ChatGPT Stats

ChatGPT was able to reach their first million users in just 5 days, making them one of the most successful and fastest-growing platforms on the market. Facebook took 10 months, Instagram only 2 months, Spotify 5 months, and Netflix a whopping 3.5 years to get to the same milestone.

March 2023: OpenAI has released GPT-4. It can understand both images and words that people say. GPT-4 is better than ChatGPT at every job it does.

Time to first 1 million users

• ChatGPT – 5 days
• Facebook – 10 months
• Instagram – 2 months
• Spotify – 5 months
• Netflix – 3.5 years

• ChatGPT got one million people using it in only 5 days after it started. This was announced by Greg Brockman, who is a co-founder of Open AI.
• ChatGPT has become the fastest-growing app in the world, recording 100 million users in just two months of its launch.
• January of 2023 marked a monumental moment for the ChatGPT website, as they experienced 590 million monthly visits and welcomed their 100th million user. (The Guardian)
• Similarweb says that the ChatGPT website has been visited 1 billion times and 25 million people go there every day.

Monthly Total Visits

• Since launching ChatGPT, OpenAI’s page views skyrocketed from 18.3 million in November 2022 to 304 million in December 2022 – a whopping 1561% increase!
• By January 2023, the website had earned its place among the world’s top 50 most visited sites due to their incredible success with ChatGPT.

Company Background and Investments

• OpenAI has developed ChatGPT and Dall-E, the latter which is an image generation AI.
• Co-founded by Elon Musk, Open AI is headquartered in the Pioneer Building within San Francisco’s Mission District.
• Open AI’s valuation skyrocketed to $29 billion after ChatGPT was released.
• Microsoft has already poured in $1 Billion and promised to commit an additional investment of $10 billion into Open AI, further demonstrating their belief that this is a bold endeavor worth pursuing.

Cost to Run and Revenue Model

• According to Sam Altman, CEO of OpenAI, the fee for a single chat with ChatGPT is probably lower than 10 cents.
• According to Tom Goldstein, an AI ML Professor at Maryland University, the daily cost of powering ChatGPT is around $100,000 and its monthly expenditure stands at a whopping 3 million USD. His calculations are based on Azure Cloud costs (which is where ChatGPT runs). He further added that the model has 175 billion parameters while it receives 10 million queries per day (as seen on his Twitter page).
• ChatGPT is free to use during its research preview. You don’t have to pay anything.
• ChatGPT Plus is a new tool you can use. To try it out, you can pay $20 for a subscription.
• For ChatGPT Plus users, Open AI has introduced “Turbo” mode to ensure faster response times and greater satisfaction.
• ChatGPT could make $200 million dollars by the end of 2023 and $1 billion dollars by the end of 2024. (Reuters).

Model Training

ChatGPT is a model designed to perfection, based on the sophisticated GPT-3.5 (text-DaVinci-003) technology.

Text-DaVinci-003 takes the incredible potential of InstructGPT – an earlier open AI model that surpasses GPT 3 in its capacities to comprehend and respond to human requests – a step further.

Bing AI is now running on an advanced form of the ChatGPT model known as Prometheus.

What was the dataset size for training ChatGPT?

• ChatGPT’s learning was based on an accumulation of 300 billion words, a 570 GB dataset that included web-crawled content, books and Wikipedia articles. (Analytics India Magazine)
• ChatGPT has been developed using Reinforcement Learning from Human Feedback (RLHF), a revolutionary Large Language Model (LLM) that is trained by human feedback.
• RLHF is an approach that utilizes human AI trainers to provide feedback during the initial training process of ChatGPT, which helps it find ways to enhance its responses.

Limitations of ChatGPT

While ChatGPT can help you get creative with tasks like crafting movie scripts, launching your own online course and generating AI-generated stories, it is important to remember that there are certain limitations.

• ChatGPT Cannot Access the Internet. As an AI language model, ChatGPT cannot access the internet or external links; it can only respond to questions and supply information based on its training material.
• ChatGPT’s knowledge is limited to its training data which has the cutoff year of 2021.
• ChatGPT can be unreliable at times, often appearing to exude confidence in its answers even though they may not always be accurate. It is important for users to verify the information before relying on it.
• Due to the potential for errors in all LLM models, Google has been reluctant to roll out Lamda – its own ChatGPT-like conversational AI – to a global audience.

