ChatGPT and the Magnet of Mediocrity
It has only been a little under four months since ChatGPT came out, and yet I already feel super behind in publishing this story. I guess this is a testament to how quickly things are moving in AI today. The space is evolving fast and will only get faster as investment pours in and we use new technologies to build even more powerful tools.
Generative AI feels like one of those moments that is fundamentally transformative for our species, like the printing press, the telegram, or the internet. The main difference is that, while those technologies have pranced into our lives on a horse, or have driven in like an old lady going out for a Sunday drive, Generative AI feels like it has come in on a unicorn, strapped to a jet engine, leaving a trail of glitter in its wake.¹
And it’s so damn useful! Developers are using it to write large chunks of their code, artists are using it to quickly iterate their visual ideas, and businesses are able to automate increasingly complex tasks. All of this, just within a few months of the technology being released.
However, we all know that the faster we travel, the more aware we need to be and the further ahead we need to look in order not to crash. It’s the same with Generative AI. Due to its speed of development and adoption, we need to be more aware and look further ahead at the impact it may have on our society. Here we will explore one of these impacts.
Let’s go.
Imagine we have some activity, let’s say, for instance, writing. This could be any activity that Generative AI is able to do; coding, making art, making movies, etc. but let’s stick with writing for illustrative purposes.
There are amazing writers, writers who aren’t very good, and everything in between.
Anyone wanting to become the next Shakespeare will start at the very bottom and slowly get better and better until, hopefully, they make it to a genius level (very few actually do).
At some point along this journey, the person is able to make some money through their writing (in green). This helps to pay the bills and allows the person to spend more time honing their craft until they are at the top tier.
Enter ChatGPT (and all the other generative AI models). If we look at the content it produces, it is quite mediocre. It can write a decent blog post, a passable article, etc. but it can’t make anything completely original and won’t be inventing a new narrative structure anytime soon.
In fact, it is (in pretentious art critic voice) inherently derivative. And this makes sense since the AI models are designed to provide the most probable (average) output based on the input it has been given.
However, there is something that ChatGPT is very very good at. Speed.
It can churn out the same content that it might take a person hours to do, in a matter of seconds. Even if we get a genius to write a blog post, they still wouldn’t be as fast as ChatGPT. It might be the most beautifully written blog post ever, but it will almost always be slower.
And for all of the jobs where a “mediocre” writing is good enough, it just makes sense to use ChatGPT. Think about all the corporate newsletters, recipe blogs, release notes, etc. written every day. Generative AI can churn out results that do the job in a fraction of the time.
So all of the jobs that were previously available to people that are at, or below ChatGPT’s skill level, will be replaced by ChatGPT (in red).
On top of that, because ChatGPT is so damn fast, in most cases, it is worth it to sacrifice a bit of quality for speed.
Chat GPT acts like a magnet, a magnet of mediocrity, that pulls in all work below its skill level and even some from above.
So we are left with a landscape where only two tiers of skill exist
- ChatGPT “good enough” skill
- Top-tier, genius-level skill
Some of you might be wondering “What’s the issue with this?” After all, we have AI that does all of the menial mediocre tasks, while creating space for the geniuses to produce high-quality work that we can all enjoy.
Yes, there are benefits, but it can also have some negative consequences in the long run.
Let’s go back to our novice who wants to become the next Shakespeare. In order to get there, they still have to put in the same effort as before and climb through all of the various skill levels before making it to the top.
However, there are two key differences from a world without Generative AI.
First, as a novice, you have, at your fingertips, a model that is much faster and better (despite being mediocre) than you are. Why spend two hours writing something when the model can do it in two minutes?
Second, let’s look at the jobs now available to help fund your writing career as you spend time honing your craft. The only jobs available to humans are those at the top level. Everything else has been taken by ChatGPT.
Since most of us need to earn money to pay the bills, you will likely have to take up a job that is not writing, which means less time you are able to spend getting better at writing. This makes it harder for you to put in the hours required to get to the genius level.
While generative AI hasn’t been around for enough time to see its long term effects, we can begin to speculate what might happen.
We can assume that if it is more difficult and there are fewer incentives to put in the hard yards to get really good at something, there will be fewer people who eventually end up at the top skill tier.
Of course, if models continue to get better and better it might not be long before even the geniuses are astonished by what AI is able to produce.
This is nothing new. We have seen how machines have replaced skilled labor and while we may have lost those skills individually, we can build even better things using the technology available.
For example, a 15th-century door carver was probably better than a door carver today, using the same tools. However, nowadays we have CNC machines that can do things that the 15th-century door carver could only dream of in a fraction of the time.
Humans have had to create and adapt to different professions as the level of technology has increased and there is no doubt that we will continue to do so. The challenge now is that there seem to be fewer and fewer things that we are still able to do better than machines²
So. If you are just starting out in a specific domain and want to become really good, what can you do? Are you doomed?
Well, not necessarily. The first thing you can do, if you are looking to master the craft, is to resist the temptation to use Generative AI to do the tasks for you.
Of course, if you aren’t looking to be the next Shakespeare, go ahead and get it to write those blog posts. But if you want to get better, don’t delegate it to AI.
You can even use Generative AI to help you improve. For example, when creating something, give AI the same prompt to see what it would generate.³ Now you have a reference for mediocrity. See if you can do better!
Try to integrate Generative AI into your process. This isn’t the same as simply delegating the entire task to AI, instead, it’s seeing how it can be used to augment your abilities in new and perhaps unexpected ways.⁴
ChatGPT may be a magnet of mediocrity, but as any nine-year-old knows, magnets can be a lot of fun to play with.
So I encourage you to go out and experiment with ChatGPT and all the other Generative AI tools. Figure out exactly how it will take your job and then figure out how you can use it to make sure it doesn’t. Find out the unintended consequences and share them with the rest of us.
It’s the only way we can learn to live with and love the jet-powered unicorn that is Generative AI
Notes and References
- I was struggling to come up with a nice analogy, so I asked ChatGPT for help and this is what it came up with. It doesn’t really work but it’s entertaining so I’m leaving it in.
- Here is is a nice Kurzgesagt video that explains this issue in more detail https://www.youtube.com/watch?v=WSKi8HfcxEk
- https://twitter.com/paulg/status/1618676092163997696
- An example of how it can help teachers https://oneusefulthing.substack.com/p/using-ai-to-make-teaching-easier
