+++
title = "On generative content"
author = ["Michał Sapka"]
date = 2023-12-20T22:47:00+01:00
categories = ["blog", "update"]
draft = false
weight = 3002
abstract = "I don't believe that AI will kill us, but I strongly believe it will lessen us. Here I try to describe that by looking at impact on art, culture, tech, and potential benefits."
aliases = ["/articles/generative-content/"]
+++

It seems that generative content from "modern" AI models has been with us forever.
In reality this is still a new fad.
The feeling is here because it seems that every few days we hear new product or controversy.

However, I still remember that computers were to be unable to replace humans in _creative_ work.
That the _human_ part is irreplaceable, and machines can only reproduce.
I think it still the case.
But, somehow, artists, techies, and all kinds of other folks all over the world are afraid of loosing their means of living.
Rightfully so.
But why?


## 1. Mass-art market {#1-dot-mass-art-market}

When I was younger, I was very into cinematography.
We had so many directors with their voices - Jarmush, Lynch, Smith, Tarantino, Cronenberg, Carpenter, Boyle, Ritchie, Anderson, or Fincher.
And that's just the USA!
They had the voice - even when imitating, the movies were distinctly _theirs_.
They got chances, so we all knew them.
It is no longer the case.
The mid-budget movie is almost non-existent, and that's where the creativity strived.
The popular movies, the ones that make all the money, are indistinguishable copies of each other.
What earns the money is the same, big budget CGI fest without any real meaning or personal touch.

In photography the end product is so removed from the original, that often it is impossible to see similarities.
Everything that is not perfect is corrected and removed.
We don't see people from posters on the street because they don't exist in the same reality we do.
This is what we call _content_, this shapeless blob filling platforms.

The mass-market "art" (and I use the term here very loosely) was removed from the _human_ a long time ago.
Why take chances, when we know what will work?
And if you know what will work, why even bother with humans?
An AI can create the script, and we can reuse that 3d model of Bogart - because why not?
What we see and what we pay for is no longer _human_.
It's more machine than that - endlessly modified to be as close the blob of mass appeal as possible.
Indistinguishable from each other.

And this is where generative algorithms strive.
When we exactly know what we want to produce, where there is a mathematical equation of beauty, we're no longer in **human** creativity.

Since the most popular _content_ is made from the same mold, LLM models can create it as well.
In fact, it can create it better, as those _imperfections_ that a human might have missed, an AI can easily remove.

All the Dall-E pictures are perfect. They are impossibly complex and complying to the popular norms.

They are also boring.
It's the imperfections that make _art_, well, _Art_.
No human creation is perfect and this why computer generated _content_ can not be called art.
It's too studied, too ideal, too perfect.

But the mass market pulp is what allows _artists_ to live.
This is where they make the money.
This is what pays the bills, what puts the food on the table.
Why pay 50EUR and wait 2 weeks for a Fiver order?
You can get just as good result in 10 minute chat with a bot.

LLMs have reversed humanity.
I fear that we are getting back to the state where _art_ may come only from _suffering_;
where artists live to create that one piece which may immortalize them, but this life is not what modern world promises.
We are to no longer see _hungry_ people in developed countries.
We are to no longer experience _pain_ in pursue of _happiness_.
Why would anyone choose it?

Therefore, I am afraid that this will decrease the number of people who choose creativity as their living hood.
Without them, we will be destined to status-quo of mediocrity
It will be perfect, but it will be soulless.

Just as the mass market is now, just without anyone pushing it forward.
And with us, running the hamster wheel of never ending _content_.


## 2. Tech market {#2-dot-tech-market}

The other, very popular, use case is code generation.

GitHub Copilot can translate a short query into an evaluable code.
It is sold as a mean to automate the _boring_ and _repetitive_ tasks - creating boilerplate, configurations, loops, or simple algorithms.
But are those really lesser tasks than the big ones?

I've been a professional Software Engineer for 10 years now.
My journey has not been the typical, one where one finishes either IT college, or a boot camp.
I'm self-thought, and I joined It Crowd from other occupation by sheer luck.
The company needed _Ruby_ developer and boom - there I was.
Not the perfect candidate, but I was capable, eager and hungry.

