Last week, I had the chance to work with 120 midwifery educators from across Australia. Many told me the same thing: "I feel overwhelmed by this AI thing...and now I'm just trying to catch up." If that's you, then you’re not alone. GenAI arrived fast, and the conversation has been full of jargon, hype, and big promises. So here’s a practical, non-technical primer (part 1 of 3) to help you get oriented. (Note to AI professionals - this deliberately is non-technical. Please forgive me for over-simplifying a complex emerging field). First: What is GenAI, really?
Artificial Intelligence (AI) has been part of our daily lives for years, it has been recommending what we watch, buy, and click and more recently even helping diagnose malignancies. That older form of AI was good at recognising patterns, but not at creating anything new.
Then, around 2017, a new approach changed things: transformer models. These models don’t just recognise patterns, they generate text, images and speech that sound human-like.
That’s where tools like ChatGPT, Gemini, Claude and Llama come in. The recent powerful jump came when models became multimodal, meaning they can:
- Read and interpret images
- Understand and generate speech
- Generate visuals
- Translate languages
- Summarise documents within a single conversation.
This is why things feel like they've suddenly ramped up, they actually have.
What These Models Do Well
GenAI tools today can:
- Summarise and explain documents in plain language
- Extract themes from handwritten workshop notes or flipcharts
- Generate convincing podcasts and even interact with a live audience (I demonstrated this Google Gemini capability during my recent workshop) and we even showed how it could debate the topic of midwifery education between artificial 'agents'
- Translate languages instantly
- Generate lesson plans, discussion prompts and reflective questions
- Draft reports or emails you can then refine (for example, I generated this newsletter, then got Chat5 to help turn it from a lecture to something hopefully more engaging)
- Write basic code (with comments) and generate experimental apps in just a few minutes (last week I demonstrated using Google's Gemini Pro model during a 45 min talk in which we interactively created a fun game for new parents based on what to expect in 24 hours with a new infant).
Limitations
Be aware that most GenAI systems are particularly strong at language, but not numbers. They struggle with:
- Arithmetic and numerical reasoning (they are paradoxically excellent at higher mathematics which involves symbol reasoning) - but just be aware they often cannot count reliably!
- Reading and 'understanding' very long documents
- Knowing when they don’t know something Because they are trained to always produce an answer, they sometimes hallucinate — confidently making up details, citations, or facts. That’s how some law firms and consultancies ended up submitting reports referencing cases that never existed.
So a core rule I can offer to you is this: Use GenAI as a reasoning 'buddy' to test your thoughts, but not as a source of truth.
The bottom line is that human judgement is absolutely required, and we cannot as yet rely on the outputs from GenAi systems. That may change with new approaches being developed but I find it helpful to think of the GenAi systems as a brilliant but somewhat untrustworthy teenager.
BUT - we can use OTHER GenAI models to help us cross check. For example, when researching ideas I often ask Perplexity.ai to examine my material and cross check for citations. And then I click through each citation to ensure I am completely happy that they are correct. In my academic work I tend to use ResearchRabbit.ai to help find highly cited sources that is a good starting point for researching a topic.
The Bigger Concerns
Two other issues are worth keeping in mind:
- Bias: We don’t usually know what data these GenAI models were trained on, and this can shape the answers they give. Many advanced models from overseas have 'airbrushed' out uncomfortable historical events. Bias can be more subtle than this: there have been cases of job applicants being discarded on the basis of the AI making assumptions of gender since that was a bias in its training data.
- Privacy of data can also be an issue: the 'free' versions of the most powerful commercial models often have limitations with respect to handling your data with confidentiality. 'Pay for' models can usually be restricted to operate within your corporate domain, and / or provide you with the ability to restrict the models from using uploaded data or documents for training of future models. Nonetheless, this is a variable aspect, and a concerning one. My rule of thumb is never to upload private or confidential data, corporate information or ANYTHING that can be identified (unless that data was already in the public domain, such as on a website) unless the model is entirely locked down to a defined corporate domain.
I also like to check and know if the uploaded data will flow out of my country (most do) since that is a definite NO for many healthcare-related uses.
So Where Should a Professional Start with GenAI?
Firstly, check with your corporate policy about what you can and cannot use any public GenAI system for - most organisations will currently have at least some restrictions (for good reasons).
Provided that your work policies permit you to use them for work purposes then I recommend you start by using GenAI to help you with work you already do so you can easily integrate some aspects into your life, and you can check for its usefulness. For example: Drafting emails Drafting emails or reports: Ask the model to suggest a first draft, then edit it with your expertise.
Or give it a draft you have prepared and instruct it to rework the draft for the audience you have in mind. For example, in a conference I just attended there was a expressed requirement for people to tailor communications for neuro-diverse people. These sorts of challenges are excellent for GenAI to have a go at. Preparing workshops or lessons Try asking it to generate discussion questions or case scenarios. In my case I asked Chat5 and Claude to suggest some examples that might resonate with the audience I anticipated to attend the workshop.
Understanding unfamiliar concepts Ask for a plain-language explanation with examples. You can ask GenAI to produce a primer on AI for you (as an example) and you can instruct it to use non-technical language.
Summarising policies or guidelines Give it a PDF (one you are comfortable uploading) and ask: “Summarise the document into key themes for educators”. Or give it workshop outputs you have taken as phone photos of flip chart outputs and ask it to summarise and theme the outputs: I have found Chat5 and Gemini Pro are brilliant at this. If you take one thing away You are not too late to get started! You are entering at a time when the tools have become usable, stable, and genuinely helpful for everyday work assistance.
Just bear in mind that none of them are perfect and they WILL make mistakes and potentially mislead. You need to be vigilant. You pay for what you get: the most powerful models sit behind pay walls, generally the free versions are limited and shallower in their reasoning and / or do not protect your data privacy.
I hope my goal of this newsletter has encouraged you to start to become a confident, discerning user who knows when, and when not to rely on these systems.
In my next primer Newsletter, I will show some setting up and prompting ideas and how I use a variety of the models (including Copilot) to undertake some common useful business and research tasks. In my third in the series I will then cover Open models that you can run on your own systems, which is something I do for sensitive material.
Let me know if this is helpful and if you have other tips to help a newbie start with GenAI. Paul
Leave a Comment