NIJAS SALIM

Models were pretty once – My job No job no more – AI Slop Shop – AI Can’t Count – Survival of the wittiest

The world you belong to is changing at breakneck speed. Or so they claim. Nothing has excited Wall Street, founders, executives, makers, the curious & the bored like AI has in recent history. AI is everywhere. It is in Word documents, it is in Excel sheets, it is in design, in video, in people that you think are real people but are not. AI is in the music you listen to on Spotify by an artist whose soul is defined by compute.

Some say AI will replace you. Some say AI is another buzzword you attach, like a pair of bells to your cat, to raise valuation.

To top all that, you have the inter-AI wars. Mark Zuckerberg went AI personnel shopping and dropped a few Bs. But the-new-AI-model-that-replaces-all-AI-models changes like the gloves of a doctor. It has settled into a sort of equilibrium of an English summer, where change, like rain, comes with only one question: when is it happening next? We have the Gemini-OpenAI-Anthropic revolving door with the interjection of a Chinese model momentarily.

Models were pretty once

Here’s the quick brief –

• John Von Neumann worked on neural networks in the 1940s.
• AI has theoretically been worked on since then. (We had a course on it in 2002 at our university.)
• The computing power that GPUs (read NVIDIA) & massive amounts of storage provided changed the world of AI simmering underneath.
• People started tinkering with different models (AI models are like feeding random things, without a measuring tool, to a food processor and hoping to get something edible out of it. Too much salt and you’ve got to spit it out).
• The huge processing power allowed people to scrape huge data available because of the internet.
• LLMs (or large language models) are just non-deterministic (non-deterministic systems are systems that might have different outputs with the same inputs) models.

LLMs know the probability of what text is supposed to come after a particular word or sequence of words. “It is raining, let me open” has “the” 45% of the time, “my” 25% of the time, and so on, as the word coming next, which it has learnt by scraping data. Now, depending on how the model is tuned, it spits out the next word (you can see the exact distribution here). So essentially, it does not think outside the box.

You can read Stephen Wolfram’s explanation here or just watch one of his videos.

My job. No job no more

A lot of people hate their jobs. And the reasons that a lot of people hate their jobs are that it becomes routine, rote, boring, tedious. Creativity has been sucked out of the major part of their lives.

My observation is that anyone that routinely, repeatedly does something via inputs that are objective and via a process they have imitated or learnt from somewhere else, without any personal value-add whatsoever, stands to lose. The process will become an n8n workflow or a Claude skill, and outputs will be churned out in seconds and not in a week.

If your job is just to search and retrieve and present, there is a problem. If your job is to barely be a blood-pumping robot, there is a problem. If your job is to just use a voice and mechanistically order, follow up, or inform, there is a problem. Anything machine-like with a heartbeat will most certainly be automated with an AI without a heartbeat.

AI Can’t Count

One of the earliest things I encountered with AI was its hallucinations. I once asked ChatGPT (2.5, I think) what the number of football fans in the UAE was. It gave me a number that resembled closely the population of the UAE. A human with basic intelligence would tell you instinctively that the numbers couldn’t be true because all the people living in a country certainly couldn’t be football fans.

But if you understood how LLMs work, this was an inevitable answer. In the absence in the data set of the occurrence of the statement “There are X number of football fans in the UAE,” it could not come up with an answer and instead outputted the number it had, probabilistically.

I give LLMs my credit card statements every month and ask it to categorize the transactions with the category spend. It always got the answer wrong (up until the last iteration of the models). Because the odds of the same combination of numbers existing in hundreds of values on the internet or in training data is next to zero, it just gave me an answer with the closest approximation from its data sets. I always had to add them in Excel to get the right answer. (They have started getting it right since September.) Now they call external tools to do the addition and fix it (credit: ChatGPT).

AI Slop Shop

Eventually everyone will get AI into their workflow, just like everyone Googles first before asking their mates. And one will see AI-generated content everywhere. Articles that are different but still the same. And they will fill the shop everywhere. Mass-produced content with economies of scale that overwhelm you. AI slop feeding AI machines generating AI slop. It’ll be like Netflix with their data points producing content that fills your time, but rarely a moving experience that gets you excited. 11pm media for when your brain can’t function anymore.

And then it will get worse some more.

And then you will get your turn if you are truly an expert. If you are creative. If you bring your human element into the mix.

Survival of the wittiest Or The Guide On How To Survive

Wit is a high form of creativity. An old joke is not a joke anymore. If it has been said and used, it is not funny anymore.

People with creativity, drive, humanity, empathy and desire cannot be subdued by any AI.

They will use AI as a tool to reduce the mundane, the non-critical, the non-essential, while focusing all their faculties on both identifying and solving problems. In figuring out a way to stand out from the slop. One outwitting, the other imitating.

So how do you survive in a world exploding in investment and tech geared towards AI?

In a world where Shein imitates and mass-produces, be the designer that changes culture. While AI slop fills social media posts, design (or generate) the one that sticks out. Become more right-brained. Become prompt engineers with your own magic potions.

Remember, it is fast fashion that made people yearn more for designers. And when you stand out, charge more. And when more people follow suit, they will be a diminishing return for AI plastic.

AI is here to stay. The question is how you are going to make sure you not only are untouched by it but become anti-fragile to it. Luxury will be redefined – experiences made by the human hand will find its status enhanced further and the question on minds will be how do we be more human.

Categorizing credit card statements is not going to stop for a while, though.