ChatGTP was developed by US company OpenAI, which specializes in artificial intelligence tools for human use. OpenAI describes ChatGPT as a sibling model to another of its tools, InstructGPT, which is trained to follow human prompts and provide detailed responses to queries.
Owing to its chat box format, which opens up a dialogue between human and AI, ChatGPT allows users to request information on highly specific topics—from quantum computing and processing code to the mechanics of the human heart or the principles of flower arrangement. In this way it apes the brain-dump, question-and-answer format of Google search, although instead of producing multiple answers on a search engine results page (SERP), ChatGPT acts as the sole authority delivering responses.
The human user can refine the search query in response to what ChatGPT produces or ask follow-up questions to home in on specific information. This information, according to its boosters, can then be incorporated into CVs, assignments, work tasks, presentations and reports. ChatGPT can also pen songs, write movie scripts, and compose poems on topics of the user’s choosing. It can add levity and humor when requested, or adopt a formal tone.
The tool was released to the general public last year as a way of improving upon its functionality through feedback, and since then OpenAI claims to have received information from millions of free users across the globe on its strengths and weaknesses, which it intends to incorporate into its updates and modifications. It is estimated that more than a million people accessed ChatGPT in its first five days alone and two months after its launch it had attracted 100 million monthly active users.
There are two versions of the tool at present: the free iteration, which tends to be more prone to errors, and the paid version, known as ChatGPT Plus, which became available to users worldwide, in February. It costs $20 US a month.
OpenAI claims the three main benefits of the paid tier are: general access to ChatGPT, even during peak times; faster response times; and priority access to new features and improvements.
How Does ChatGPT Work?
ChatGPT is what is known as a large language model tool, and it is trained in vast reams of information—articles, textbooks, the internet—and then taught to replicate human speech and modes of information delivery. It is known as GPT-4, and is a successor to the less advanced GPT-3.
What sets ChatGPT apart from its predecessor is its subtle generative AI capacities for human-like expression and interaction, which is continuously refined in a process known as Reinforcement Learning from Human Feedback (RLHF).
OpenAI’s human AI trainers create conversations in which they play both sides—the user and an AI assistant—for greater authenticity of responses.
Trainers are also provided with access to “model-written suggestions” to compose their responses and help train the tool in speech patterns, written expression, translation, text completion and similar tasks.
Because it provides answers in a text box, it acts as a search engine, a chat bot or a virtual assistant in one.
