Even a couple of years ago, AI applications were very few in number and didn’t have a significant impact on global society.

This has all been turned upside down quite recently. In 2022, the global artificial intelligence market size was estimated at around $120 billion, and it was predicted to reach $1,600 billion by 2030 with a CAGR of 38.1%.

A big part of why AI and machine learning are the most hyped and fastest-growing technology of current times are AI chatbots. The most notable example is ChatGPT, a popular chatbot developed by OpenAI.

It is a vast understatement to say that ChatGPT made noise when it was introduced. As we will see later, it has become the fastest-growing platform in history. And with a good reason.

ChatGPT Statistics in a Nutshell

ChatGPT was created by OpenAI, a company founded in 2015 by Elon Musk, Sam Altman, and others, as a non-profit.
ChatGPT is the fastest app to get to 1 million users (just 5 days)
It broke the record for the fastest-growing platform (2 months to reach 100 million users)
Its website has over 60 million visits per day (17th most visited website in the world)
The newest version (GPT-4) is already at human expert level in many fields
Microsoft invested $10 billion in OpenAI in early 2023, having reported a 49% stake in the company
US users are leading in numbers, followed by India, Japan, and many developing countries
It is estimated that the current version (GPT-4) has between 1 and 170 trillion parameters
The average response length by ChatGPT is estimated to be 30 words
OpenAI’s revenue is projected to be around $1 billion in 2024
ChatGPT’s operating cost is estimated at around $700,000 per day (36 cents per query)
Twice as Americans believe that ChatGPT is going to be bad for society than good

Growth, visits, and users

1. The company OpenAI which produced ChatGPT was founded as a non-profit in 2015 by Sam Altman, Elon Musk, and other partners

Initially incorporated solely as a non-profit, OpenAI split into two parts in 2019: a non-profit OpenAI Incorporated and a capped for-profit OpenAI Limited Partnership. In the latter, profits are limited to 100x on any investment.

Elon Musk left the company in 2018 and, in 2019 admitted he quit because he was not in favor of aspects of what the OpenAI team wanted to accomplish. Nevertheless, OpenAI achieved extraordinary success by 2022, largely due to ChatGPT.

2. ChatGPT is by far the fastest platform to achieve 1 million users (5 days)

Starting from its launch on November 30th, 2022, ChatGPT has become, by far, the fastest platform to reach 1 million users. Even powerhouse platforms that we all use today with stellar growth pale in comparison.

The previous record holder was Instagram, with 75 days to reach its first 1 million users, but ChatGPT blew that out of the water with only 5 days needed. An astonishing statistic.

3. It broke the record for the fastest-growing platform (only 2 months to reach 100 million users)

But the platform didn’t stop there. After its first month, it already had around 57 million users, and it took it another month to reach its first 100 million users.

By doing so, it broke the record for the fastest-growing platform in history.

4. Its website (openai.com) has around 1.9 billion monthly visits (over 60 million per day)

According to SimilarWeb, the website used to access ChatGPT (openai.com) recorded over 60 million visits per day in May 2023, or around 1.9 billion total monthly visits. The growth was particularly stellar in its first months, almost doubling every month during the first 4 months.

However, even though the visits are still growing, it seems that the meteoric growth has slowly tapered off, with the platform recording “only” 100 million more visits in May than in April 2023.

5. This makes openai.com the 17th most visited website in the world (May 2023)

To put these numbers into perspective, openai.com is currently the 17th most visited website in the world. This statistic is especially mind-blowing when we consider how young the company is, especially ChatGPT, which is the major traffic generator for the website.

6. It is currently available in 163 countries and territories but not in China, Russia, Ukraine, and some other countries

One of the major advantages of ChatGPT is that it’s available globally. According to the official website, it is currently available in 163 countries and territories.

Some notable countries missing from this list include China, Russia, Ukraine, and more. Italy also banned ChatGPT in 2023 due to safety and privacy concerns but lifted the ban in late April 2023.

7. U.S. users are leading in numbers, followed by India, Japan, and many developing countries

It shouldn’t be a surprise that U.S. users are in the largest numbers (13.22%) on the platform. The company, after all, originated in the U.S., the world hub of innovation and innovation adoption. India firmly holds second place with around 7% of total users, with Japan in third place.

What may come as a surprise are the 4th and 5th countries on this list: Colombia and the Philippines. This just shows how strong the power of the internet is in equalizing access to the newest technologies for all citizens of our planet.

Another plus is that ChatGPT currently understands dozens of languages, with some sources citing 26 and others as high as 95.

