What is GPT?
Before explaining ChatGPT-3 and ChatGPT statistics, it’s better to investigate the GPT language model.
GPT, which stands for Generative Pre-trained Transformer, is a generative language model and a training process for natural language processing tasks.
OpenAI Company created GPT-1, GPT-2, and GPT-3 at the times you see in the list below:
- GPT-1: it was released in June 11, 2018
- GPT-2: it was released in February 14, 2019
- GPT-3: it was released in June 11, 2020
First, GPT-1 was introduced in 2018, which was trained on a huge BooksCorpus dataset. It could learn wide range dependencies and obtain massive data on a diverse corpus of contiguous text and long stretches. (360 Digi TMG)
Then GPT-2 was released by OpenAI Company in 2019. Compared with GPT-1, it used a bigger dataset with more parameters to make a richer language model. As you can see in the picture below, the number of GPT-2 parameters increased to 1.5 billion, which was only 150 million in GPT-1!
GPT-3 introduced by OpenAI in 2020, which was stronger and more rapid than previous GPTs with more parameters (around 116.5 times larger than GPT-2!) and decoder layers. The table below shows the differences between these three GPT models based on the number of parameters, decoder layers, context token size, hidden layers, and batch size.
Now it’s time to start talking about ChatGPT (Generative Pre-trained Transformer), which can act as a tool for creating text. We start from the ChatGPT definition and end to a complete list of ChatGPT statistics.
What is ChatGPT? All There Is to Know About ChatGPT!
To put it simply, ChatGPT is an assistant and a tool that generates texts via the use of GPT language models. The breakthrough in AI technology created by American OpenAI Company!
In other words, ChatGPT simulates human beings to make conversation with users and chat with them. The users ask their questions and ChatGPT responds to them fast!
ChatGPT uses Reinforcement Learning with Human Feedback (RLHF), adjusted for conversational dialogue.
ChatGPTs are produced based on huge amounts of data so they aim to be human-like and the more time passes, the richer they become.
ChatGPT is as good as the quality of the data it has been trained about. The more data you give to ChatGPT, the more knowledgeable and functional it will be.
What do you think about Chat GPT-3? Is it amazing or frightening?! Or maybe both of them!
Let’s look at the ChatGPT statistics step by step, then you can shape it better in your mind.
1. What is ChatGPT?
As stated above, ChatGPT (Artificial intelligence Chatbot) is an assistant or a tool that simulates human beings to chat with users and customers. Users ask questions and ChatGPT answers them instantly based on the vast amount of trained data with the help of GPT language models.
2. What does ChatGPT do?
- chatting with the users in a conversational manner
- remembering what the user said before in the conversation
- replying users’ questions like a human being
- producing well-formed and text based content (story, social media posts, essay, poetry, lyrics, etc.)
- understanding, writing, and debugging computer codes
- being able to take tests
- text summarization
- managing and manipulating data in bulk (producing data in tables and add indexes)
3. When ChatGPT was created?
ChatGBT was initially released on November 30, 2022.
4. Who Created ChatGPT?
OpenAI, the artificial intelligence (AI) company established ChatGPT in 2022.
5. What is OpenAI Company?
OpenAI is a research lab with two main companies. For-profit corporation (OpenAI LP) and its parent company, the nonprofit OpenAI Company.
6. What is OpenAI known for?
OpenAI Company tries to enhance friendly AI “to ensure that artificial general intelligence benefits all of humanity” according to the OpenAI website.
7. When was OpenAI Company founded?
OpenAI Company was established on December 11, 2015.
8. OpenAI website link
9. Who created OpenAI?
Elon Musk, Greg Brockman, Ilya Sutskever, Sam Altman, and Wojciech Zaremba founded OpenAI Company in 2015.
10. OpenAI headquarters
San Francisco, USA
11. Is OpenAI owned by Microsoft?
No, Microsoft is one of the partners of OpenAI Company. In 2019, Microsoft invested in OpenAI Company.
12. How much did Microsoft invest in OpenAI?
Microsoft invested US $1 billion in OpenAI Company in 2019.
13. Who is OpenAI CEO in 2023?
Sam Altman, the former president of Y Combinator, is the CEO of OpenAI Company. He is an American entrepreneur, investor, programmer, and blogger.
14. Who is OpenAI chairman in 2023?
Greg Brockman is OpenAI chairman and president.
15. Who is the top researcher of OpenAI in 2023?
Ilya Sutskever is chief scientist and top researcher of OpenAI Company.
16. ChatGPT positive impact on OpenAI’s value
ChatGPT improved OpenAI’s value to USD $29. Now OpenAI is one of the most valuable startups in the USA.
17. How many users does ChatGPT have?
In Q4 2022, ChatGPT’s users crossed over 1 million, just after 5 days from its establishment!
18. How many users will ChatGPT have by the end of 2023 based on predictions?
It is predicted that ChatGPT will have approximately 1 billion users by the end of 2023!
