Generative AI can be overwhelming. But over time, it’s possible to find ways to make it work for you. Here’s how AI use works in practice for me…
I’ve been a bit critical of AI lately. First I wrote a long set of thoughts about privacy and the need for caution, and then I did another long set of thoughts about hallucinations and the need for caution.
Yes caution is important, but with so much of it sloshing around the sensible woman’s part of the world, you might be wondering if I in fact use AI at all.
Dear reader, I do indeed use AI. A lot, in fact. And thought it might be useful to share the ways in which I do that. I’ve kept a running list of how I use AI over the course of several days, and here, grouped by theme, is the result.
Everyday uses of AI
On Tuesday night, I made a one-pot chicken dish that is now a staple in our house – originally generated by ChatGPT. And tonight I’ll be trying one for pasta and bacon from a new kid on the block Epicure AI, which promises to ”leverage the power of AI and machine learning to explore science-backed flavour pairings and generate recipes”.
Tomorrow, if I remember to take my phone with me on the dog walk, I’m will take a picture of the pretty yellow flowers coming up as weeds on our pavement, and I’m going to ask Gemini or ChatGPT to tell me what they are. Because even weeds have names.
A way in which I didn’t use it, even though tempted
On Monday this week, I needed to write the texts of my social media posts so that they could be sent to Anya (my son’s girlfriend who is also my paid assistant) for scheduling.
The posts are based on whatever blog post I wrote the previous week – which this time was a personal musing about what it means to be grown-up. Since (regrettably) my main social media platform is LinkedIn, that meant I needed, theoretically, to extract some business angle from the post.
Now, I’ve tried this using various LLMs before, and have used the (edited and rewritten) results. Over time, when I’ve seen the resulting posts in real life on a social media platform, I’ve always been struck by how much they simply don’t sound like me.
So I sighed, and put on my thinking and writing cap, and wrote the blessed post from scratch.
A way in which AI was just useless
A business partner and I are emailing a long list of experts to ask them to take part in a special report we are writing. The names were grouped, with several people per country needing to get the same email. I looked at the spreadsheet with all the countries and names of people who needed to get an email, and eyeballed it to assign half to me and half to my partner, typing our names in the various rows.
Then, since this was a Google sheet, with Gemini integrated, I asked Gemini to tell me how many times my name appeared in the sheet (because counting them physically on the screen is a sure way to get it wrong). Gemini insisted my name appeared not at all.
So I used the old-fashioned find function, and established that my name appeared 13 times. It should have been 16, and so I changed it to make our shares of the work equitable.
It’s probable that this example of uselessness comes from my not understanding how Gemini interacts with a sheet (rather than a text document). But there was a simpler way to get the job done, and I didn’t go down the rabbit hole of trying to make Gemini do what I wanted.
(If you are wondering about the privacy implications of all those email addresses sitting in Google Drive, know that I set the document to be viewable only by team members.)
Using AI without actually choosing to
In the course of my work, I do a lot of internet searching, usually on Brave and sometimes on Google, when Brave seems not to be delivering. These searches now usually produce an AI-generated summary of the results, and a list of the sources used to generate that summary.
I don’t set much store by those summaries – they’re often wrong, or inaccurate (in a recent search to try to solve a pesky problem in Microsoft Word, information in the summary had come from a link that dated back 2004. I don’t think so, dude). But I do skim them and I do skim all the sources given, and choose to look at the ones that seem as if they will work for me. Essentially, I’m using these summaries to show the sources in a compact form rather than scrolling through endless sets of links.
Using AI to store things, and research things
Three times this week I’ve made NotebookLM collections of links related to articles I’m writing, none of which have imminent deadlines. But now that I have the notebooks created, I’ll be able to stash links that I come across there, safe until I need to use them.
When I do write those articles, I’ll be asking NotebookLM questions about the content of the links, or for suggested outlines for the article, or to find the source of something I remember reading in one of those links.
READ: The AI tool everybody should be using – NotebookLM
Using AI to save me lots of time
I’m working on revamping my website, and one of the things I want on the site is a comprehensive page that lists all my offerings, with prices, and examples of how that pricing works in the real world.
