What is ChatGPT?

This is the first of two posts on Generative AI.

Late last year (November), OpenAI released ChatGPT to the public. However, it took a few months for the internet at large to catch on and realise the potential of this new tool. The trend started to take off in December, and by February everyone was talking about it.

Well, what is ChatGPT? ChatGPT is an LLM, or Large Language Model, released by the company OpenAI, founded in 2015.

A language model is a program that has been trained on a huge amount of text data, like books, articles, and websites. It learns patterns and relationships between words, so it can predict what words or phrases are likely to come next in a sentence. We see an example of this in predictive text in cellular telephone chat applications.

The "large" part means that the model has been trained on a massive amount of data, which helps it become even better at understanding and generating human-like text. It's like having read millions of books and articles and being able to recall that information to have conversations.

When you interact with an LLM, you can ask it questions or give it prompts, and it will generate responses that make sense based on what it has learned from its training. It tries to understand the context and provide helpful or informative answers.

Of course, it's important to remember that while the model can generate text that sounds human, it doesn't truly understand the meaning behind the words. It's more like a mimic that tries to match patterns it has seen in the training data.

But why is ChatGPT called a "Generative AI"? The clue to this is the term "generative", implying creation. For convenience, let's call AIs that are not generative, "regular" AIs.

• Regular or "Narrow" AI (Artificial Intelligence): Regular AI, often referred to as "narrow" or "specific" AI, focuses on building systems that can perform specific tasks or solve specific problems. These AI systems are designed to operate within predefined boundaries and excel at specialised tasks. For example, a regular AI could be created to classify images, play chess, or process natural language. Another example is the AI that chooses your words in predictive text. Narrow or regular AI is typically trained using supervised or unsupervised learning techniques and relies on large datasets for training. It aims to optimise performance on a specific task and often follows a rule-based or algorithmic approach. Supervised learning means that a human monitors and/or controls the learning process. Unsupervised means that the computer is left alone to do its own thing.
• Generative AI: Generative AI, on the other hand, refers to a class of AI models that are capable of generating new content, such as images, text, or even music. Generative AI models, like ChatGPT, are trained to learn patterns and relationships in data and generate new samples that are similar to the training data. These models use techniques like deep learning and neural networks to generate content by capturing statistical dependencies and structures in the data they have been trained on. They can create new and original outputs based on learned patterns, and their ability to generate realistic content has improved significantly in recent years.

So, in summary, regular AI refers to AI systems designed for specific tasks, while generative AI focuses on models that can generate new content based on learned patterns. Generative AI is a subset of AI that leverages advanced techniques to produce original outputs, while regular AI covers a broader range of AI systems aimed at specific tasks.

ChatGPT has many powerful applications. A lot of educationalists and researchers, particularly in higher education, are concerned that ChatGPT can be used to write paragraphs or even essays, without crediting it, and therefore give a false impression of the competence of the alleged human author. Students might, for example, get ChatGPT to write an essay for them on a topic that they are not clear on, and submit the essay, and receive a good score, even though they themselves did not "deserve" that score. I see the opportunities presented by ChatGPT as follows:

• Information retrieval: ChatGPT can help researchers and educators find relevant information quickly. You can ask it questions or provide prompts related to your research topic, and it can provide you with relevant facts, explanations, or even suggest further resources to explore.
• Concept exploration: If you're trying to understand a complex concept, you can use ChatGPT to engage in a dialogue. You can ask it to explain the concept in simpler terms, provide examples, or even compare it to other concepts. This can help deepen your understanding and clarify any confusion.
• Writing assistance: ChatGPT can act as a writing companion for researchers and students. If you're working on a paper, thesis, or any other written piece, you can use ChatGPT to brainstorm ideas, outline your thoughts, or even get feedback on your writing. It can provide suggestions, help rephrase sentences, or offer alternative wording. For example, you can directly ask it to paraphrase sentences to avoid plagiarism problems.
• Language learning: ChatGPT can be beneficial for language learners. You can engage in conversations with it to practice your writing or speaking skills. It can correct your sentences, offer vocabulary suggestions, or provide explanations when you encounter difficulties with grammar or usage.
• Problem-solving and critical thinking: If you're facing a complex problem or trying to think through a challenging scenario, ChatGPT can act as a thinking partner. You can discuss the problem, articulate your thoughts, and receive alternative perspectives or ideas. It can help you explore different angles and encourage creative thinking.
• General knowledge exploration: ChatGPT has access to a wide range of information, so you can use it to explore various topics and expand your general knowledge. It can provide historical facts, scientific explanations, or engage in discussions on current events. It can be a handy tool for discovering new areas of interest or staying updated on different subjects.

Note that the current version available on OpenAI's website, ChatGPT 3.5, does not have internet access, and was only trained on data up until 2021, so its information is out of date. You have to subscribe to access version 4.0, which does have internet access. The updated version also has plugins which perform other tasks.

To avoid accusations of plagiarism when using ChatGPT or any language model, researchers and students should follow these guidelines:

• Understand the model's limitations: Remember that language models like ChatGPT generate text based on patterns they've learned from training data. They don't have original thoughts or expertise. Acknowledge that the generated content is not your own work but a product of the model. ChatGPT sometimes also experiences "hallucinations", that is, inappropriate or otherwise "weird" data that makes it say strange things. You therefore have to proofread everything it says because it may contain something offensive or strange.
• Use generated text as a reference: Treat the output from ChatGPT as a source of information and ideas rather than directly copying it. Take the generated text as inspiration and reshape it in your own words to demonstrate your understanding and originality.
• Combine multiple sources: Incorporate information from multiple sources, including the output from ChatGPT, to create a well-rounded and comprehensive piece of work. By including various perspectives and citing relevant sources, you show that your work is a result of thorough research rather than relying solely on the language model.
• Attribute and cite appropriately: If you directly quote or paraphrase any information obtained from ChatGPT or any other source, make sure to attribute it properly and provide citations. Follow the citation style required by your institution or academic guidelines.
• Review and revise: Carefully review the content generated by ChatGPT and ensure that it aligns with your own writing style, knowledge, and understanding. Revise and edit the text to incorporate your own insights and ideas, making it a genuine reflection of your work.
• Seek human feedback: Consult with your peers, instructors, or mentors to get feedback on your work. Discuss the ideas generated by ChatGPT and seek input on how to integrate them appropriately while maintaining academic integrity.

By following these guidelines and demonstrating your understanding and originality in your work, you can use ChatGPT as a helpful tool while avoiding plagiarism concerns. Remember, responsible and ethical use of technology is crucial in academic and research settings.

If you are skeptical of the power of LLMs and/or ChatGPT at this stage, please note: half of the above article was written by ChatGPT (mostly the bullets).