ChatGPT is a revolutionary AI-enabled chatbot that provides long-form answers to complex questions. Trained to learn what humans mean when they ask a question, ChatGPT may be the most impactful tool since the modern search engine.
1. ChatGPT can create content in seconds
ChatGPT is the latest addition to a suite of AI-enabled natural language processors and writing assistants. It was developed and trained on a large dataset of human-to-human conversation transcripts and has been specifically fine-tuned to generate responses in a chatbot context that are generally relevant and appropriate for conversation.
Users can ask it a question such as, “What are the five key messages from the work of French philosophers Gilles Deleuze and Felix Guattari?” and the system will generate a written, concise and generally correct response. It can also offer its analysis in several languages. Already a research paper written by ChatGPT - with some editing by Xiaoming Zhai of the University of Georgia - has been published.
Released in November 2022 by OpenAI Foundation, ChatGPT can generate code that can be deployed in software programs. In one case study, it generated complex code in a few minutes — a process that normally takes three to five hours. It can also code in various programming languages. In the example cited, it used Python.
ChatGPT is not connected directly to the Internet and has limited knowledge of the world and events after 2021. It will make explicit its own limitations and will not generate specific responses to questions where its data sources are either limited or non-existent. For example, asking the system to answer “What is the current weather in Toronto?” will trigger a response that because it can’t browse the Internet it can only provide general information about Toronto weather. It will also challenge incorrect responses to its results.
See more at: https://www.youtube.com/watch?v=Wj_SPwjqFxE
2. ChatGPT is one of a variety of AI-enabled writing tools
There are a growing number of natural language programs that are similar to ChatGPT and can support the writing process. According to a recent article, these include:
- Writefull – Writefull’s AI helps you write, paraphrase, copy edit and more
- ecree – A writing tutor
- TooWrite – A step-by-step guide to writing a paper, with prompts and questions
- CoAuthor – Human-AI collaborative writing dataset
Other tools that are focused more on corporate copywriting include:
- CoWrite: https://cowrite.com/en/
- ParagraphAI: https://paragraphai.com/
- Compose AI: https://www.compose.ai/
The Washington Post has been using Heliograf to write stories for its business and sports pages since 2016, and later expanded its use for political and economic stories.
3. ChatGPT is currently free to use
Because ChatGPT was developed and deployed by OpenAI, it is free to use. Users need to create a log-in to have access to the program anywhere and at anytime, although access is sometimes restricted because of the high volume of users at certain times.
On its FAQ page, OpenAI makes clear that “free use” is for a research preview phase. There is no indication either of how long this phase will last or what (if any) pricing model may apply once the phase is completed. Other services from OpenAI Foundation, such as the AI-enabled image generator DALL-E, is a pay-per-use service, with monthly free credits available for registered users.
4. ChatGPT learns from its interactions with users.
When ChatGPT is used in a chatbot application, it can continue to learn from the interactions it has with users. This can be done through “fine-tuning,” a process in which the chatbot is fed a new dataset of conversations that are specific to the task or domain the chatbot will be used in. This allows ChatGPT to learn more about the specific language and content that is relevant to the application, and to generate more relevant responses.
One of its main advantages is that it can be easily fine-tuned for specific tasks or domains, allowing it to generate relevant responses to a particular conversation or user need. It is also able to handle a wide range of conversation styles and can generate responses that are appropriate for different types of conversations, including casual, formal and technical.
It's important to note that ChatGPT is not able to learn in the same way that a human does, but it can continue to improve its performance and generate more appropriate responses through fine-tuning and exposure to new data.
5. ChatGPT (and similar systems) will have a significant impact on programming and code development
A study of reactions to ChatGPT suggests that the most significant near-term impact will be on software development, given the capability of the system to both generate code and to explain how the code it generated works, making software documentation both simpler and quicker.
6. ChatGPT has some limitations
Among the limitations identified since the release of this service are:
- Dependence on large amounts of data: ChatGPT, like many other language models, relies on large amounts of data to learn patterns and generate responses. Although this can be an advantage in some cases, it also means ChatGPT may not perform well in tasks or domains where limited data is available.
- Lack of common sense: Language models like ChatGPT do not have the ability to understand or use common sense knowledge in the same way humans do. This can lead to responses that are only sometimes appropriate or relevant to the conversation.
