What is ChatGPT?
ChatGPT is a large language model developed by OpenAI. It has been developed, using a massive dataset of internet text and is designed to generate human-like text in response to questions set to it. It can be used for a variety of natural language processing tasks such as language translation, summarisation, question answering and more.
At least, that’s what ChatGPT says about ChatGPT. For the everyday user, it provides answers to questions that are written the way a human would write them. There is a wide range of aplication to how ChatGPT can be used, but it still has limitations – so it won’t become self-aware and attack humanity, in a Skynet/Terminator-style attack.
How does it work?
ChatGPT works by using a type of neural network called a Transformer (sorry, the dayjob nerd crept in), which is trained on a large dataset of text. The model learns to predict the next word in a sentence based on the context of the words that come before it.
When given a question, ChatGPT uses the context of the prompt to generate a response. The model generates a probability distribution over the vocabulary for the next word, and then selects the word with the highest probability as its prediction. This process is repeated multiple times to generate a complete response.
The model is also fine-tuned on specific task datasets to improve its performance on those specific tasks like question answering, chatbots, etc.
In summary, ChatGPT is a digital neural network that generates a human-like text response based on the context of a given question, by predicting the next word in a sentence with the highest probability using a transformer network.
Benefits & Challenges Of Chat GPT in Education
- Personalised learning: ChatGPT can be used to generate personalised learning materials and exercises for students based on their individual needs and abilities.
- Tutoring assistance: ChatGPT can be used as an AI-powered tutor, providing students with immediate feedback and assistance with their coursework.
- Improving writing skills: ChatGPT can be used to help students improve their writing skills by providing suggestions for grammar, style, and organisation.
- Language learning: ChatGPT can be used to help students learn a new language by providing real-time translations and assisting with grammar and vocabulary.
- Research assistance: ChatGPT can assist students in their research by providing information on a specific topic and suggesting relevant sources.
- Time-saving: ChatGPT can be used to automate routine tasks, such as grading and providing feedback on student work, allowing educators more time to focus on other tasks that are pupil-focused.
- Gamification: ChatGPT can be used to gamify education, making learning more engaging and interactive for students. This has proven benefits in research and in ICT settings like Hour of Code, or coding in Minecraft (as one example)
- Flexibility: ChatGPT can be accessed from anywhere, at any time, making it easy for students to continue learning even when they are not in a traditional classroom setting. The major issue at the minute is that the usage of Chat GPT is so high that it sometimes happens that the servers.
- Lack of common sense: ChatGPT is trained on a massive dataset of internet text, but it does not have a built-in understanding of the world and the context in which the text is used. This can lead to nonsensical or irrelevant responses when the model is prompted with certain questions or situations because it cannot define context.
- Biases in the training data: The training data for ChatGPT is sourced from the internet, which can include biases and stereotypes. This can lead to the model generating text that reflects these biases, which can be harmful or offensive.
- Difficulty in controlling the level of creativity: Since the model is trained to generate text based on patterns it has seen in the training data, it can sometimes generate text that is too similar to the training data, or text that is too creative and diverges from the prompt.
- Computational and Memory requirement: GPT models are large and require a lot of computational power and memory to fine-tune and run. This can make it difficult to use them on devices with limited resources.
- Limited ability to handle more complex tasks: While GPT can handle a wide range of natural language processing tasks, it may not be as effective for more complex tasks that require understanding of the underlying meaning, context or significance of the text.
Why are Educators Getting Interested?
ChatGPT, more than other language models before it, poses several potential issues to teachers, some of which include:
- Plagiarism: As ChatGPT is able to generate human-like text, the potential exists that students may use it to plagiarise assignments and homework. This tool makes it more difficult for teachers to identify and prevent plagiarism.
- Increased difficulty in evaluating student understanding: Since ChatGPT can generate text that is highly similar to text written by your or I, it may be difficult for teachers to determine whether a student truly understands the material or if they are simply using the model to generate an answer.
- Lack of critical thinking: ChatGPT can generate text that is highly similar to text written by you or I, which may lead to students using the model to simply repeat information without understanding or thinking critically about it.
- Misrepresentation of student’s work: Students may use the model to generate text and present it as their own work. This can misrepresent the student’s abilities, and can also lead to cheating and academic dishonesty. This would ultimately have greatest consequence with exams (at 16, 18 or in Higher Education) where the student can be stripped of their grade if academic dishonesty can be proven. A costly mistake, but none more so than at university, were fees and loans are involved.
