There has been a lot of discussion in the last year about when and where it is acceptable to use artificial intelligence (AI) to craft text. And, very belatedly, high level discussions about identifying and controlling AI have begun.
As it happens, I was browsing a website called Stack Overflow, where individuals who write computer code can get help from others (fellow humans). There’s a page there titled “Why can’t I use Artificial Intelligence tools to generate answers?” that provides reasoning applicable well beyond the scope of the one website.
For example, AI derives its answers from existing text and as a result its answers can sound quite believable, even if inaccurate or in some cases, just plain wrong. AI may not consider whether its answer is accurate and relevant and also may add a lot of unnecessary or irrelevant details.
Furthermore, if an individual wanted an answer from AI, they’d probably already have consulted AI themselves.
And finally, AI rarely cites its sources and even when it does, those citations may not be relevant and the “sources” may not exist at all.
All good reasons why, in almost any situation, if a person asks another person a question, they deserve and should get a response from that person and not from AI.
An example I’ve been encountering recently is when I seek help from WordPress support. Inevitably the first response I get is from a bot/computer/AI. It’s mentioned at the top that it’s a computer response from “Happy Bot (Automattic)”* but I didn’t notice that at first. It sounded like it was from a human and it addressed the issue. But often it addressed only part of my question. Or ignored or overlooked a specific item. And very often it would suggest I consult the copious amount of do-it-yourself guides that WordPress has posted — guides which, if I thought they could help, I’d have already consulted.
In many ways, a computerized response like this makes sense. The AI might very well answer the question, thus saving a WP employee from having to respond. (My questions, of course, are far too advanced/clueless for any mere bot to answer.) The message closes with the note that if it hasn’t answered my question, more help is available: “Still need help? Just reply to this email.” (The additional help, finally, will be human.) So, yes, AI has its uses. No doubt it saves WordPress support staff a lot of time and probably answers a lot of basic questions with its links to existing information. But if a question gets specific about a certain detail in a particular post, the bot’s answer can be less than useful or just plain wrong by suggesting a solution that doesn’t apply in the situation described. It may even suggest trying something that I stated specifically I’d already tried, with no success.
Certainly AI has its uses. But one should never forget its shortcomings. Relevant, perhaps, is the old (1960s) journalist’s rule: Always confirm with at least two other sources.
___________________
*Automattic owns and operates WordPress. The name is derived in part from the name of developer and founder Matt Mullenweg.
I have’t bothered with WP Support for so long that I think I can be sure when saying I’ve never had one of those preliminary bot answers, Colorado. And my reason for not asking isn’t that I’m so on top of everything – far from: it’s that WP has changed virtually everything, and I’ve completely lost interest in any attempt to ‘play’ with my blog or improve it.
I consider Automattic to be the perfect example of a staff who have nothing to do but change stuff – most definitely for change’s sake.
You get a rueful chuckle from me. The constant WP changes have accomplished little more than giving WP employees something to do — whether dreaming up new complications or answering pleas for help from people like me. I forget the timing but the bot response might have originated with the introduction of AI to the masses. I get very frustrated with all the new stuff but I owe it to WP to say I’ve always gotten responses and followups when I’ve asked for help.
Indeed…verify!
Absolutely. And always.
I was disappointed when I saw that WordPress had a “let AI write your post” option that was pretty dodgy to say the least. Seriously? What a fabulous opportunity to propagate garbage!! In a world where people pretty self-select themselves to bubbles of misinformation, what could go wrong… after all, that misinformation is there for the AI to use.
Indeed. Why bother with a blog or any kind of online presence if you’re going to let AI write it for you? I don’t see the point. A spell checker or grammar checker, sure. But having AI write it for you smacks of so much laziness and disinterest that I don’t know why a person would bother with a blog in the first place.
Exactly!