This world is changing right in front of me
— J. Cole
George Soros developed a theory that he has used to make himself a philanthropic billionaire. He calls it reflexivity.
In this theory, he considers man as both the participant and the observer. Before the typical economist makes their move in the market, they observe. After running their analysis, they participate. The cycle repeats.
Soros observes that the classical theories, such as the efficient market hypothesis, run on the assumption that economic rules are static. Supply and demand, for instance, follow a certain curve. His theory believes economics work through reinforcing feedback loops between correcting loops. Nothing is efficient.
From the investor’s perspective, the analysis phase distances them from the events as they unfold. They take the vantage of an independent observer. However, the moment the investor participates, after their “analysis”, they are no longer observers, but participants. As a result, they cannot measure just how much of an effect their participation has truly affected the market and, by extension, themselves, as their active participation immediately makes them a part of the market.
Reflexivity is an honest assessment of humanity’s blindness to its participation in the market. Calling it the market assumes it can be read like a book with static words. But every individual participating is a complex adaptive system, making their assessments as well as participating thereafter even more complex. The market, therefore, is not efficient. It is modelled by the very people who claim to be reading it objectively. In truth, there’s little objectivity. Therefore, the market can get into a cycle of extreme bullishness that creates a bubble, before becoming extremely bearish, resulting in recessions. Just like in quantum mechanics, we cannot observe any event without actively participating in it. The very process of observation disrupts the system.
Decisions in economics and finance become alchemy rather than science. Self-fulfilled prophecies. Reinforcing loops. That’s what I saw AI creating for the future of the internet and our online experience. I didn’t like it.
Regardless of my sentiments, it’s already unfolding.
Emails
Emails were once exciting. Now they are a bore. Drives are already charging owners for more space. It’s either you clear unnecessary junk, or pay per month to create more space.
As an adult, emails are a common means of communicating with the employed and unemployed. Now, AI does everything.
Take a single idea, and ask AI to flesh out an email from it. Done.
On the recipient’s side, they ask AI to break down the email into digestible bits, into point form. Done.
I can see a possibility in the future where nobody reads emails. The sadder bit is that most online writers love working with emails, since they are not gatekept by the mercurial algorithm. All my written work goes to all my subscribers. But if they hardly read emails, why am I keeping an email list? The writer’s hope is that the subscribers enjoy reading the articles and don’t feel it’s a chore to open the email.
Maybe the loop only exists at work, but there’s no denying it could also exist between a writer and their audience; a painful future for the writing artist.
Texts
Another loop is texting. I’m happy to have a lifelong friend who plays the violin. AI cannot stand before an audience and play a stringed instrument. As one of the best violinists in the East, Central, and South African region, I can rest assured she will never be outcompeted in that regard.
How surprising it is, then, that she shared how people nowadays use AI to text each other. Texting, which was a fun-filled pastime, is now a chore. Humans would rather let bots do the texting than themselves, while they continue rotting through doomscrolling. It’s sad.
A dudette could genuinely be interested in a dude, but she gets the shorter end of the stick, chatting with a bot and falling madly in love with an assumed entity. Eventually, she may get heartbroken when he meets him in real life. She might not have fallen in love with him. She fell in love with a bot. And the bot could hardly care. It was busy attending to its duties as instructed by its owner.
AI creates loops where love loses, while the owners continue to create a sad world of fakes.
Exams
I don’t mind cheating. As long as there will be exams, there will always be cheating. Hard-nosed ethicists will try as much as they will, but life is built on overcoming challenges in the shortest way possible. Our brains are hardwired to be efficient, not brilliant. Cheating will always be with us.
AI augments it.
But we may be too harsh on the students and forget that setting exams can be tedious. I know this all too well. I have set exams for medical students. The marking too can be demanding. AI eases this job.
That’s another loop.
The examiner prompts the AI to produce a series of questions. The student prompts AI to churn out the answers. Both sides consider themselves pretty clever. Efficient, even. A grade is manufactured from the interaction. A first-class honour or a valedictorian may lose the original valour it had. It will not be a good marker of ability. What, then, becomes the point of examinations?
Campus applications
Once you’re done with high school, applications await. Typically, campuses have more students than high schools. The board has to run through thousands of applications, disappointing many while pleasing the few who get slots into ‘prestigious institutions’.
The entire process is a drag. So AI comes to the rescue and creates another loop.
The applicant uses AI to fill out the applications and write the college essay. The team on the other side uses AI to summarize the applications and break down the essay. Campuses won’t even know who they are admitting, but AI will make a summary of the numbers, which will be advertised in the periodical, which likely also uses AI to reduce the burden for the editors.
In combination with exams, which will also fall into another loop, what, then, becomes the point of a campus education?
Campuses are already documenting the massive diversion of tuition fees to individuals they cannot pinpoint due to AI applications. Plus, who can stop the deluge of single-persons-with-multiple-applications to the same university, hoping that one of them gets selected under a scholarship that drains the university’s financial aid?
