It should be obvious that, like many other amateur technoweenies, I’m fascinated by AI. Or, better said, I’m fascinated by what AI can finally do.
Many of the components, tools, and algorithms current generative AI is based on have been around for a lot longer than most people realize. It’s only in the last few years that the hardware and software have reached the point where initiatives like OpenAI have been able to put all the pieces together to give us something resembling the computers from the original version of Star Trek. (Which, if you think about it, were pretty dorky compared to what we have now. If you lobotomize ChatGPT, give it Majel Barrett’s voice (definitely a missed opportunity there by OpenAI), add some chattering background noise, and throw in some sort of seizure for good measure, you would probably come close to the overtly metallic and stilted interface that 1960s network TV envisaged Captain Kirk and friends would be interacting with some 150 years from now.)
I’m similarly fascinated at how quickly such revolutionary inventions are copied. For instance, it took Apple about two and a half years to develop the first iPhone and, by all accounts, its debut in January 2007 was very premature. (There are numerous accounts about how the Apple engineers were all fervently praying (when they weren’t even more fervently drinking) that the multiple prototypes used during Steve Jobs’ keynote address at Macworld 2007 would not crash and burn despite all the other tricks and fakery that were in play.) But in October 2008, only 16 months after non-Apple people could finally get hold of an iPhone, the first copy of it in the form of the HTC-Dream, complete with a from the ground up new, Android OS, came out. Admittedly, HTC and Google had a blueprint to work with (whereas Apple had to create theirs), but they still had to reverse engineer most of it.
And so it is with generative AI. Within months of ChatGPT exploding on to the world stage, literally dozens of other generative AI applications have also hit the big (computer) screen, enabling people to generate music, art, and even entire comedy routines in the style of someone else (as well as to frequently, if very mistakenly, claim them as their own creative creations). And the list of new uses goes on and on.
Never being one to shy away from biting the hand that I feed, even WordPress is getting heavily in the AI act …
Through its Jetpack extension, WordPress has recently begun offering its users diverse AI tools to facilitate and even improve their blog writing. I’d stumbled upon some of these tools already and and was rudely made aware of some of the latest ones while writing my last blog when some new AI writing tools suddenly popped up with their suggestions without even asking. (You know, like Clippy, that moronic Microsoft office assistant that no one ever used.) The WordPress tools will check your text for complex words (which are highlighted in yellow), long sentences (in purple), and unconfident words (in green). If you let them, that is.
I won’t …
For starters, I don’t need WordPress telling me that my sentences are too long. I know that my sentence are too long. People have been telling me this literally for decades now and my father reminds me every time he reads one of my new blogs. (Which is ironic considering that he’s German and so should not be calling this Kessel schwarz when it comes to long sentences.) In any case, flick the switch to show which sentences in my blogs are too long and the purple haze that Jimi Hendrix sang about is nothing compared to what WordPress throws back at me.
And you know what? I don’t care. You want short sentences? Try Twitter (now known as X) …
The complex-words tool is even better because it’s just so, well, bad. Here is a list of all the words that it’s flagged in this blog until now (in italics) and the “simple, direct” corrections that it’s suggested instead (immediately following the complex word in parentheses):
- where initiatives like OpenAI have been able to (can) put all the pieces together
- There are numerous (many) stories … during Steve Jobs’ keynote address (discuss).
- complete (finished) with a from the ground up new, Android OS
- whereas (since) Apple had to create theirs
- and thereby to frequently (often), if mistakenly
- begun to offer its users numerous (many) AI tools to facilitate (help) and even improve
- immediately following (at once next) in parentheses
Many of the corrections are just dead wrong and sound like the invisible-idiot game where you had Google translate some phrase into a foreign language and then back again. (Some of the classic examples of mistranslation here might have been made up but there were still apps to automate the process for you. And they often came up with similar bits of nonsense.) As for the others, it would appear that the Jetpack AI has a thing against syllables, which would also explain why its writing service is called Writing brief with AI instead of the grammatically correct Writing briefly with AI.
As for unconfident words, the AI hammers me each and every time I use words like “probably” or “could”. And I literally mean each and every time. For instance, even though their usage in the first sentence of this paragraph is completely justified, the AI flagged probably and could nevertheless (and in this sentence too for good measure).
And, truth be told, who really needs these writing tools anyway? Or wants them? A lot of blog writers are writers and, so, probably know how to write in the first place. Or, more importantly, want to express themselves in their own words, not some standardized schlock that an infinite number of CPUs with access to an infinite number of keyboards churns out. When was the last time that any computer won any one of the Pulitzer, Booker, or Nobel Prizes anyway?
One tool that I admittedly had high hopes for was for the AI to generate suitable images for the blog. To be honest, one of the hardest parts about writing this blog is finding (or making) suitable, public-domain pictures for it. The words are easy, the pictures less so. It sometimes takes me hours to find or make good ones. If AI can do it for me in a couple of seconds, well, so much for the better.
Too bad that it can’t.
To the right is the AI’s free attempt to summarize this blog as one, featured picture. It’s very, very good, but what the hell is it exactly? And what the hell does it have to do with the content of this blog? I’m not 100% happy with the featured image that I did find (they should be computer chips, not poker chips …), but at least it captures the “all in” aspect of the title.
Fortunately, it is possible to get the AI to deliver an image according to some instructions so I asked it to make me a picture of “a computer poker player going all in but with computer chips and not poker chips” …
… and got an angry, older, and washed up Ryan Reynolds instead. (Still no computer chips either.)
Another, potentially useful tool is the one to suggest titles for the blog (although the official verb being used is “improve”, not “suggest”), another aspect of my blog writing that often costs me more time than I’d like. But, given these suggested improvements, I’ll happily invest that time.
- The impact of generative AI on technological advancements
- WordPress Jetpack AI tools: benefits and limitations for blog writing
- Enhancing blog content with AI-generated images: expectations vs. reality
Sounds like their AI was trained on admin memos …
The final, kinda fun tool is for the AI to generate feedback about the content and structure of the blog (which you can request over and over again until you get one that you like!). I’ll just leave you with my preferred version uncharacteristically uncommented.
The content touches on the fascination with AI and its rapid development, discusses WordPress Jetpack’s AI tools, and expresses disappointment with the AI’s image and title suggestions. The AI’s writing tools are criticized for providing inaccurate corrections and being unnecessary for most experienced writers. The AI’s inability to deliver suitable images is highlighted, along with receiving unexpected results. The blog titles generated by AI are considered useful, and the feature to generate feedback about the blog’s content and structure is mentioned as well. The feedback is comprehensive and provides a good overview of the content, but it could benefit from some reorganization to improve flow and coherence, especially in the section discussing the writing tools. A more focused approach to analyzing the specific features of the Jetpack AI tools for blog writing could enhance the overall presentation.
The Blog About Fog 