4 hours ago
4
I am not, by nature, an early adopter. There comes a point in our lives where change becomes more irritating than exciting and, I suspect, I reached it sooner than most. But when a workplace recently tasked me with exploring practical applications for AI, I spotted an opportunity to cast off my luddite inclinations.
It turned out AI was very good at mimicking most of the things I could already do. Irrespective of quality, it could churn out articles, reports, presentations, fiction, even podcasts with stammering hosts. That was no use to me. What I wanted help with was all the stuff I was useless at. There was an obvious target: DIY.
It helped that this tasking coincided with our purchase of an old house in need of extensive TLC. Before settlement, I had managed to convince myself – if not my partner – that I would be able to conduct many of the smaller repairs the property desperately needed. This was quite an achievement, as I come from a family of great renovators and tradespeople and I have been made aware, from a very young age, that those particular genes were not passed on to me.
My experience of AI in the workplace is that it behaves like the worst member of any team. It will do the bare minimum, it will invent nonsense instead of checking its facts, and it will consistently lie to you about whether it actually did the work. Nonetheless, I embarked on our renovations with my idiot artificial assistant as my guide.
My approach was to treat ChatGPT as a searching machine. I didn’t want it to create but advise. Isn’t this what robot assistants traditionally did in science fiction? Surely it could go and scrape the internet for the information I needed to know in order to paint a bedroom, build a soak well and restore some ancient windows.
For the most part, this worked. It was able to talk me through the purchase of the right equipment, the right paints and primers, the right brushes. It could compile shopping lists and help me choose between products on the shelves. This may have used up essential water stocks – one of the chief objections to AI being its environmental impact – but at least it spared me the embarrassment of asking 100 stupid, manhood-sapping questions of the nearest shop assistant. It was also able to talk me through the process, step by step, and assess my work on the fly.
What I wasn’t prepared for is how needy AI is – or, perhaps, how needy it thought I must be. Instead of giving simple feedback on my progress, it lavished praise on each layer of primer. What an amazing job you’re doing. Most people would have made a mess of this. This hole you’re digging is a really professional piece of work. This was intensely irritating. I needed advice, not approval.
What AI is very good at is building your confidence – whether you deserve to be confident or not. It will always give you the second opinion you’re wanting, if not the one you need. This might be fine, if you know enough to spot the gaps in its thinking. The builders I went to with its plan to build a new subfloor atop an existing floor with rotten stumps did not take long to identify potential issues.
The dangers, when you are applying those opinions without expertise, are obvious. While AI gave me the confidence to attempt a small soak well, it also insisted I would need two tonne of drainage gravel to fill it (I actually needed about 20kg). It was only as I considered fitting it all in the boot that I decided to make my own calculations.
While AI can give you confidence to attempt something outside your comfort zone, success will rely on human judgment and critical thinking. The issue is that apps such as ChatGPT make that judgment harder, speaking as they do with absolute authority (while flattering the hell out of you). To use them effectively, you need to be able to parse these suggestions for what a bot thinks you want to hear and apply the one thing it cannot – experience of the real world.
It worried me how quickly I became dependent on AI to make decisions around my DIY projects. I could feel that dependence leaking out into other areas, from dinner choices to movie nights. I see others talking about becoming reliant on bots to make decisions from the trivial to possibly-violating-local-building-codes. Decisions can feel oppressive – why wouldn’t you outsource them?
Because, of course, our choices tend to define us. We have to live with those choices, while the bots can only speculate what living might be like. If anything, the emergence of these decision-making machines highlights the importance of critical thinking. Our kids will be presented with more – and less trustworthy – information than ever. Information that, like my assistant’s DIY suggestions, will be carefully tailored to appeal to their particular interests or biases.
My hope is that being suckered by AI will largely be a symptom of old age. Already I see AI videos being dismissed as “boomer slop”. Our kids have, quite rightly and repeatedly, mocked me for asking ChatGPT for renovation advice.
I am nonetheless grateful for the unearned confidence it gave me as a DIY newbie. Because it didn’t take long to realise that, actually, maybe I could trust my own judgment a bit more. Nothing does more for your ego than realising you can make a better decision than a bot with all of human knowledge at its digital fingertips.
Either that, or those dormant DIY genes are finally kicking in.