- Lack of empathy: ChatGPT is not able to understand or respond to emotions in the same way humans do. It may not always be able to generate appropriate responses in emotionally charged situations.
- Limited personalization: Although ChatGPT can generate responses that are appropriate for different types of conversations, it is not able to personalize its responses to individual users in the same way a human can.
- It is stand-alone: In the launch and research phase, ChatGPT is a stand-alone AI platform not connected to the Internet.
7. ChatGPT doesn’t always get it right.
Because ChatGPT sometimes delivers wrong information, StackOverflow, the go-to question-and-answer site for coders and programmers, banned ChatGPT generated answers. Users were using ChatGPT to flood the site with AI responses, but StackOverflow’s administrators confirmed the answers posted have a high rate of being incorrect.
In another example, Dr. Jacqueline Antonovich, an Assistant Professor of History at Muhlenberg College, wrote on Twitter that she put a question from her midterm essays into ChatGPT and the paper it produced “would earn an F. Probably an F - if that's possible.”
The implication of this for users is that they must develop a critical and analytic process to evaluate the outputs produced by ChatGPT and verify the suggestions it makes.
8. The arrival of ChatGPT is causing some concern for educators
An evaluation of the performance of ChatGPT on the United States Medical Licensing Exam (USMLE), which consists of three exams, was undertaken in December 2022. ChatGPT performed at or near the passing threshold for all three exams without any specialized training or reinforcement. Although some see this as a way of supporting students, others are concerned that the quality of ChatGPT responses could lead to a significant increase in cases of academic misconduct, especially for online examinations or take-home essay assignments.
There are other general concerns:
- The potential for ChatGPT to reproduce biases and stereotypes that exist in the data it was trained on. Because ChatGPT is trained on a massive amount of text data from the Internet, it is likely to contain biases and stereotypes that are prevalent in society. This can lead to ChatGPT generating responses that are offensive, discriminatory or harmful to certain groups of people. For example, if a student asks ChatGPT a question about a particular group of people, the model may generate a response that is based on stereotypes or prejudices, rather than on accurate or objective information.
- The potential for ChatGPT to reduce human interaction and personalized feedback in the learning process. By relying on a machine-learning model to generate responses, students and educators may be less likely to engage directly with each other, which can limit opportunities for meaningful dialogue and the exchange of ideas.
- The potential for ChatGPT to create a “one size fits all” approach to learning. When all students receive the same information and feedback based on the model’s output, it can limit their ability to develop their own unique perspectives and voices, and to benefit from personalized instruction or guidance from human educators.
- The risk that ChatGPT does not generate accurate or reliable information. Just like Google search results, there is no guarantee that the information ChatGPT provides will be correct or complete. This can lead to students receiving incorrect or misleading information, which can have negative consequences for their learning and understanding.
9. Measures can be taken to reduce the risk of academic misconduct
A writer asked ChatGPT to respond to the question, “What can be done to reduce the risk of ChatGPT being used by students to cheat?” ChatGPT responded with the following:
“To avoid ChatGPT being used for cheating, there are a few different steps that educators and institutions could take. For example, they could:
- Educate students and educators on the ethical use of AI technology, including ChatGPT, in academic settings.
- Develop guidelines and policies for the use of ChatGPT in academic work, and make sure that students and educators are aware of and follow these guidelines.
- Monitor the use of ChatGPT in academic settings and take appropriate action if it is used for cheating or other unethical purposes.
- Use ChatGPT in ways that support learning and academic achievement rather than as a replacement for traditional forms of assessment. For example, ChatGPT could provide personalized feedback and support to students rather than as a tool for generating entire papers or exams.
- Incorporate critical thinking and ethical reasoning into the curriculum, to help students develop the skills and habits necessary to use ChatGPT and other AI technology responsibly.”
10. ChatGPT is causing a stir
Since its launch at the end of November 2022, ChatGPT has caused a stir.
A day hardly goes by without a report or opinion piece in popular media about ChatGPT. Twitter is ablaze with comments about its threat, inaccuracies or promises. Even Google reacted with an internal declaration of a “code red” challenge to its search engine. The company is set to announce a series of AI-related developments in May 2023.
What’s clear at this point is we are only seeing the beginning of the impact of ChatGPT on higher education and our society more generally.