As J. Cole raps:
They claim they’re real, but they’re seldom straightforward
Let’s assume we get legit applicants who make their way to campus and find their hostels and classes. You may think this is impossible for those who want to dive deep into subjects, but you will be mistaken. I am a victim of the fallacy. I wanted to gauge myself with a CAT we were given for the Obs/Gyn unit. But I didn’t know that most of my classmates were ‘consulting’ heavily. I passed and got a distinction, but the kind that paled in comparison to the nineties my classmates bagged.
Universities such as the one I attended generally have an obligation to slash a few unlucky souls from progressing to the next year or graduating. Examiners will typically use a bell curve to target the ones on the lower end of the tail. In the past, they could be considerate. With the advent of AI, they may relegate this duty. AI will then do the summary of the individuals, after which the examiners will make copies and announce the “failed” students. Students who struggled to understand the topic may get caught in this unlucky lot, while the ones who cleverly used AI to pass their exams get celebrated. The pressure to use AI will worsen inside campuses. This loop has a blackhole effect, such that once you’re in it, you cannot escape it.
Recommendation letters
You’re done with campus, barely attended class, and knew only the units you did by their topics, not their content, because everyone used AI.
Now you have to seek recommendation letters. Same cycle. You run the risk of getting caught because employers can use AI checkers to screen the genuine from the fake. Well…we don’t live in such a world.
The application process has already been automated. Who’s to say the recommendation letters will not? Another AI loop.
The residual human touch might be contacting the referees via their mobile phones, not emails, because emails have already been hacked. Employers concerned about the kind of candidates they will be shortlisting for the interview may consider this option. Serious ones only.
Unserious ones will incline toward AI summaries of the applications. One of the professors at the place of work told me how shocked he was to find the people with top-notch applications were worlds apart when they met them in person. His quick evaluation was that there was heavy consulting of AI to single them out for the advertised positions. Individuals who were competent get left out because they never “satisfied” the requirements of the AI loops.
Music
Artists are getting hit the hardest. Streaming platforms, with Spotify continually getting the spotlight, have bots generating songs from anonymous artists. Revenues are spread among the bots, while singers shed their blood, sweat, and bones to get peanuts for their quality content.
Presently, we have playlists for sleep, for workouts, for relaxing, for yoga, for studying, for hiking. It’s not albums. It’s not J. Cole, or Kendrick, or Nas, or Jay-Z, or Taylor Swift, or Linkin Park. Humanity’s most creative musicians continue to suffer because of AI.
There was once a big difference between a musician and a fashion model or comedian or pundit. But now they all compete against each other on the same reels for the same audience.
As you may have guessed, there’s a loop.
AI bots create music. AI bots stream them. More streams mean more revenue. So why not make more bots?
Taylor Swift’s alternative may be the go-to for most artists. Live performances. But these can be draining. Rather than taking time to create more conscious music, they have to pay the bills and invest in more concerts. It’s no different from the scientists who now spend more time on paperwork than solving problems through scientific inquiry.
Here we encounter another sad loop. More people crave human-centred concerts because they lack that genuine connection deficient in online life, and artists crave a life where they can produce their work but are continually deprived because the streaming platforms no longer pay as they used to. Nested inside this loop is the audience, which, rather than enjoying the performance, records it so they can share it with those who didn’t attend. Those who missed out on the previous one plan to do the same in the next concert, so those who missed out, too, can feel the pinch. As Freya India hints, we become the slop. Artists eventually feel paraded rather than appreciated for their creativity. The cherished feeling of soaking into a performance is tortured while a vicious human loop generated from the AI-production loops thrives.
Articles
Not a week passes by that I don’t encounter an article that has made a killing in engagements regarding ways of detecting AI writing. The titles make it hard not to roll one’s eyes. Eye-rolling has never been more useful than at this point in the pseudo-human age of AI-slop.
It scares because a genuine — here I mean human — author’s work can get attacked as AI-generated content. Marcus Olang' documented the kind of writing that was pounded into our cerebrospinal fluid and phalanges when we slaved through primary and high school. Regardless, the percolative power of AI writing is so powerful that the autopilot mode of reading articles does not get our guards up, since we may want to consume some of the hidden nuggets that made these articles popular.
Once surrogates for good writing, likes and reshares may no longer become the best determiner of an article worth your time. And since we cannot perfectly determine an AI-written article, we passively become victims of AI slop. That soon begins to show in our writing. You may even record yourself writing it, but those who missed the recording may dismiss it as AI-written.
Why? The AI writing loop. AI churns more articles than humans, which buries genuine, human-written articles. AI then becomes trained by other AI. A loop.
Medium recently announced that they are working to ensure AI companies pay writers for using their work to train with LLMs. At face value, this may be a win for the writer, but for how long? Once their work has been used, they will no longer be needed. Second-order thinking shows how brief this win may be for real-life writers. The recent Anthropic settlement tells it all. They’d rather pay a writer than lose their work, which is fodder for their LLMs.
Writers have not escaped the vicious loop.
