Further to my thoughts on the dangers of generative AI not just to our livelihoods but to how we relate to Nature in its broadest sense, I also want to flag up a smaller issue I have with this new technology. The biggest impact of AI in my working life so far has been improved captioning, and spelling and grammar checks. They’re still quite* inaccurate, but as someone who learned to do a lot of things the long way because there were no other options (or no other affordable options), I appreciate how hugely time-saving they are. I appreciate how much they’re improving. The inaccuracy matters, though.
Most generative AI suggestions are great reminders that fundamentally, what this tech is doing is taking an input, comparing it to other similar things, and suggesting an option that’s closer to commonly used patterns. You end up with a grammatically correct piece of text that, at best, has had its individuality sanded away and, at worst, means something entirely in contradiction to what you’re trying to say. In the case of Grammarly, for example, it offers to ‘improve’ your text, but really, it’s a kind of normalisation.
For transparency, I should say that I use Grammarly regularly, and its spelling and grammar checks are the best I’ve found, though I reckon I only accept 75-80% of the suggestions, and I’ve never been happy with the generative AI suggestions.
It’s the latter I want to highlight. I tried re-writing a paragraph from my previous post with Grammarly’s generative AI. Here’s the original:
Continue reading Lost in Improvement: a Quick Postcript to My Previous Post
Paul Burgess 