A while ago, I put all the documentation (scrappy notes mostly, and long AI-generated tracts of copy) about what I want on the revamped site into a notebook on NotebookLM.
On Tuesday, I asked NotebookLM to take all the scrappy bits and turn them into the text that I want for that website page. It did the job in under five minutes, saving me the hours-long chore of reading all those old notes again, and synthesising and writing.
I then spent 40 minutes or so rewriting and reformatting that text – but I have to say there was no reorganisation to be done at all. NotebookLM had correctly reflected the thinking I’ve done on this, including incorporating information I had previously done on actual pricing.
Overnight though, it occurred to me that some of that pricing came from oldish documents. Also it was time to do a job of work on pricing in general that I have simply never had the courage to undertake. That job of work means taking a high-level look at the work I do, how long it takes and what I should be quoting.
So on Wednesday, I turned to Google Gemini and its Gems, and built myself a quick “business money” coach.
I put in another set of scrappy documents where I had previously, over a period of years, worked out quotes for jobs. I entered an hourly rate (which I work out every year, using a tried-and-tested formula) and uploaded the package document I’d made the day before, calling it the Target Document. Then I asked the coach to populate the target document with updated pricing, based on the hourly rate and all the other information I had given it.
READ: How to build yourself a virtual coach with AI (for free)
And my “coach” did it, in five minutes (I was using the reasoning model of Gemini, which I pay for). I manually added the new prices to the original NotebookLM document. I also double-checked the maths, which was correct in every case. In one service I offer, Gemini had neglected to provide a real-world example. I asked it to do that, and added it into the text that will make my web page.
That took me about 40 minutes. Something I’ve been putting off for months, and which would have taken hours, has now been crossed off the list.
Every time I do something like this, I think how grateful I am that I decided to embrace this tech, and learn about it. Used with care, it has made a huge difference to my working life.
The lessons learned
* Don’t use AI to do something that you can do better (even if it takes more time).
* Sometimes, it’s quicker to use older, tried-and-tested ways to get things done.
* If it seems that AI has not delivered what you want, take into account that it might be something you are doing wrong.
* If you’re putting people’s personal details into AI (or anything else that is internet-connected), take measures to protect their privacy.
* A fair amount of your browsing of the Internet is now being mediated by AI – keep your wits about you, and know what you are looking at.
* Use AI to your heart’s content when you are working with text or ideas that you already understand, and can assess.
Main picture: Árpád Czapp, Unsplash
Other things I’ve written
Privacy and Generative AI – what you need to know – I’ve done some research on the question of privacy and Generative AI. My findings and thoughts…
What you need to know about AI hallucinations (hint: they aren’t hallucinations) – When AI makes mistakes, it’s called hallucinating. There’s a better, if ruder, word for it. That means this post contains what the dictionary calls vulgar slang. You have been warned.
What does being grown-up mean? – There’s adulting, and then there’s being grown-up. I learned the difference between the two early one morning…
The AI tool everybody should be using – NotebookLM – Research assistant and note-taker rolled into one – allow me to introduce you to NotebookLM.
How to build yourself a virtual coach with AI (for free) – So you know how to use ChatGPT (or Claude, or Gemini). But you’re sure there’s more you could be doing. Here’s how to build yourself a virtual coach with AI (for free).
A beginner’s guide to charging for editing services – It’s the bane of every self-employed person’s life – how to charge for the services you give customers. A beginner’s guide for editors starting out on their journey….
LinkedIn – can I learn to like it? – January brings the annual review of my social media activity – and also the annual LinkedIn crisis. Why be there if I dislike it so? (Written in 2023)
How can I help you make order from chaos?
Join the Safe Hands AI Explorer learning circle!
Sign up for my Sensible Woman’s Guide to AI and Content Creation, which comes out fortnightly.
Or sign up for my personal fortnightly email newsletter In Safe Hands (it’s a place to pause and think about things).
Book a free half hour of my time here. Find out how I can help with writing, editing, project management, general clear thinking, technical hand holding, or an introduction to AI.
Contact me via email.