I've made a lot of mistakes, I've wasted a lot of time, I've taken down the production on a few occasions.
All of those could have been avoided if I used a code generation.

But it's impossible of overstate how important those menial tasks were in making me into a real _Software Engineer_.

With every mistake, I learned.
With every issue, I became swifter to jump into action,
With every boring, repetitive task[^ruby] I gained insights into how stuff works.
[^ruby]: with _Ruby_ and _Ruby on Rails_ there aren't many of those but still, you do the same things from time to time.

But the biggest growth came from the most hated task of all - writing tests.
It is there where I learned how to write a usable contract; it is there where I learned the value of documentation-as-a-code[^cdac]
Would I learn anything from "hey copilot, write tests for this class"?
I doubt.
[^cdac]: very often the best documentation of a contract is the test for it.

This was what every intern/junior would do - the dirty work.
With the dirty work comes the realization that the real world is not perfect, not every code is good and not every developer is good at being a developer.
And with Copilot we are loosing it.

We are losing it two ways: first, juniors will not learn; second junior will not be hired.
We are already seeing that there are much fewer offers for sub-senior positions.
Why would it be different?
A senior aimed with code generation can do the work of many junior and one senior.
It makes _perfect_ economic sense.

But this is also the suicide of the industry.
Without new blood, the tech crowd will shrink.
Companies want infinite growth, and will not stop at anything to accomplish it.
We've seen all the visa scandals, the inclusivity actions, the offshoring[^contr].,
It's clear that all those had only one goal: to increase the pool of _cheaper_ candidates.
The typical programmer is expensive; someone fighting to leave poverty is not.
But now the competition is no longer _someone_ cheaper, but rather something that's never tired, and can create infinite number of creations, but is still _cheaper_ than a human can ever be.
When was the last time you tried to outrun a train?
When was the last time you tried to fight a machine to give your children a better start?
[^contr]: this is not the place where I want to address those subjects, but their existence is important for this article.


## 3. Maybe I am wrong {#3-dot-maybe-i-am-wrong}

But what if I am wrong?
We've seen similar things before.
Ever since the beginning of industrial revolution, more and more industries were mechanized and automated.
The machines needed someone to take care of them, to design them.
And the rest of the populi moved to other areas.

Will it be the same?
Well, how much work does an _algorithm_ need to operate?
We are still in development phase, so we see a lot of people working there.
But when we will reach plateau?
There will be a time when it will be good enough.
A moment when companies will buy it and not expect it to be better.
How many people will be needed then?
What will be left to do?

Will Universal Basic Income be enough?
It assumes infinite growth, and this may be in the hands of very few, gigantic companies.
We already see how big influence Altman and Open AI has.
There will still be rich, the 1%, so money will still be an issue.
Will we simply do yet another October Revolution?


## 4. Utopia that is false {#4-dot-utopia-that-is-false}

The best case scenario is that humanity, as a whole, will ascend.
Free from the shackles of menial task, we will all be poets and explorers.

This is what _Star Trek_[^old] is all about - a world where we no longer need to create and work to survive.
In place of that, we are free to pursue adventure, knowledge.
With nothing to gain, why one would need power?
Money?
We can focus on the humanity as one entity, to make it better.
[^old]: old Trek, at last.

In _Star Trek_ this was not created by removing jobs but by presence of _replicators_ - a device capable of creating virtually anything.
No longer food is scarce, we can simply create it.

Replicators are like LLM, but they create things of intrinsic value.
This is what removed the shakles.
And this is the biggest difference: LLM don't create anything that will actually free us.
It will never make our lives better and freer.
It will make it easier for _few_ on the cost of _many_.


## 4. Summary {#4-dot-summary}

I believe that LLMs will steal of us younger talent, who has yet to make a dent.
They won't have the chance to learn and earn on the basic stuff, so they will not be in position to create the big thing.
The cost of using LLMs is so low (and will only become cheaper), that there will be no place for them.
And I am afraid that the post-LLM market can find worthy place for them.