Using digital tools

Keeping up with the AI revolution

By 29/03/2024

This blog is the first of a two-part overview on AI which will cover the basics of using AI in our day-today-life. AI is here to stay and will only improve with time; we all need to make sure we understand the technology well enough to use it to our advantage – or risk being left behind! Make sure you stay tuned for the second part – it will be published on Friday 5 April, and will cover the uses and limitations of AI in research.

Artificial Intelligence (AI) has rapidly evolved from being a futuristic concept, to an increasingly integral part of our daily lives, hobbies, and jobs. We’re all familiar with virtual assistants like Siri, Alexa, and Google Assistant – but how well do you know the next-level companions?

Today we’re delving into the versatile world of AI  language models like ChatGPT. But first, some distinctions: 

You may have heard terms like AI, Large Language Models (LLMs) and Generative AI being thrown around. Although they are related concepts in computer science, it’s important to know the distinctions between them.

AI, in general, refers to the development of intelligence systems that can mimic human behaviour and decision-making processes. Generative AI tools generate new data similar to the training data they have seen; depending on the type, they can synthesise images, generate text or even music. LLMs, for example,  are a specific application of generative AI which provides a basis for natural language processing (NLP) tasks. ChatGPT (which stands for Generative Pre-Trained Transformer) falls under the category of LLM.

Since they have been trained on large amounts of text data in order to learn the patterns, grammar and semantics of human language, LLMs are able to generate a coherent and relevant response when given an input – such as a sentence or a prompt. 

But ChatGPT can do more than just answer user queries; the AI powered chatbot can write articles, create lesson outlines, and much more. In the emerging market of generative AI chatbots, ChatGPT already has competition in Google Gemini (formerly known as Bard), Microsoft Bing Chat, and others.

So far, there have been mixed feelings about the potential and utility of language models. Whether you have tried using AI and found its fabricated facts annoying, or you are afraid of contributing to its learning because it might make us all redundant in the future, this blog will help you harness AI as a useful tool for boosting your productivity.

Demystifying AI language models 

Language AI models, like ChatGPT, work by learning patterns and relationships in vast amounts of text data. During training, the model processes diverse examples of human language, understanding how words and phrases relate to each other. It learns grammar, context, and even some reasoning abilities. Once trained, the model can generate text or answer questions based on the patterns it learned. 

So who should be using AI and when should they start?

The answer is everyone, and now! AI is not going away, so it’s best not to bury your head in the sand – ignoring it won’t safeguard your job. The key is for us to use AI to help us work more efficiently, and use the freed-up time to be more creative and productive. ChatGPT and its counterparts are already making waves across a variety of fields, and it isn’t confined to the realm of technology and science. From assisting writers in content creation, summarising complex concepts, or helping researchers analyse data – the possibilities are vast. It’s about tailoring AI to streamline your workload, and make your life a little bit easier. 

Bridging the gap between expectations and reality

If you fall into the camp of AI users who can’t seem to get what they want out of AI, the chances are you may be falling into some very common pitfalls. The next section of this blog will explain how to avoid them, and how to overcome some of the challenges that may arise.

Craft a precise prompt 

The Achilles’ heel of language models is vagueness. ChatGPT is great, but it is not a mind-reader . Users often make the mistake of assuming  it knows what they want. The solution? Be precise. In your prompt, provide context, outline your desired outcomes, and be explicit about your requirements. The more context you give, and the more specific your prompt is, the better the result will be.

Bad prompt: “Write something about technology.”

This prompt is vague and lacks specificity. It gives ChatGPT little guidance on the desired outcome, leading to a broad and possibly unfocused response. The result may not align with the user’s expectations, and the output could be generic or scattered.

Good Prompt: “Compose a 300-word blog post highlighting the impact of artificial intelligence, specifically ChatGPT, on everyday communication. Discuss its applications in diverse fields, emphasising its role in enhancing productivity and creativity. Use a friendly and conversational tone.”

This prompt is specific and provides clear instructions to ChatGPT. It outlines the desired word count, the main topic, and even suggests the tone. By being detailed and explicit, the user increases the chances of getting a tailored and relevant response.

No matter how good you think your prompt is, it’s unlikely that you will get exactly what you had in mind on the first attempt. This is an opportunity to experiment with different queries and refine your instructions, until you get to the desired output.

Do’s are better than don’ts 

Positive reinforcement is key. Instead of telling AI not to do something – for example, saying  “don’t use a formal tone” – instruct it to “maintain a conversational tone”.  ChatGPT responds much better to clear instructions of what you want it to do, rather than what you want it to avoid. This is because the former is a restriction to creativity, whereas the latter channels it in the right direction.

Feedback is valuable 

The conversation between the user and the model is a two-way learning street. Whether you receive a perfect or totally rogue response, take a moment to provide feedback. Highlight what worked well, what missed the mark, and share specific examples.

This is especially important given ChatGPT’s infamous tendency to fabricate facts or provide inaccurate information. These are called ‘hallucinations’. It’s important to approach AI-generated content with a critical eye, and verify information through reliable sources. Don’t hesitate to question a response and seek clarification, and provide explicit instructions to prioritise accuracy. 

This feedback loop contributes to the model’s ongoing learning process, making it more attuned to your needs over time.

Guard sensitive information

AI models generate responses based on patterns learned from a wide array of sources. They are not able to either comprehend or respect confidentiality. This can be dangerous – so you need to be careful about what information you feed into AI prompts. For instance, passwords, financial data, or any information you wouldn’t disclose in a public forum. 

Have a go!

If you’re still not convinced about how AI can be of service to you, or you’re not sure about how it fits into your personal and professional life, here are some very practical everyday uses:

  1. Summarising a long, complicated document into a simple, quick read
  2. Proofreading and reviewing documents
  3. Drafting outlines for articles, blog posts, or emails
  4. Creating to-do lists and organising tasks
  5. Real-time language translation

What to read:

Beyond applying these guidelines, we recommend arming yourself with the right resources to take your knowledge further.Here are some we found particularly useful:

Who to follow: