs for ChatGPT range somewhere between $100,000 and $700,000 per day. This pricing is based partially on the costs of Microsoft’s Azure cloud infrastructure, which charges around $3 per hour for one A100GPU super computer cluster. The same experts suggest each word costs around $0.03 to generate.

Of course, the price of super computer technology isn’t the only thing OpenAI needs to account for. OpenAI currently employs around 524 people according to RocketReach, and many are positioned in high-paying roles. For instance, Glassdoor estimates OpenAI salaries range from $47,922 per year to around $284,231 per year for the CTO.

Despite relatively high running costs, ChatGPT can be accessed by users for free. The free version provides access to a complete research preview of the app, which allows OpenAI to gather feedback from early adopters, and test strengths and weaknesses.

However, only the premium version of ChatGPT, (ChatGPT Plus) supports the latest GPT-4 algorithm and newest features. This is offered at a $20 per month fee. There’s a waitlist for accessing the service, and a separate waitlist for developers and businesses looking to use an official ChatGPT API.

While it’s expected that OpenAI will create multiple pricing plans for ChatGPT everyday users, businesses, and developers, the company has said it will continue to provide a free version.

illustration of a cat climbing a ladder

The Value and Revenue of ChatGPT and OpenAI

The value of OpenAI and ChatGPT have continued to accelerate as the company’s user base has grown. CNBC ranked the OpenAI brand as its number 1 disruptor brand for 2023, with a valuation of around $30 billion. By 2024, Reuters predicts the organization will generate around $1 billion in revenue, making it an extremely profitable brand.

Notably, OpenAI has also had quite a lot of money offered up by investors over the years. The company, founded by Sam Altman, Elon Musk, and a variety of other innovators, earned around $1 billion in investment from its initial founders, and a further $3 billion from Microsoft over the years.

In 2023, Microsoft also announced it was investing more money into the ecosystem as part of its third partnership phase, which brought more generative AI to Microsoft’s features. Some industry analysts estimate the company may have lost money at one point too.

According to “The Information” the company may have lost around $540 million in 2022. Of course, as OpenAI continues to roll out new versions of ChatGPT, and new pricing packages, its overall earning potential is likely to grow.

illustration of a cat climbing a ladder

How Long Did it Take to Train ChatGPT?

As mentioned above, OpenAI spent around 5 years creating the ChatGPT model we know today, and producing the innovative tool took significant work. Some specialists predict that if the model was trained on a single Tesla (NVIDIA) V100 GPU, it would take 355 years to create.

Since it’s believed OpenAI used around 1,023 A100 GPUs to train the model, this indicates the training may have taken as little as 34 days. Lambda Labs also estimated the cost of this intensive training was likely to be around $5 million dollars.

The dataset for training ChatGPT-4, the most recent version of the model, is estimated to include around 100 trillion parameters, much more than the training data used for ChatGPT-3. The training data included feedback from users of ChatGPT-3, as well as insights from more than 50 AI security and safety experts.

The ChatGPT-3 dataset included text-based data from 5 sources, each with its own proportional weighting. Additionally, 60% of this dataset was based on a filtered version of common crawl data, made up of metadata, web page data, and text extracts.

Another 22% of the dataset for ChatGPT-3 came from WebText2 content, which consisted of popular Reddit posts, and 16% came from internet-based book collections. Finally a further 3% of the dataset included insights from the English-language version of Wikipedia.

illustration of a cat climbing a ladder

How Intelligent is ChatGPT?

Clearly, ChatGPT is built on a massive dataset, making it an extremely intelligent tool. ChatGPT is designed to work primarily in English, but it also understands inputs in 95 other languages, and the dataset is continuing to grow.

It’s also worth noting that ChatGPT was written primarily in the Python language for computer programming. However, it was exposed to various other coding languages throughout its training, which means it can now understand and produce responses in a variety of other computer languages.

ChatGPT can process content in languages such as JavaScript, C++, C+, Java, Ruby, PHP, SQL and more. Despite all of this phenomenal data, it’s worth noting that ChatGPT still isn’t 100% accurate or factual in its responses to every query.

ChatGPT-3, in particular can struggle to answer a lot of questions from more recent years. The previous version of the app is only trained on data spanning through to 2021, which means it’s not familiar with a lot of information from recent years.

illustration of a cat climbing a ladder

What Can ChatGPT Do? The Applications of ChatGPT

Like most disruptive tools, ChatGPT has various pros and cons to consider. It’s also a solution that’s constantly growing and evolving, based on new input and training from OpenAI, and feedback from users of the free version of the model. However, the solution has already proven to be extremely versatile in its use cases.