19. How to use ChatGPT?
20. ChatGPT example
The picture below shows the conversational dialogue between CUFinder as the user and ChatGPT. CUFinder asked ChatGPT to “write a paragraph about lead generation”. As you can see, Chat GPT-3 wrote over 8 lines about lead generation instantly based on its trained data.
21. Chat GPT-3 dataset size
- 175 BN parameters
- 300 billion words
- 570 GB Data
22. ChatGPT infrastructure
23. What is Chat GPT based on?
“ChatGPT is fine-tuned from a model in the GPT-3.5 series, which finished training in early 2022.” (OpenAI)
24. When was Chat GPT-3 training stopped?
ChatGPT training stopped in 2021 so it cannot fetch the data after 2021.
25. How was ChatGPT trained?
ChatGPT was trained by applying Reinforcement Learning from Human Feedback (RLHF).
26. What is human reinforcement learning?
“Using methods from reinforcement learning to directly optimize a language model with human feedback. RLHF has enabled language models to begin to align a model trained on a general corpus of text data to that of complex human values” as mentioned in Hugging Face.
27. How many days did it take to train GPT-3?
According to OpenAI, ChatGPT-3 was trained in 4 days using 2048 TPU v3 chips.
28. Does ChatGPT have access to the internet?
ChatGPT has no access to the internet. It is restricted to its training information.
29. What are the limitations of ChatGPT? A complete list of ChatGPT insufficiencies
- May give wrong data; the data should be verified before using
- Every search provides only one result
- Has limited data of the world as its training was stopped on 2021
- Usually overuses certain phrases
- May generate destructive and wrong instructions
- Cannot be used for shopping and product searching unlike Google
- There is no source for the given data
- ChatGPT-3 is sensitive to the phrases entered as input. It means that you may ask a question and it cannot answer but if you rephrase, it may reply easily!
- May produce biased data
- When it doesn’t have the response, it shows irrelevant data instead of saying that there is no data for your query
- Low-speed of data providing compare with Google search engine
- Has no API; only supports a text interface
- No access to the internet
- After searching, it doesn’t provide suggestions related to your search
- Only produces text output; no image or video unlike Google
30. ChatGPT price
31. How many GPUs does Chat GPT use?
It cannot be fitted on a single GPU. A single ChatGPT query uses at least 8 GPUs.
32. ChatGPT’s revenue prediction
OpenAI Company predicts that ChatGPT revenue will be around US $ 200 million by the end of 2023, and surprisingly, it will jump to US $1 billion by the end of 2024!
33. Can ChatGPT replace Google?
ChatGPT cannot replace Google at least in near future. Google is a search engine while ChatGPT is a language model. They are different from each other.
ChatGPT understands the meaning of data and then responds to the questions. It should find out the meaning of the message and the user’s intent to be able to reply.
However, Google doesn’t produce any data on its own! It is just a source of data added by web users and it provides results based on keywords.
34. Which countries do not support ChatGPT?
ChatGPT and GPT-3 are not available in some important countries and regions including:
35. Why ChatGPT is not available in Ukraine?
ChatGPT and GPT-3 API do not work in Ukraine as OpenAI placed Ukraine in the list of terrorists, the same as the decision for Iran and Russia!
36. Can Chat GPT-3 understand the users’ intent?
Previously, if the user asked an ambiguous and confusing question, ChatGPT couldn’t understand the meaning and hence could not reply but Chat GPT-3 can find out the users’ intent even if their request is unclear.
37. Can ChatGPT replace journalists in the future?
According to The Guardian, “The ability to generate human-like written text has prompted suggestions that the technology could replace journalists”. However, we need to wait for the future!
In The Atlantic’s “Breakthroughs of the Year”, Derek Thompson wrote “As a journalist, I’m most excited by their potential to help me plow through complicated bits of writing”.
He continued “When I feel myself stuck on a definition or transition sentence, I could give a prompt to the AI, which, having feasted on a banquet of online writing, would generate sentences that I could edit back into my own voice.”
38. Is Chatbot a robot? Are robots coming to do our jobs?!
What was expressed by Paul Krugman about Chatbot and Chat GPT-3 in December 2022 was notable: “… in the long run, we are all dead, and even before that, some of us may find ourselves either unemployed or earning far less than we expected, given our expensive educations.”
39. How has ChatGPT impacted cryptocurrencies?
ChatGPT-3 caused AI token prices to jump to over 76.7% when GPT-3 started!
However, not all tokens experienced a jump. Look at the list below showing the tokens’ improvements and declines when GPT-3 launched (all in the same period, from 30 November 2022 to 6 December 2022):
- DeepBrain Chain (DBC): +76.7%! (the most increase)
- Numeraire (NMR): +54.5%
- Fetch.ai (FET): +10.1%
- SingularityNET (AGIX): +7.0%
- Alethea Artificial Liquid Intelligence (ALI): -3.2%