8. The majority of people can not tell the difference between human and ChatGPT-generated writing

A very interesting Tooltester survey found that the majority of people couldn’t recognize that an AI-generated text was actually produced by the program. While the text was completely generated by ChatGPT, only 16.7% of respondents recognized it as such.

It is interesting to note that most respondents thought it was generated by AI but edited by a human, and 36.3% thought it was completely written by a human.

Some other interesting statistics from this survey include:

71.3% of respondents would lose trust in a brand if it used AI-generated content without explicitly telling users;
80.5% think that online publishers should only use AI in online copywriting if they are explicitly going to disclose when they have done so.

We could already see this sentiment in action in early 2023 when some online publishers faced criticism for publishing AI-generated content without telling the users.

9. Google searches for ChatGPT hit a plateau in late April 2023

According to Google Trends, peak interest for ChatGPT was in April 2023, probably because of the launch of the newest version (GPT-4). It has since somewhat tapered off.

Specifications and performance

10. It is estimated that the current version (GPT-4) has between 1 and 170 trillion parameters

The growth in the number of training parameters for different versions of ChatGPT has been mind-blowing. Starting from 120 million for GPT-1, 1.5 billion for GPT-2, 175 billion for GPT-3, and trillions for GPT-4.

There is no official number, but estimates range from 1 to 175 trillion, with the most common number cited being around 100 trillion. In any case, there have been major improvements in performance across the different versions, all in relatively short time intervals.

Even though we are talking about trillions of parameters, It is estimated that training the model took just 34 days. It has been reported that OpenAI used 1,023 Nvidia A100 GPUs to train ChatGPT.

11. The free version uses GPT-3.5, and ChatGPT Plus paid version ($20/month) uses GPT-4

The free version available at chat.openai.com uses an advanced version of GPT-3 named GPT-3.5. OpenAI has also rolled out a premium paid version ($20/monthly) under the name ChatGPT Plus which uses the GPT-4 language model. It can also be accessed by using Microsoft’s browser Bing, where it’s merged with its search functionality.

Some of the advantages of GPT-4 are:

Unobstructed access to Chat GPT, even during peak times;
Quick response times;
Plugins;
Using images as input;
Maximum response of up to 25,000 words;
Priority access to new features.

OpenAI also released the Chat GPT API for developers and also declared that it would continue offering customers the free version of ChatGPT.

12. Both versions have similar results in many fields, but GPT-4 made huge leaps in law, statistics, physics, chemistry, and more

As far as real-world knowledge goes, there have been staggering advances made in many fields in the last version of the program. When compared to human examinees, GPT-4 is at the very top (80th percentile or higher) in fields such as law, math, statistics, and quantitative and verbal skills.

Some of the fields where, despite significant advances, it is still quite average include calculus, chemistry, and physics. There were just a few fields without visible improvements, such as writing, English literature, and competitive programming.

13. It is estimated that ChatGPT likely serves hundreds of millions of queries per day

Some early estimates in the early days of ChatGPT put the number of daily queries at around 10 million per day, and some recent estimates go as high as 1 billion per day. There is no official number, but it is likely to assume from the previous data points that it’s somewhere in the hundreds of millions per day.

14. The average response length by ChatGPT is estimated to be 30 words

ChatGPT is not the most talkative person in the room. At least if you don’t make it such. The average response length is reportedly around just 30 words, showing that most users use it to answer fairly simple questions.

However, the maximum response length is reportedly 25,000 words, which is about the length of a short novel.

Financials

15. OpenAI is projected to have around $200 million in revenue in 2023, and around $1 billion in 2024

According to Reuters, OpenAI brought in about $10 million in revenue in 2022, but that number is estimated to explode to $200 million in 2023 and $1 billion in 2024.

However, keep in mind that these are all estimates and that ChatGPT has very high operating costs.

16. ChatGPTs operating cost in 2023 is estimated at around $700,000 a day (36 cents per query)

Based on this estimate, the operating cost of running ChatGPT is a hefty $700,000 a day. That comes out over a quarter of a billion dollars a year, or about 36 cents per query.

The cost is so high because ChatGPT runs on Microsoft Azure supercomputers, over 3,500 of them to be precise, and with about 30,000 GPUs, according to the same estimate. There have been some rumors that Microsoft is working on its own chip to cut costs.

17. Microsoft invested $10 billion in OpenAI in early 2023, having a reported 49% stake in the company

It is clear that OpenAI is currently spending much more money than what it brings in. However, the private company already had several major outside investments, most notably from Microsoft.

The tech giant invested $1 billion in 2019, $2 billion in 2021, and finally $10 billion in 2023.

According to Fortune, Microsoft is entitled to 75% of profits until it recoups its initial investment, and thereafter, it will take 49% of OpenAI’s profits.

