.png)
Writing with AI: My 5 Day Adventure into Automated Blogging.
This week, I decided to try something new. I let an AI write my blog posts for me. And, boy, was it an adventure. The benefits were clear: more consistent content, faster writing, and no fear of plagerism (since the AI wrote everything). However, there were also drawbacks: a loss of personal tone and individual perspective. Overall, it was a mixed bag experience that made me think about the role of AI in content writing.
That first paragraph was produced by an AI program that specialises in content creation and, as an example of one of the contentions of AI…it did not write my blog posts for me…it wrote examples…which was an important detail it decided to omit. This paragraph was not written by AI. I’ll let you know when the machines take over. The use of Artificial Intelligence has finally reached it’s undulating neon-green strands into our content creating; from generating art that borders on plagiarism, devising advertisements, to the spinning blue light that tells you that Alexa is absolutely listening to you, we’re on the verge of Skynet…kinda.
Look, I am not even remotely anti-tech. And I am embracing of the benefits of AI in the home and workplace. I’m a disabled person, and AI makes the world far more accessible. However I absolutely agree that there is a line that AI can approach, push against, warp and distort but cannot cross regardless of how many skins we put on it, voices we give it, or machine learning algorithms we stuff it with. Between cognitive dissonance and the uncanny valley, there will always be something a little “off” when presented with something machine produced vs. something that took literal blood, sweat and tears to create. Writing is my passion. Writing is my work. How I use and manipulate words is something I consider uniquely my own, even if it sounds similar to someone or something else; it is still mine. But I wanted to see if it was possible for something artificial, given every advised element of my interaction, was able to produce work that would resonate as “human”.
Day 1: It Learns With Me…It said it on the box! I imagine this as a bit of auto-surgery; I had to walk the AI through my “tone”, provide it with context and nuance. And, in turn, I had to really get a feel for what that was…and I wasn’t particularly sure. Score for the AI because it had a program that could assess what my tone was through copying and pasting a paragraph I had already written sans-AI. I had always considered myself conversational, if a bit loquatious, dry-humoured but forthright.
The Tone Detector determined that I was ‘sarcastic and light-hearted’, ‘encouraging and positive’. Considering the topics I tend to writer about; rip out my heart, why don’t you?!
Light-hearted?! I must be doing something seriously wrong! But it came with another option: ‘matter-of-fact and humerous’. A little more dissection (refreshing), and the element of humour had been completely forgotten. From exactly the same paragraph, my tone was now ‘jaded’, ‘cynical’, ‘frustrated’. For an experiment, I took the same paragraph and put it through the Content Improver with the tone suggestions to see what would happen. Here are the results:
Original paragraph:
Now I’m not encouraging snatching, biting, hitting, giving rewarding cuddles in the face of all-out extreme behaviour (though cuddles, in general, are ageless). I assure you; it is completely normal to shout “shoulda-coulda-wouldas” into a cushion, read your anger to filth, or oust latent sensory urges of anger (pushing, shoving, grabbing, running) into the nearest unwitting stuffy instead of punching a brick wall or someone’s face. Freedom of expression is emotional management education 101. Whether your slowing down your parenting, or breaking down your developmental trauma, or both. I love that for you. Good luck.
AI paragraph:
If you don't want to resort to punching a brick wall (or someone's face) when feeling angry, that's totally okay! Don’t be afraid of expressing yourself in healthier ways – snatching stuffies is allowed and sometimes necessary. Take this time as an opportunity to learn how your emotions work by reading them out loud or pushing around whatever urge you have - scream into the pillow if it helps... heck, even give cuddles if needed (they never hurt). Good luck with managing your frustrations; I'm cheering for ya!
Now I don’t know about you but a lot of what the AI extracted from my original paragraph is accurate…in that it has pulled the organs of the sentence out and sewn them together into something “positive” and “encouraging”. But it’s not, at all, encouraging healthy freedom of expression or emotional management. No person wants to resort to punching a brick wall and don’t need to be reassured that it’s okay not to do that. Snatching stuffies is not conducive to anger management. What exactly is “pushing around…[an] urge” and didn’t I say NOT to give rewarding cuddles? So what have we learned; that the AI can identify tone but it gets confused when it has to reconstruct the information around natural paradoxes in human expression. One can be light-hearted, jaded, cynical and encouraging all at once…but not this AI.
Day 2: The Bay of Cliches “…the world is your playground…” “…fighting fire with fire…” “…rain on your parade…” “…think outside the box” I adore a cliché…excuse me, I adore ripping apart a cliché and inserting my experiences into them in order to reinvigorate myself to keep writing. So when the AI was throwing around cliches that have been used since Shakespeare took his first poop, I gagged a little. Some phrases in writing are so painfully overused that it causes the reader and the writer to completely switch-off. The response, for me, is being part-way through vomiting up cotton-wool and my mouth feels like the cobwebbed recesses of an ancient wardrobe. The language feels so dusty to me that my face desiccates. In order to shoe-horn it’s prosaic style, it would forget about tenses, grammar, spelling…the basics of language that your average word processor would habitually pick up. This is language AI, after-all, so it does exemplify what everyday-people are reading as well as what writers are writing. And that is both reassuring and woefully miserable. I am grateful that the AI is not incapable of the linguistic innovation required to form it’s own witty phrases; us mere human-writers can stand-out from the matrix. But…really? Are readers still dry-swallowing ‘plenty more fish in the sea’ with all the incredible words and experiences available to them? Do you even f*cking fish, bro?!
