I is for IMPROVE

The future arrived last week in the form of WordPress with enhanced AI functions. I was not happy that this future downloaded and installed itself as a non-consensual update. The AI STUDIO button appears in the toolbar by default but disappears when you close the sidebar. I used the next paragraph to run some tests.

Clicking on the AI STUDIO button opens the right sidebar with four pre-set options and one custom option. The pre-sets are IMPROVE, FIX GRAMMAR, MAKE SHORTER and MAKE LONGER. The blue button at the bottom of the screen opens an AI chat. I already know that MAKE LONGER will make a text longer without adding any new information or content and that MAKE SHORTER will do just that, only less noticeably. FIX GRAMMAR, I’m in two minds about. Lately, I’ve been using some experimental grammar in the hope it will make my writing easier to identify as not AI-generated. I’m also more relaxed about typos for the same reason, as well as using a mix of British and American spellings. This should be enough for me to hit the FIX GRAMMAR button and conduct a test. I’ll copy and paste and save the code for what I’ve just written as I expect this new AI capability will be making live edits, although I hope I’ll be left the agency to reject and revert. Separately, I wonder how far I’ll be able to get before I’m asked to pay for some kind of AI token?

Here’s what was returned. The FIXED grammar was displayed in a box in the sidebar with options for me TRY AGAIN, DISCARD or REPLACE which was the preferred.

TEST 1: FIX GRAMMAR

ME

ME, FIXED

Clicking on the AI STUDIO button opens a right sidebar with four pre-set options and one custom option – see below right. The pre-sets are IMPROVE, FIX GRAMMAR, MAKE SHORTER and MAKE LONGER. The blue button at the bottom of the screen opens an AI chat. I already know that MAKE LONGER will make a text longer without adding any new information or content and that MAKE SHORTER will do just that, only less noticeably. FIX GRAMMAR, I’m in two minds about. Lately, I’ve been using some experimental grammar in the hope it will make my writing easier to identify as not AI-generated. I’m also more relaxed about typos for the same reason, as well as using a mix of British and American spellings. This should be enough for me to hit the FIX GRAMMAR button and conduct a test. I’ll copy and paste and save the code for what I’ve just written as I expect this new AI capability will be making live edits, although I hope I’ll be left the agency to reject and revert. Separately, I wonder how far I’ll be able to get before I’m asked to pay for some kind of AI token?

Clicking on the AI STUDIO button opens a right sidebar with four preset options and one custom option – see below right. The presets are IMPROVE, FIX GRAMMAR, MAKE SHORTER, and MAKE LONGER. The blue button at the bottom of the screen opens an AI chat. I already know that MAKE LONGER will make a text longer without adding any new information or content and that MAKE SHORTER will do just that, only less noticeably. I am in two minds about FIX GRAMMAR. Lately, I’ve been using some experimental grammar in the hope it will make my writing easier to identify as not AI-generated. I’m also more relaxed about typos for the same reason and am using a mix of British and American spellings. This should be enough for me to hit the FIX GRAMMAR button and conduct a test. I’ll copy, paste, and save the code for what I’ve just written, as I expect this new AI capability will be making live edits, although I would hope I’d be left the agency to reject and revert. Separately, I wonder how far I’ll be able to get before I’m asked to pay for some kind of AI token?

“Not bad!”, says smug me. It “fixed” I’m on the sixth line to I am but left I’ve as I’ve on the line after but I can see the sense in that. My boldface was removed and I can also see how that makes sense as those words were already capitalized for emphasis. However, I capitalized and boldfaced those words as they were button names. It was more about how words look rather than their grammar.

TEST #2: IMPROVE

ME

ME, IMPROVED

Clicking on the AI STUDIO button opens a right sidebar with four pre-set options and one custom option – see below right. The pre-sets are IMPROVE, FIX GRAMMAR, MAKE SHORTER and MAKE LONGER. The blue button at the bottom of the screen opens an AI chat. I already know that MAKE LONGER will make a text longer without adding any new information or content and that MAKE SHORTER will do just that, only less noticeably. FIX GRAMMAR, I’m in two minds about. Lately, I’ve been using some experimental grammar in the hope it will make my writing easier to identify as not AI-generated. I’m also more relaxed about typos for the same reason, as well as using a mix of British and American spellings. This should be enough for me to hit the FIX GRAMMAR button and conduct a test. I’ll copy and paste and save the code for what I’ve just written as I expect this new AI capability will be making live edits, although I hope I’ll be left the agency to reject and revert. Separately, I wonder how far I’ll be able to get before I’m asked to pay for some kind of AI token?

