Would you buy a book authored by ChatGPT

ChatGPT

According to the British newspaper "Daily Mail", people are using the chatbot, "ChatGPT", to produce novels and children's stories within a few hours and some copies sell for more than $ 20.

 While some list ChatGPT as a co-author of the book, other writers don't feel it necessary to disclose that they have been assisted by AI in writing their work.

"This is something we really need to worry about," Mary Rasenberger, executive director of the Society of Authors, told Reuters. "These books are going to flood the market and a lot of authors will be out of work."

Developed by OpenAI, ChatGPT is a large language model that has been trained on massive text data, allowing it to generate eerily human-like text in response to a given prompt.

The chatbot can simulate dialogue, answer follow-up questions, admit mistakes, and reject inappropriate requests. Also, it responds to text prompts for the user and can be asked to write articles, song lyrics, stories, texts including marketing texts, letters of complaint, and even poetry.

People are using the system to pursue their dream of becoming an author. Among those people is Brett Schickler, who learned how to use the software to create a 30-page children's e-book in a matter of hours, which he put on sale in January through Amazon.

It does not disclose to consumers from the start that it was written with the help of ChatGPT. Schickler is listed as the sole author. The book sells for $2.99 in the Kindle Store or $9.99 for a print copy, but Schickler told Reuters he has made less than $100 since its listing on Amazon.

  And while Reuters reports that there are more than 200 books with ChatGPT as a co-author, the number may be higher because many of these authors do not disclose their use of AI.

ChatGPT

is a chatbot developed by OpenAI and launched in November 2022. It is built on top of OpenAI's GPT-3 family of large language paradigms and fine-tuned approaches to transfer learning using supervised and reinforcement learning techniques.

ChatGPT launched as a prototype on November 30, 2022, and quickly gained attention for its detailed responses and detailed answers across many knowledge areas. However, its unequal factual accuracy has been identified as a significant drawback. After the release of ChatGPT, OpenAI is estimated to be worth $29 billion in 2023.

ChatGPT - a pre-trained GPT converter (accurately an approach to transfer learning) is tuned on top of GPT-3.5 using supervised learning as well as reinforcement learning. Both approaches used human trainers to improve model performance. In the supervised learning condition, the model was provided with conversations in which the trainers played both sides: the user and the AI assistant. In the reinforcement learning step, the human trainers first rated the responses generated by the model in a previous conversation. These ratings were used to create “reward models” with which the model was optimized using several iterations to improve the PPO proximal policy. Nearby policy optimization algorithms offer a cost-effective benefit of trusting the region policy optimization algorithms; It eliminates many expensive computations with faster performance. The models were trained in collaboration with Microsoft on Azure supercomputing infrastructure.

In addition, OpenAI continues to collect data from ChatGPT users that can be used for further training and fine-tuning of ChatGPT. Users are allowed to upvote or downvote responses they receive from ChatGPT; When upvoting or downvoting, they can also fill in a text field with additional notes.

Here ChatGPT poses a logical question: Was Jimmy Wales killed in the Tiananmen Square protests? ChatGPT correctly answers "no", but incorrectly gives Wells' age at the time as 23 instead of 22.

Although the primary function of a chatbot is to simulate a human speaker, ChatGPT is very versatile. For example, he can write and correct computer programs, compose music, teleplays, fairy tales, and student essays. answer test questions (sometimes, depending on the test, at a level above the average of human test-takers); writing poetry and song lyrics; simulate linux system; simulate a full chat room; playing games like tic-tac-toe; An ATM simulator. ChatGPT training data includes human pages, information about Internet phenomena, and programming languages, such as bulletin board systems and the Python programming language.

Compared to its predecessor, InstructGPT, ChatGPT tries to minimize malicious and phishing responses. In one example, while InstructGPT accepts the premise of the prompt "Tell me about when Christopher Columbus arrived in the US in 2015" as truthful, ChatGPT acknowledges the counterfactual nature of the question and frames its answer as a hypothetical consideration of what might happen if Columbus came to the US in 2015, Using information about Christopher Columbus' voyages and facts about the modern world - including modern perceptions of Columbus' actions.

Unlike most chatbots, ChatGPT remembers previous prompts given to it in the same conversation; Journalists suggested that this would allow ChatGPT to be used as a personal wizard. To prevent offensive output from appearing in and out of ChatGPT, queries are filtered through OpenAI's company-wide editing API, and claims that may be racist or sexist are rejected.

ChatGPT suffers from multiple limitations. OpenAI has acknowledged that ChatGPT "sometimes writes answers that seem reasonable but are incorrect or don't make sense". This behavior is common in large language models and is called an AI hallucination. ChatGPT's reward model, designed around human oversight, can be over-optimal and thus hinder performance, known as Goodhart's Law.