Real-World Impact

• ChatGPT has proven itself time and time again, having bested the United States Medical licensing exam, law school exams and Wharton’s MBA test.
• Unfortunately however, it is also being used for some malicious activities like producing hacking scripts or phishing emails.

CyberSecurity Threat

• ChatGPT is a worrying trend as it can be employed to craft convincing phishing emails and generate malicious code, even among users with little-to-no programming abilities. (Source: TechCrunch)
• For instance, ChatGPT can compose a medical article with the panache of a physician or create legal documents that give off an aura of authenticity (even though they were not written by lawyers).
• ChatGPT has the remarkable ability to mimic the writing style of famous celebrities and authors.
• This technology can provide potential hackers with the capability to generate working code for malicious operations, like ransomware attacks and other exploits, through simple English prompts. By having access to this feature, these criminals will no longer be hindered by their lack of programming expertise!
• According to Blackberry Research, the majority of IT professionals anticipate that a successful cyberattack will be orchestrated with ChatGPT by year’s end.
• Through ChatGPT, malware writers have developed an info stealer, encryption tool, and a variety of other malicious scripts for the dark web.

The Peril of Higher Education: A Look at the Current Challenges Facing Academic Institutions.

• ChatGPT makes it easy to craft compelling essays and research papers. While tools may exist that can detect AI-generated content, they are not 100% reliable
• Consequently, AI can be exploited in academia: students may submit assignments crafted by artificial intelligence and still pass without actually gaining any understanding.
• In the not-so-distant future, Open AI has declared that they plan to “brand” content created with ChatGPT in order to discourage academic dishonesty and spam.
• In order to ensure accountability, Springer (a renowned research publisher) has requested that researchers omit ChatGPT from their list of authors. Instead, it is recommended for them to mention the use of this revolutionary bot in either the acknowledgments or methods section.

Other Reactions to ChatGPT

• Stack Overflow, a popular coding question and answer platform, has banned the use of ChatGPT because it can be used to give seemingly legitimate but actually useless or wrong answers, making it harder to find relevant answers that work.
• Amazon has warned its employees not to share sensitive information with ChatGPT.
• Artificial Intelligence tokens such as DeepBrain Chain and Numeraire witnessed a sharp rise in prices after the debut of ChatGPT. According to CoinGecko, the prices of AI tokens increased by up to 77%.

Alternatives & Competitors of ChatGPT

Google Vs ChatGPT

Google’s CEO, Sundar Pichai recently unveiled a new AI search assistant named Bard AI; however, the demo response was unfortunately inaccurate and resulted in an immense financial loss of over $100 billion to the company.

Bard’s answers reveal how the James Webb Space Telescope (JWST) has been instrumental in taking the very first pictures of exoplanets, or planets existing beyond our Solar System.

This was incorrect… the first photos of exoplanets weren’t taken by NASA – they were actually captured in 2004 via the European Southern Observatory’s VLT and then confirmed through NASA.

ChatGPT has been unrivaled in its ability to provide direct answers to user questions and the capacity for follow-up questions as well, which Google still lags behind on.

Introducing Google Lamda, a groundbreaking language model from the innovative minds of Google’s AI research. First demoed in 2020, LAMDA can converse with users on any subject matter and even craft jokes or poems – just like ChatGPT!

With its cutting-edge technology, you’ll be sure to get your hands on the latest and greatest advancements in natural language processing.

• Google was not sure about using LAMDA because Artificial Intelligence (AI) models can make mistakes. Sometimes ChatGPT, Bard and Bing may seem sure of themselves, but they could give wrong answers.
• Google PaLM is another language model presented by the tech giant. It has been trained on an astounding 540 billion parameters, dwarfing ChatGPT which only boasts 175 Billion!

Microsoft has been a step ahead of Google in launching an AI search assistant. Several users have already been able to preview Bing AI “the co-pilot for the web”. See examples of Bing AI’s responses.

Which model is behind Bing AI?

OpenAI made a new kind of language model. It is better than ChatGPT and GPT-3.5. It works faster, is more accurate, and can do more things. This has since been confirmed to be GPT-4.

Youchat

Unlike ChatGPT, Youchat is connected to the internet and can consequently engage in conversations with its users – just like ChatGPT.

Wordtune

ChatGPT has a pitfall of presenting statistics and definitions without attributing them to any source, but thankfully Wordtune is here with its conversational model that offers citations.