Some of the most common uses for ChatGPT include:

  • Writing and debugging code: Since ChatGPT can understand a variety of different computer programming languages, it’s becoming quite popular in the coding world. The tool can provide full snippets of code for professionals, and explain what each part of the code is responsible for, though it may still make errors.
  • Copywriting: One of the most common uses for ChatGPT is to create content for websites, ads, and other online platforms. By entering a description of the copy you want the system to write into the platform, you can prompt ChatGPT to create essays, blogs, opinion pieces, and more, though most content may still require some basic editing.
  • Creating newsletters: ChatGPT is excellent for creating email newsletters, and the system can even offer guidance on how frequently to connect with customers. The solution is also suitable for providing ideas on what kinds of content to create.
  • Writing resumes: Entering all of your qualifications, work experience, and background knowledge into ChatGPT can help it to create resumes and cover letters for new jobs. Although some companies may frown on this usage.
  • Creating summaries: If you don’t have time to read through long documents, or you want to convert a large piece of writing into something smaller and more concise, ChatGPT can summarize the key details of a document for you.
  • Chatting: ChatGPT is primarily a chatbot, which makes it excellent for conversations. It’s good for answering questions and giving advice about situations. Many companies are using similar tools for their self-service chatbots on websites.

ChatGPT can also produce jokes, songs, and poems, because it has a huge amount of source material to work from, and can produce content in a variety of tones. In some cases, it can even be used to produce a basic version of legal documents and contracts.

Microsoft’s investment in OpenAI has also expanded the functionality of ChatGPT in recent years. In 2023, Microsoft introduced the new AI-powered version of their search engine, enhanced by the same technology used by OpenAI for the chatbot. Google has since taken a similar approach, creating it’s own generative AI solution (Bard) and adding AI capabilities to the Google search engine.

illustration of a cat climbing a ladder

Does ChatGPT Have Any Limitations?

While ChatGPT has certainly proven itself to be a highly versatile and useful tool, it does have some limitations. OpenAI, the creators of ChatGPT have openly acknowledged there are still issues with the tool. This is part of the reason why the company continues to offer the free to access version of ChatGPT, so it can enhance the solution based on user feedback.

ChatGPT can sometimes produce answers to questions that seem plausible, but don’t make a lot of sense. Fixing this problem can be complex, as Reinforcement Learning training can’t always verify the truth. Changing the model to be more cautious and accurate would mean it may decline to answer certain questions in the future.

Notably, input phrasing in prompts can have a huge impact on the overall performance of the tool. The way a user inputs a question can make a difference to whether ChatGPT can understand and respond to the query. If ChatGPT struggles to answer a question, it will also guess at the response, rather than asking questions to clarify a user’s meaning.

ChatGPT also has some ethical and security-based issues. For instance, it can sometimes exhibit biased behavior based on certain inputs, although OpenAI is working on fixing this problem. ChatGPT isn’t programmed to produce content that’s violent or gory, and it won’t respond to questions that condone unethical or illegal behavior.

illustration of a cat climbing a ladder

What are the Main Competitors to ChatGPT?

The success of ChatGPT on a global scale has prompted countless other technology companies to begin experimenting with both large language models and generative AI. Google, for instance, launched early access to its own conversational AI system, Bard, in March 2023. Like ChatGPT, the solution provides searchers with synthesized information based on a huge database.

Though initially popular, Google’s Bard live event led to a significant loss in stock value for the company, amounting to around a 7% reduction. This is because the platform generated the wrong answer to a question during the live video.

Baidu, a Chinese tech giant, also announced it was releasing its own generative AI chatbot in 2023, known as “Ernie”. Initial testing was completed in March this year, and shares in Baidu rose by around 13.48% in value based on the press release.

Microsoft is perhaps one of the biggest investors in the generative AI landscape today, with a significant holding in the OpenAI company. The latest version of the Bing search engine now includes ChatGPT technology. Additionally, Microsoft has developed a generative AI tool named Copilot based on OpenAI’s input, which is available throughout Microsoft tools like Office and Teams.

illustration of a cat climbing a ladder

The Ultimate List of ChatGPT Statistics

Although OpenAI has been working on ChatGPT for a number of years, the world has only begun to scratch the surface of what’s possible with this platform. The OpenAI company has earned significant funding through the years, and continues to expand the quality of its chat tool.

Going forward, it seems likely that ChatGPT will continue to grow in both functionality and popularity as OpenAI continues to implement feedback from initial testers.

We’ll try to keep this list of ChatGPT statistics up to date, as new information is revealed about the evolving application.


Source link