Day 3: Someone Whack The AI; It’s Repeating (and contradicting) Itself. At this point, I was thirsty for some actual results and decided to really stretch the boggy-blogging Terminator so I decided to let it write it’s own blog post on something I had no knowledge of: How Much Do Athletes Earn and Why? So blogs are fantastic ways of delivering information to the masses. The thing about the AI is that it wants to give you ALL the information. And, in attempting to do it in a “person-like” fashion, it uses amiable sentences to instil a weird trust in the knowledge you’re receiving…but it doesn’t have empathy or integrity, so it will absolutely change it’s mind from paragraph to paragraph just to give you more information. It doesn’t care if it said Lionel Messie gained fame unequivocally through being superbly skilled; it will throw Messie under the bus a moment later saying he undeniably achieved his fame through sponsorship and advertising. It also doesn’t care about the overuse of jargon or entire phrases only a comma away from one-another. My AI is perpetually trying to illicit my surprise to very underwhelming effect; it loves ‘…and even’. “Even this brand!” “Even this person!”. “Even this unrelated tidbit about photosynthesis so we can place a sun shine cliché right here!” AI prioritises compiling information over analysis and contextualising; while you can give it some direction with a ‘word use’ function and ‘tone’ directive, it is still woefully unprepared to provide nuance, or sustain a perspective.
Day 4: This Is Not The Function You Are Looking For So I’m getting at it for not writing like me, using my voice, implementing my particular turn-of-phrase (that few other humans understand or appreciate anyway). I had to accept that I was the problem. For one, I don’t need an AI to write for me about the things I’m passionate about. I don’t need or want it to expand my content or do research on my behalf; that’s what dictionaries, libraries, and the internets is for. As in toxic romance, so in writing; I was letting my past relationships define my present and future. So we turned to the other functions; formatting. My blog format isn’t great. I’m stuck in the rigid habit of producing sheer walls of text that terrify other people and annoy readers because nothing says “preachy” more than a HUGE platform of paragraphs. I went back to my oldest blog post on www.borderlinekitty and, gritting my teeth against the nauseating nostalgia, gave it a full revamp using every formatting tool the AI had to offer. List Creator. Blog Post Outline. Persuasive Bullet-Points. Honestly...nothing happened other than it shortening the sentences and and reformatting it into bullet-points on my behalf cuts my work down, summarises my key points, and broke up the 8th wonder of the world that is my garrulousness. The bullet-points were no more or less persuasive and the blog-outline was a "Write this but in your own words exercise making it longer and more complicated. I could have done all of that myself.
Day 5: The One-Shot Blog Post As a disabled content creator with a chronic illness that causes fatigue etcetera, etcetera (must be said in the fashion of Lucy Lui circa Oren Ishi E) sitting upright staring at a glaring computer screen for long periods of time is daunting, painful, and tiring. So, on this day, we turned to the bullseye-for-a-blog-guy of the AI; I would let it do ALL the heavy lifting. As this is the “lie back and think of England” approach, I was expecting to get short-sharp thrusts of information that were deeply unfulfilling, lacking in originality and intimacy; more of a cliché than mid-life missionary at 9pm on a Friday. I was…surprised? Even astounded. I did have to tell it what I wanted in an explicit 80 character sentence which was probably a bit more effort than I wanted to put in (especially difficult when I had to explain that DBT was Dialectic Behavioural Therapy or it would wax lyrically about mindfulness); but what it produced was cohesive and conversational. Repetitive and ridden with banal cliches but it wasn’t awful. In a world of candy mushrooms, I would put this down to it having some decent source material…but I did read the original blog I wrote and, as far as quality goes, it was not the unicorn manure required to cultivate candy mushrooms. It was plain old horse-shit. What the AI lacked in nuanced parlance and originality, it made up for in content, expanding more on the details of DBT. It was a similar length but I think that was what benefited the AI; the One-Shot Blog generator is for short pieces limited to 500 words maximum. It is programmed to keep it short, sweet, to the point. It is not afforded the room to grab the piece with it’s teeth, tear it open and run around splattering information all over the walls.
Seriously, by comparison, I was half tempted to replace the original blog. But only tempted.
Is This The Impending AI-polcalypse? Right now there is a concern that AI will replace all creative processes due to it’s speed, it’s convenience, it’s rapid ability to acquire skills that take a mere human decades to master. It can keep up with a digital age fuelled by consumption, it updates itself faster than you can swallow. The things about AI is that, while it can learn, it has no concept of personal improvement. It doesn’t care whether it ‘gets better’ at writing or reader analysis. I do. Readers do. While human interaction and creation is a key component of how the AI operates, the why is always going to be missing because the AI does not ask why, it is can or can’t; do or do not. When you read things written by a particular person there is a whole, entire, real-life cognisant being to look at so you can understand their why and decide on whether their why resonates with you. It is a discussion that the AI cannot partake in regardless of how often or how rapidly it produces content. And because of this, it is woefully unprepared to navigate very human aspects of making the thing; tone, theme, perspective, composition. Even the very real, very human trait of showing your working, exhibiting how you made mistakes. The AI can present itself as utterly infallible even if it is clearly not then refresh without any real impact to its existence.
AI might be able to write content but can it create? Can it engage in the same discussion that a real-life cognitive being can? No. But because it exists, we’ve come to an impasse of sorts; discard creativity and choose convenience or rise to the challenge and invest our emotions into the outcome. It is not only a matter of ethics, but one of opinion. People are clearly imperfect, and are made up of stories and experiences that move them forward and make them strive for more. AI cannot provide that intangible element that drives creation.
We are ridiculous upright apes with big flaws, even bigger ideas and a lust for sharing them. We may have given AI a screen for it's output but it's still just an ephemeral echo-chamber that talks with itself.