Microsoft has doubled down on this investment not just financially but also with integration with its Bing search engine, Edge browser, and other services. Probably the most awaited integration is one with MS Office, which is reported to take place sometime later in 2023.

18. Open AI is currently valued at $27-29 billion

The outside investments from Microsoft and others, along with the latest investment of $300 million, have put Open AI’s valuation at around $27-29 billion.

It is notoriously hard to evaluate private companies, but these numbers would put OpenAI already in or at least near the most valuable private companies in the world.

Future & Impact

19. Applications

There have already been many instances of using ChatGPT to develop useful applications, and surely many more are to come. Some examples include:

Microsoft’s Bing Search, Edge browser, and many other apps;
The Indian government is planning to integrate ChatGPT into an educational tool;
TweetGPT is a Chrome extension that writes tweets using ChatGPT;
Promptheus is a Chrome extension that takes voice input and sends it as a text prompt to ChatGPT, which returns a response;
There are multiple texts and even video (Youtube) summarizers;
ShareGPT is an app to share chatGPT prompts with other users;
ChatGPT writer is an extension that creates AI replies to emails using ChatGPT;
Engage AI reads LinkedIn posts and generates a comment on your behalf based on the content contained in the post;
GPT-4 is being used in apps such as Khan Academy, Duolingo, Stripe, and many more.

20. Competitors

It currently seems that ChatGPT is standing alone on the throne of Large Language Models (LLMs). Most notable competitors include:

OPT – The Open Pretrained Transformer is Meta’s rival to ChatGPT;
Bard – Google’s Bard is a conversational AI service powered by LaMDA, which Google says seeks to combine the world’s knowledge with the power, intelligence, and creativity of their LLMs;
Character AI – This AI chatbot can impersonate fictional characters and real historical figures, allowing users to role-play with famous people and characters.

There have also been many other LLMs developed internationally in recent times, most notably in China. However, none seem to be even close (for now) to jeopardizing ChatGPT’s reign.

21. Twice as many Americans believe that ChatGPT is going to be bad for society than good

Despite all the current applications and future potential, the average American is worried about ChatGPT’s impact on society. This survey showed that twice as many (26%) Americans think that ChatGPT will be worse for society than good (13%).

It is important to note that there is a lot of uncertainty on this topic, with many respondents answering they’re not sure (28%) or that the impact will be neutral (23%).

22. Young people and active users are more optimistic about the platform than older and less active users

The same survey shows us that, unsurprisingly, young people and more active users of ChatGPT are more optimistic about the platform and its impact than their older and less active counterparts. This is understandable because the younger population is more open to innovation and also interacts with technology more, on average.

23. ChatGPT and similar technologies will impact at least 80% of the U.S. workforce in a meaningful way

According to research from Open AI, ChatGPT and similar technologies will meaningfully impact (10% of the workload or more) the vast majority of the U.S. workforce (80% of workers).

They further state that nearly one in five workers (19%) will be substantially impacted by similar technologies (50% of their workload or more).

24. The impact will be visible across all wage levels, but with higher earners impacted more

The same study concludes that the influence will span all wage levels, with higher-income jobs potentially facing greater exposure. Among these, the most notable ones are any jobs that include repetitive work, programming, information processing, writing, etc. The jobs that will be minimally impacted are mostly lower-paying jobs without degree requirements, such as in the service sector, tourism, manufacturing, agriculture, etc.

They also concluded that Generative Pre-trained Transformers exhibit characteristics of general-purpose technologies (GPTs), suggesting that these models could have notable economic, social, and policy implications.

25. Other risks and limitations

There are many potential risks and limitations of ChatGPT and similar technologies, but some of the most common ones are:

Biased, wrong, and/or harmful responses;
Privacy and data security issues around how the user’s data is stored and used;
Threat to higher education: cheating in writing papers, exams, and more;
The current version of ChatGPT is usually guessing what the user is asking;
Creating convincing phishing messages and malicious code;
False medical advice;
Imitating others’ writing style (most commonly famous persons);

It has already been used to create an info stealer, an encryption tool, and dark web malware scripts.

Conclusion

We can clearly see that there are many pros and cons of ChatGPT and similar technologies, which is why they made such a big splash recently. While some have called for tighter regulation, it is currently unlikely that doing so will hinder the development of these technologies in a meaningful way.

Machine learning and AI tools are already a part of our present and will likely increasingly be a more and more important part of our future. It is thus important to recognize the risks and limitations of these technologies on every level between developers, users, and regulators.

Doing so, we must not lose the enormous potential which is already in clear sight. ChatGPT and similar tools are the tools of the future, and we will all probably be better off learning to use them properly and productively sooner than later.