- Limited ability to handle more complex tasks: while ChatGPT can handle a wide range of natural language processing tasks, it may not be as effective for more complex tasks that require understanding of the underlying meaning and context of the text. Extrapolating signficance or presenting evaluative comments is not a strong aspect to what ChatGPT can currently do.
Overall, ChatGPT is a powerful tool for natural language processing, but it also poses several issues for teachers. Teachers should be aware of the potential for plagiarism and cheating, and take steps to prevent it. Additionally, teachers should be mindful that the model may not be suitable for certain complex tasks. Pupils more than likely do not posess the intellectual capability to recognise its’ shortcomings.
Is There Anything For Educators to Really Fear?
Yes and no.
There are the points discussed above. There’s no getting away from those potential uses and abuses.
It hit the news yesterday that ChatGPT scored between a B and a B- (minus) in an MBA exam set by a Wharton professor in the University of Pennsylvania. The professor commented that the bot is “remarkably good at modifying its answers in response to human hints,” and experts who work in both the fields of artificial intelligence and education have recognised that bots like ChatGPT could have a detrimental impact on education in the future. Particularly as the AI platform becomes more advanced.
I performed an experiment in my school this week. I asked some of the Subject Leaders to provide me with an exam question (their choice: exam board, year and question) that would require pupils to write at least half a page in response. Here are the results:
- Maths: would know it wasn’t a pupil response; does not provide answer showing method, method provided is not used in school. All exam questions are equations
- French: didn’t use language taught in class, answer doesn’t follow mark scheme, too short. Teacher guessed it was an AI-provided answer
- Technology: quite good, but a teacher could easily show how to provide a better answer.
- Business Studies: knew it was an AI-provided answer, but impressed by movements of technology and what it would mean for teachers.
- Digital Technology: not good at plan drawings (e.g. network illustrations) or at discussing the significance of an IT tool in its application to business use.
The last point came out fairly consistently – AI can give the facts, provided the question is specific and detailed enough to provide an answer to all aspects of the question, but it cannot explain significance. Of course, this point should be somewhat obvious – what is effectively a computer programme (albeit a very impressive computer programme) will deal in facts. Argument, persuasion and evaluation is nuance that is offered by an individual. Computers cannot offer that just yet.
This holds our greatest continual strength against the increasing abilities of technology. Teachers can provide pupils with nuance, skill, desterity of thought, persuasion, and meaningful feedback. It is our job to help them negotiate this world, prepare them for the workplace in however it presents and give them the necessary skills and tools to succeed. ChatGPT is now one of those tools.
I’m not going to repeat anything I’ve said elsewhere in this post. You can see the strengths, benefits, pitfalls, and challenges this will present in the classroom.
My advice is to go and play with it. See what it can do. So far I’ve used it to write module notes, lesson plans and attempt to answer exam questions. It is limited – but that’s were my expertise comes in to play. I will continue to use it to produce learning material – answers to questions like:
- What are the different types fo file compression?
- What are the main threats to data information and systems
- What is application software?
- How do the Features of Connection Types Meet the Needs of Individuals and Organisations?
- What are the Implications of Selecting and Using Different Connection Types?
But ChatGPT cannot answer these questions just yet:
- How do we choose an IT system?
- How do we resolve an ethical or moral issue?
That last question, you can get the following answer:
There are several steps that can be taken to resolve an ethical issue:
- Identify the problem: Clearly define the issue and gather all relevant information.
- Identify the stakeholders: Determine who is affected by the issue and gather their perspectives.
- Identify the options: Generate a list of potential solutions to the problem.
- Evaluate the options: Consider the potential consequences of each option and determine which one aligns best with the organization’s values and ethical principles.
- Choose the best option: Select the option that is deemed the most ethical and implement it.
- Monitor and review: Continuously monitor the situation to ensure that the chosen solution is effective and make adjustments as necessary. It’s important to consult with diverse groups of people and experts to ensure that a wide range of perspectives are considered and ethical decision is made.
I’m not happy with it because it is focused solely on what is best for the person in any environment – whatever is most expedient and beneficial to you. Morals are meant to be rooted in deeper soil than that. The answer above leads on a pathway to selfishness, narcissism and patholigical self-interest. But a teacher would know this.
So turn up to work tomorrow. Your job hasn’t been taken, and a terminator that looks like Arnie won’t be there policing you.