Let’s examine the scope of the necessary improvements. Selecting the text opens a sidebar on the right featuring four presets—Improve, Fix Grammar, Make Shorter, and Make Longer—alongside a custom option. At the bottom of the screen, a blue button initiates an AI chat. I have observed that the ‘Make Longer’ preset increases word count without adding substantive content, while ‘Make Shorter’ condenses the text more subtly. I am hesitant about the ‘Fix Grammar’ feature; recently, I have experimented with non-standard grammar and a blend of British and American spellings to ensure my writing maintains a distinct, human-centric voice. However, this context provides an ideal opportunity to test the tool. I will save a backup of my current draft in anticipation of live AI edits, though I hope for the autonomy to review and revert any modifications. Additionally, I am curious about the usage thresholds before a payment for AI tokens is required.

Unpicking this won’t be as easy so I’ll do it line-by-line in a table. Ive already explained why I capitalized the button names but WordPress now doesn’t see them as having anything more than grammatical meaning.

ME	ME, IMPROVED
Clicking on it opens a right sidebar with four pre-set options and one custom option – see below right. The pre-sets are IMPROVE, FIX GRAMMAR, MAKE SHORTER and MAKE LONGER.	Let’s examine the scope of the necessary improvements. Selecting the text opens a sidebar on the right featuring four presets—Improve, Fix Grammar, Make Shorter, and Make Longer—alongside a custom option.

I hate this first sentence, especially the chummy Let’s! This isn’t a TED talk. I’m not sure if the structure has been improved or not. Either way, I would have used parentheses rather than em dashes because this is text to be read by a reader, not spoken to a listener.

The blue button at the bottom of the screen opens an AI chat.

At the bottom of the screen, a blue button initiates an AI chat.

The AI sentence is clearer but the added comma breaks the flow unnecessarily.

I already know that MAKE LONGER will make a text longer without adding any new information or content and that MAKE SHORTER will do just that, only less noticeably.

I have observed that the ‘Make Longer’ preset increases word count without adding substantive content, while ‘Make Shorter’ condenses the text more subtly.

There’s been a shift in meaning between my less noticeably and AI’s subtly. Less noticeably is just an improvement on a bad situation but subtly implies something more artful. My meaning was that the results of MAKE SHORTER are less egregiously noticeable. The quote marks seem to be indicating something outside the grammatical meaning of the words. This is reasonable.

FIX GRAMMAR, I’m in two minds about. Lately, I’ve been using some experimental grammar in the hope it will make my writing easier to identify as not AI-generated. I’m also more relaxed about typos for the same reason and using a mix of British and American spellings.

I am hesitant about the ‘Fix Grammar’ feature; recently, I have experimented with non-standard grammar and a blend of British and American spellings to ensure my writing maintains a distinct, human-centric voice.

If our AI did a bit more cooking it would know that mixing is not the same as blending. 1) I want a plate of mixed nuts to go with my blended juice. 2) My mixed spellings show inconsistency but Mid-Atlantic is a blended accent. More importantly, I don’t like the tone of this sentence. Is AI telling me how to have a human-centric voice? Grrr. Most student-led AI text I’ve seen errs on the side of an overly-chummy TED-talk but this errs in the other direction, suggesting that the unspoken default of AI-generated text is to mimic the tone of the assumed writer so as to reassure the assumed author as well as fool the assumed reader. It would make commercial sense for AI to be designed to do this.

This should be enough for me to hit the FIX GRAMMAR button and conduct a test.

However, this context provides an ideal opportunity to test the tool.

Context? Ideal opportunity? Tool? No, no and no. This sentence in the “improved” version follows naturally from what went before, but is drifting somewhere else, as if it’s listening to itself more than me. This is not going to end well.

I’ll copy and paste and save the code for what I’ve just written as I expect this new AI capability will be making live edits, although I hope I’ll be left the agency to reject and revert.

I will save a backup of my current draft in anticipation of live AI edits, though I hope for the autonomy to review and revert any modifications.

