Everyone is using LLMs to help with their work – why wouldn’t you? Well, there is a good answer to that – which is that it is not possible to rely on the outputs it produces when it thinks for itself, even at your request. It literally guesses and often hallucinates. It cannot know enough to do what you need it to do without error. But it has other values.
Imagine, if you will, a brain still evolving. It can do certain simple tasks well – especially if they are purely instructional – and in particular if there is no reward required to repeat those tasks. As a scholar, ideas are everything – and they need to be unique, original – and build on the work of others – standing on the shoulders of giants, as it were. LLMs can’t do that – when they do, we are all cooked, although it may be sooner rather than later, as LLM power and capability are currently doubling every seven months, although costs remain the greatest barrier to infinite progress.
I use them – a lot – but for very specific and simple tasks – and each LLM has its best ‘free’ feature that can be used all day long.
One – I have been voice typing as soon as I got hold of Dragon Dictate sometime in the early 2010s – now voice typing is embedded in Windows – simply click Windows Key + H and a little microphone appears – suddenly your laptop is listening – and it will type up what you say in any content window or as free text in Word. Google Docs has a function for voice typing, easily found from the menus. Once this is in the bag, typing is never needed again – it might feel odd talking to your laptop all day, especially as you have been used to tapping keys all the time, but you slowly get used to it – this is also valuable when you make handwritten notes. Instead of typing them up with poor and imperfect typing speed, you can simply dictate them directly – faster and mostly error-free.
But even voice typing is never error-free – as how we speak is not how we think when we write – and vice versa – we also use shorthand – i.e., everyday phrases and less sophisticated vocabulary when we speak.
With all text – once on the screen, there is a simple fix. Using any non-paid LLM – I mean any – and these are ones I use: ChatGPT, Gemini, and Notebook LM – simply (voice) type this command for any text, and it fixes your text in seconds. Use the following:
Remove all typographical, grammatical or punctuation errors only: “text”
Insert your “text” wrapped in quotes and voilà – it’s done.
This fixes your language – and it’s great for people like me who have a bit of dyspraxia – and for others, too – it also adds confidence and speeds up all editing.
Two, images. I often like to add a cartoon to a blog (like this one) or improve on a bar chart – for this, a clear and precise prompt is needed – and I would recommend ChatGPT – the quality is very good – however, you only get 3 or 4 free options a day so use wisely – I wouldn’t bother with any other LLM for images.
Three, tables. These can easily lose formatting or need better formatting – for this, use Gemini and click the writing option – so it produces the table in its Canvas – once done, export to Docs, and then to Word as required.
Four, vibe coding. I read mathematics, statistics, and economics 35 years ago, and I learnt how to use SPSS on MS-DOS, with most of my time devoted to debugging syntax in order to create and recreate variables and undertake some basic testing, but with vibe coding, large language models assist greatly in producing initial templates for all the modelling I need and want.
Vibe coding is when you prompt an LLM to write the code for you, specifying what you want, what you want the statistics to do, and why, and it produces the syntax, which you then use in third-party software, in my case, R.
That’s it… this is all I need… typing, editing, formatting, and coding assistance – and the rest is my own brainpower.
There are other uses, too – and I do use them…
I am drafting an email – and I want to check its tone – I ask any LLM to do that.
I have a title that I don’t like – I want to play with the one I have – I ask…
I want to reorient an abstract – I ask.
I want to check for flow and style – I ask…
I ask – and receive abundantly – but I accept very little – as, fortunately, I have a screen built over 30 years, when I started academic writing with pen and paper – and only used a word processor in typing up handwritten masters essays… After which, I fully reverted to typing, in the end producing more quantity than quality – and now with these tools, I can revert to handwritten notes, based on quiet and clear reading, and speak them into my laptop – and compose and compile – and refine and redevelop quicker, faster and more effectively than ever.
So, the moral of the story is this.
LLMs are tools – they are not solutions.
LLMs cannot replace the brain – but they can aid in tasks.
Use LLMs to free up time for greater cognition, not less.
As the risks of cognitive off-loading – which means ultimately your brain will reduce in capacity, leading to early-stage Alzheimer’s for all of us – will speed up societal decline, making us all ripe for further exploitation until there is truly nothing left of human form but to basically exist as consumers and moaners while all our remaining capital is extracted by AI-driven cloud computing tech bros who will rule the world.
You have been warned.