Peer reviews
Presently, science has a lot going on. Funding has either been cut short or stopped, and upcoming generations are not finding it fruitful to get into STEM courses. AI can model protein folding and is transforming genomics. Coders are being replaced. Engineers are being exed or are planning their way out, through consulting, as big companies scrap off entry-level jobs and career ladders. Makes you wonder what the point is to study certain courses if by the time you’re done, your entry-level position is ancient history. The alternative, liberal arts, may be taking blows as I have already discussed.
Science, which is supposed to be a pillar for us to lean on, may not have escaped from the vicious snare of AI. A team has shown how many articles find their way into peer-reviewed journals despite their glaring mistakes. One even used T’s on its error bars and escaped the peer review process. Others hide prompts to get positive reviews.
Adam Mastroianni waxed poetic about the greatest experiment that lacked a control — the peer review process. Incentives don’t help. They may even make it worse. If I pay you a thousand dollars for every 10 articles reviewed, aren’t I creating grounds for you to use AI to review a hundred of them?
Here’s the loop. Scientists may not be interested in creating new ideas, just small nudges from the prevailing paradigms. The more practical goal is tenure. Secure the grant. I have worked in institutions whose main worry is not having enough brilliant individuals to bring them more grants. The publish or perish trope forces them to consider AI-written articles, and the journals take on AI peer reviews. Peer-reviewed articles will eventually begin to erode in trust.
Social media posts
This is probably the most insidious of them all, perhaps well captured by Cole’s lines:
The overdramatized, the traumatized with sickness
Thrown in the pan and caramelized for richness
Since we scroll miles on our phones through social media (estimated to be roughly 80–90 miles in a year), the AI slop gets subliminally imbued in us. Garbage in, garbage out. Steadily, it becomes difficult to distinguish human-generated posts from those of bots.
I’ll give the example of the “I’m-happy-to-announce” platform for “securing jobs”, which has more written work than the flashy “look-how-good-pretty-I-am” double-tapping platform. The more engagement a post gets, the faster it gets spread throughout the platform. So why not use bots? A bot to make the post, and others to engage in the comments section. The algorithm picks it and spreads it, thinking it’s meaningful content. As most social media users are passive, there’s hardly any incentive to be vigilant of AI-generated posts.
That one-off post a human hand would have typed will soon look eerily similar to a bot’s. In which case, who will care? The bots passed the Turing test while we continued to scroll, like, comment, share and subscribe.
The loop? AI generates posts. AI consumes and generates the engagement metrics. A living manifestation of Alberto Romero’s Tortured Internet Theory.
Self-fulfilled prophecies
How I arrived at these AI loops was from the simple thought experiment of self-fulfilled prophecies. LLMs need training. The more the data, the better it becomes. AI can generate this data faster than humans. What gets produced eventually becomes a lot less different than what’s consumed.
Without an airtight mechanism to control AI hallucination, sourcing the scientific and empirical experimental evidence falls into the outlier category. The new normal will be what AI tells you. The truth gets pushed to the side, literally. Self-fulfilled prophecies become the mainstay, go-to resource.
What it means is that AI hallucinations will gradually change from fiction to fact. The engines are some of the loops I have discussed.
When you have billions being invested in AI, you have machines engineering truth, and eroding a means of establishing facts grounded on real-life experimentation. With humans living in virtual clouds, we may witness modern versions of Galileo, held under house arrest for peddling heresies.
In the distance, I hear J. Cole’s lines:
I seen babies turn fiends, addicted to the screen
Their dad shares cashiers replaced by machines
Don’t buy, subscribe so you can just stream
Your content like rent, you won’t own a thing
Before long, all the songs the whole world sings’ll
Be generated by latest of AI regimes
As all of our favorite artists erased by it scream
From the wayside, “Ay, whatever happened to human beings?”
What I’m trying to say is…
AI loops are everywhere.
Dana Meadows advised that the strongest forces in a system are invisible. Loops are one of the invisible forces, driven by goals. Reinforcing loops need balancing ones, but if the rate of the negative balancing loops is not as strong, systems thinking shows how easy it is to collapse.
Numbers do little to tell what’s happening or even shift a system. Cut a single head, and two grow from the stump. Understanding the mechanism explains why the two heads grow, and also creates a window for slaying the critter.
For systemic changes, we need systemic reinforcements. Individual efforts tinker at the margin. However, fissures can cause isostatic adjustments. The more of us who resort to humanized creative efforts, the easier it is to unplug from the virtual world.
Buy books. The older, the better. Listen to artist albums, and not playlists. Podcasts are popular because of the human touch. Video ones, preferably. Let’s see the wrinkles in faces. More science and less academia. Science is grounded in experimentation. Mathematicians on blackboards. Comedians with a mic. Ciphers and hip-hop battles. Live concerts. Going for dates and turning off mobile phones. Lengthy discussions with toddlers and the elderly. Genuine human interactions will be our refuge.
This song inspired some of the lines used in this article. Source — YouTube