I bristle at the I will save and the though I hope for the autonomy but no, that’s not the situation I was describing. I was afraid the AI tool would automatically change what I’d written and I might not be able to revert. Saving a draft would have been pointless. The AI wrongly assumes we are collaborators. We are not.

Separately, I wonder how far I’ll be able to get before I’m asked to pay for some kind of AI token?

Additionally, I am curious about the usage thresholds before a payment for AI tokens is required.

This is awful. I wonder how many academic papers this has been trained on? It’s rephrased my comment but no, my comment wasn’t additional but separate from everything that went before. It’s an extra thought rather than an elaboration of the previous one.

Overall, the AI improvements to my text make me sound far more pretentious than I like to think I am.

I fed that sentence back but its grammar could not be “fixed” or the sentence “improved” upon. I even asked in simple words in the box provided “Is there a way this can be better said?” but there was no response. It looks like I’ve used up all the free tokens I didn’t even know I had. Nevertheless, I did ask WordPress AI to generate a feature image. I specified a 16:9 photographic image illustrating a human-AI writing collaboration and, sure enough, I’d hit the limit.

Wearily, I hit the Upgrade plan button. The Complete package is S$32 the first year and then S$65 per year after. The S$ is Singapore dollars because my host server is in Singapore. S$65 is equivalent to US$50.88 or in Chinese yuan CN¥345.33 per year. This is about the same as one Starbuck’s grande latte every six weeks but all this tells us is that Starbucks charges far too much in China.

I then went and made some more edits to a post that wasn’t published at the time and discovered I still had some tokens left. It looks like my Free package has a limited number of requests per post but most likely for some limited but unknown period of time. I asked WordPress AI to improve my final paragraph.

For now, let’s not even think about how AI can turbocharge this. Calling for an architecture that fosters more emotional engagement may be intended either as heroic defense or a cynical selling point against an architecture substantially generated by an AI not known for its emotional intelligence. AI-generated art, music and writing already exist and, as with art, music and writing, all architecture has to lose is some emotional bandwidth but who’s going to even notice?

And this was the AI-generated rewrite.

For now, we should set aside the ways AI might augment this process. The call for an architecture that fosters emotional engagement can be interpreted as either a principled defense or a cynical marketing tactic against designs largely generated by AI—a technology not recognized for its emotional intelligence. AI-generated art, music, and literature are already a reality; as with these fields, architecture may risk losing some emotional depth, but the question remains whether anyone will truly notice the loss.

It’s out of its depth. As it was earlier for less noticeable vs. subtly, augment is not a substitute for turbocharge. These things have never been said before so what gets returned are rephrasing that are unnaturally smooth. Botoxed text. Botexted? Botext? This last example is okay ish if I ignore the pompous academic tone but I really think I’m going to have to pass on this opportunity to be part of the future. I prefer my not-improved style of writing because there’s still human fun to be had with language and writing and some of that enthusias,m shows with leaps of logic, being relaxed about divergent grammar, bumpy parsing, made-up words, inconsistent spellings and the occasional typo. I don’t wish for an EXPAND THIS THOUGHT INTO 1,500 WORDS button with “in the style of misfits’ architecture” as an option. There’s no fun in that.

My problem with AI writing is that whoever made this “tool” wants us to think writing is something onerous. We had a similar situation exactly one hundred years ago when disingenuous prophets wanted to relieve artists and craftspersons of the mindless drudgery of making things and too few and too slowly at that.

• • • 

1. https://mypoeticside.com/poets/su-tung-po-poems ↩︎
2. https://www.tropicalhainan.com/the-life-and-legacy-of-su-dongpo-poet-politician-and-exile/ ↩︎
3. There are many renderings of this poem. As is the way with translation, some are more literal while others emphasize the essence of the thoughts. Rhyme and rhyming patterns aren’t so important in Chinese poetry. This next link is to a comparison of various translations. https://thebamboosea.wordpress.com/2014/07/30/su-tung-po-tracks-in-the-snow/ I combined and worked them into the one above that works for me. It looks freestyle and is but, in English, adhering to the strict rhyme and choice-of-word rules of Chinese Song Dynasty poetry is never going to work. Meter simply isn’t something one plays with, as you can guess from how the poem looks in Chinese. ↩︎
4. https://allpoetry.com/poem/8536827-Remembrance-by-Su-Tung-po ↩︎

• • •