Day 1: AI tools and general terms
Monday 29 April
Hello and welcome to day 1 of 6 days of AI. Today’s task is relatively simple.
We would like you to become more familiar with AI tools, general AI terms and the various software available.
What is generative AI?
Generative Artificial intelligence (Gen AI) is a series of tools that use the technology of artificial intelligence to enable the generation of content, data, or certain outputs. This could include text, image, video, or audio generation.
Generative AI is distinct from General AI, which is the work to create an intelligent system that could think and act like a human mind. It is important to keep this distinction in mind considering AI tools.
AI Chatbots
AI Chatbots, also known as LLM (Large Language Models), are tools designed to mimic human conversation. These tools allow the user to pose questions to the chatbot in plain, conversational language, and then receive a response. Follow up questions can then be asked to refine or expand the original answer.
Essentially, users input texts prompts that will create AI-generated content that can include text, images, or videos.
Copilot
There are various AI chatbots available to us, but here at Sussex we recommend Microsoft Copilot, you can find more information regarding Copilot from the Educational Enhancement webpages. You can log in to Copilot using your University of Sussex credentials. Make sure you select the second option, ‘Sign in with work or school account’.
You will then be presented with the usual Microsoft sign in window. Put in your shortened email address (ab123@sussex.ac.uk) and enter your password.
If you have logged in correctly, on the top right-hand side of your screen you should see the green icon ‘Protected’, as shown in the below.
Why do we recommend Copilot?
Microsoft Copilot is a chatbot developed by Microsoft and was launched on February 7, 2023. It is a conversational AI chatbot that uses language processing techniques to generate humanlike interactions in dialogue form. It has the capability to answer queries and craft a wide array of textual content, encompassing articles, posts for social media, essays, programming code, and email messages.
If you're logged into Microsoft Copilot with your University of Sussex details, what goes in and what comes out is not saved or shared and your data is not used outside of the organisation. Simply put, Copilot is ‘Protected’.
Alternatives to Copilot
We’ve also listed some alternatives below.
- ChatGPT
- Gemini (previously called Google Bard)
- Claude 2
- Hugging Chat
- Pi, your personal AI
How do we use these AI tools?
Getting good responses from AI is an art in itself and is referred to as ‘Prompt Engineering’. There are many introductory guides and even a dedicated online course available from Coursera, called Prompt Engineering for ChatGPT, that is worth checking out.
The great thing about generative AI tools is that you can refine the output over and over again by asking follow on questions, e.g. to extract a more detailed or specific outcome, or to explore variations on a theme. We'll look at prompts in more detail on day 2 of this course.
All this is well and good, but what are the potential pitfalls with AI? Well, there are many. There is of course the element of ethics and the environmental impact (see the post on 5 April 2024) AI is having. This is something we will cover in more detail on Day 6, our last task on 7 May.
It is essential to remind ourselves that, while the outputs of GenAI tools may appear accurate and the tools themselves seem intelligent, this is not the case.
There are also concerns about the type of information and where it comes from. One of the key considerations is that AI is based purely on probability, so to find out more, I asked Chat GPT:
Me:
is AI based on probability
ChatGPT:
Artificial Intelligence (AI) encompasses a wide range of techniques and approaches, and probability theory is indeed a fundamental aspect of many AI algorithms and systems.
In many AI applications, especially those involving machine learning and statistical methods, probability theory is used to model uncertainty, make predictions, and make decisions under uncertainty. For instance:
- Probabilistic Graphical Models (PGMs): PGMs, such as Bayesian networks and Markov random fields, represent the probabilistic relationships between variables. They are used for tasks such as reasoning under uncertainty, pattern recognition, and decision making.
- Machine Learning: In machine learning, algorithms often rely on probability distributions to model the uncertainty inherent in data and make predictions. Techniques such as Bayesian inference, probabilistic classifiers (e.g., Naive Bayes), and probabilistic graphical models are widely used.
- Reinforcement Learning: In reinforcement learning, agents learn to make decisions by interacting with an environment. Probability distributions are often used to model the uncertainty in the environment's dynamics and in the agent's policy.
- Natural Language Processing (NLP): In NLP, probabilistic models are used for tasks such as language modelling, part-of-speech tagging, named entity recognition, and machine translation.
- Robotics: In robotics, probabilistic techniques such as Bayesian filtering (e.g., Kalman filters, particle filters) are used for localization, mapping, and sensor fusion, allowing robots to estimate their position and make decisions in uncertain environments.
While probability theory is a foundational component of many AI techniques, it's important to note that not all AI approaches rely solely on probability. For example, symbolic AI techniques focus on explicit representations of knowledge and logical reasoning, while evolutionary algorithms are inspired by biological evolution. However, even in these cases, probability theory may still be used in certain aspects or in combination with other methods.
Your task for today
Your task is to access Copilot and ask the same question that I asked ChatGPT. Do you think it will be a similar answer?
How to access
You will need to sign into Copilot with your Sussex information. See instructions above.
Give your feedback
Please feel free to give your feedback and answer the questions below using the 6 days of AI Padlet. There may be other issues or questions that you might want to explore. The aim of the discussion area is to get a variety of different opinions about AI. Many of our views about AI are still evolving and might change as we try out these new tools – you might just want to use this discussion area to test out some of your thoughts.
- How did you get on with the task? – did you find it useful?
- Did you notice any similarity in its response?
- Were there any surprises?
- Did it give you any other ideas how you could use Copilot in your teaching? If so, how else could you use it?
More information
You can find out more from the Educational Enhancement web pages.
Across this website, there are many terms which may be new to some of you. We’ve defined a few in our glossary of general AI terms.
Please note: This course is designed to provide an introduction to the more popular AI tools. However, unless explicitly stated, the AI tools discussed are not supported or endorsed by the University of Sussex. All are free to use but may require users to provide personal information (e.g. name and email) to access.
Please DO NOT upload your own or others' personal or private data, INCLUDING STUDENT WORK, to any such unsupported tools. Doing so could be breach data protection regulations.