The world is all abuzz with AI, and I'm always interested in how the Internet has grown and changed over the years. It often surprises me how certain things take off (like Facebook, TikTok and Instagram) when other things didn't (like MySpace). Sometimes I think it's just the timing.
So, last month I had a play with AI and let CoPilot loose with the following prompt:
Write me a blog post discussing the impact that AI and ML might have on wastewater treatment facilities. Include positives and negatives. Include opportunities and hurdles. Add in a few case studies too.
I didn't do this just because I'm lazy and never get round to blogging much these days (though hopefully it will motivate me to get blogging again). I am interested to see what it can produce. The general guidance on using these LLM tools is they're OK to get you started, but you need to check them for accuracy as they can hallucinate!
 |
| Photo by Ehimetalor Akhere Unuabona on Unsplash |
The Good
So, what did I like about the blog? Overall, it was OK. I liked the structure, following what I asked for in the prompt, with sensible section headings and all. In terms of the content, it identified several ideas that made sense to me, including AI and ML used for predictive maintenance, process optimization, and anomaly detection. It also identified data quality and concerns over cybersecurity as hurdles to wider adoption. The examples were OK, but they lacked detail and there are others out there that may be better.
The Not-so-good
The key opportunities listed were a bit weak. It started out with ideas about using AI for achieving regulatory compliance, which I guess is plausible, but not an obvious thing to focus on. Energy efficiency is a more-obvious thing to focus on, which it noted, and of course it had to mention digital twins, though it thought of them just as tools for doing scenario testing, which is just one application and not the most common. In the category of challenges and risks, it listed high initial cost and workforce skills gap. I guess these are valid, but they're not the topics I'd focus on at first.
The Bad
Argh, I hate hype, buzzwords, and corporatespeak so the title about killed me: "How AI and Machine Learning Are Transforming Wastewater Treatment Facilities." AI and ML are NOT transforming wastewater treatment facilities: not yet anyway, and not in any widespread way. So, I hated the title. Also, they just had to get the word "sustainability" in the text didn't they. What the heck AI and ML have to do with sustainability, I don't know, but you'd better throw that buzzword in there if you're doing anything remotely associated with environmental engineering. Oh, the last sentence is precious: "Facilities that embrace these technologies early will gain a competitive edge in efficiency, sustainability, and compliance." Kill me now.
The Weird (or plain wrong)
So far, I've noted what I liked, what niggled me a bit, and then some things I disagreed with but maybe you could argue it is just my opinion of what's bad (poor AI is fed on whatever it can scrape from the Internet after all). But there were a couple of things that were just plain wrong. Firstly, it said that AI and ML could be used for resource recovery. Que? And of course, it gets to talk about this as a wonderful thing to help with the circular economy (see previously about hype and buzzwords). Yeah, no, I don't see it. Maybe at a stretch you could apply AI and ML to try to do anything, but there's nothing magic about AI that will help us with resource recovery, I don't think. The other error, which was the weird one, was the graphic I asked it to make. It was supposed to be an infographic to summarize the whole blog. Why does CoPilot (and probably other AI) struggle to incorporate words into graphics? The title it came up with is precious: "How AI and Machine Learning Are Wastewater Treatmater Treatment Facilities." Weird. I'll say no more.
Summing Up
So, overall, I think the AI did OK. I published the blog, warts and all, with no tweaking, refining or additional prompts to refine it. As a way to get a blog started I think it did a reasonable job. I could see me using it to generate some initial thoughts and ideas on which I could add or remove things I didn't like. On the flip side, I've also used AI to help summarize things or rewrite something. It makes me a little nervous that it might miss something, but again it's on me to check what I'm getting. For now, I'll leave the training wheels on and be careful in my use of CoPilot or other AI to generate stuff. Mostly I'll stick to using it for curated searches and maybe summarizing things, but for idea generation I'll stick to using the old grey matter for as long as it functions normally, or maybe until I do my own hallucinating (but how will I know?). And of course, if I need help describing the Transforming Treatmater Sustainability in the Circular Economy, I know where to go!
