An image generated by Midjourney.
We’ve reached the next pivotal moment in the history of the tech industry. The newest capabilities of Artificial Intelligence have grown exponentially in value. Depending on who you are, what you do, and what you enjoy, you may believe one of two things: 1) AI is going to eradicate society and take away everyone’s livelihood, or 2) AI is going to be a big help for a lot of people, maybe even myself. If you have been mostly paying attention to social media and the news, you may be in the former category. And, don’t get me wrong, this is certainly a possibility. I’m not going to try to convince you that you shouldn’t be worrying about a potential AI doomsday scenario. However, I would like to share how I believe the latter is the more likely outcome.
The entertainment industry has reached a new pivotal moment as well. The entertainment companies are now reckoning with the compensation practices that they’ve been using for the past few years, and the writers (and at the moment of writing this post, possibly the actors) have decided to strike in order to persuade the companies into fairer compensation and employment standards. To briefly summarize the complaint of the writers: due to the nature of video streaming, and the fact that it is now the primary medium for consuming entertainment, TV shows are not making as much money on the backend from syndication like in previous decades. And even if a show is doing well enough to garner decent compensation on the backend, the streaming companies may make the decision to take it off of their platform specifically because they no longer want to pay the royalties.
On the frontend, writers’ time in the writing room has been significantly reduced. Writers are paid for their labor as they work writing on shows, but what used to be possibly 10+ weeks of income from writing scripts has been reduced to closer to 6. In an industry where the next opportunity is never a guarantee, writers believe that the entertainment companies have forced their craft to convert from career positions to mere gigs. With this reduction of compensation on both the front and backends of the process, writers are not able to afford working in this industry the way that they used to. The development and implementation of their creative ideas is no longer providing them a benefit, and for some, even a livable wage. To make matters worse, now that AI has the ability to write and develop ideas for shows, companies have already began exploring the use of AI for its future shows. With all of this going on, there’s plenty of justification in why writer’s feel they need to strike to secure their future. If you’d like to see the full demands of the WGA, which authorized the strike, you can view them here.
Picketers protesting on behalf of the writer's strike.
The New Machines
The irony, and the cornerstone of this conversation, is that the newest advancements in AI have made ideas and creativity are more valuable than ever. I know that it may not seem that way, and I understand why. The newest AI achievements have displayed potential of becoming amazing tools for different careers. The most important part about that previous sentence, however, is the word “tools”.
Surely, at this point, you have heard of OpenAI’s newest product, ChatGPT. This AI has the ability to summarize and explain almost any topic that you ask it. It doesn’t matter the field or discussion, it can usually provide some insight. Then, there’s the release of Stable Diffusion and its derivative, Midjourney. These AI tools have the ability to produce whatever you can possibly imagine, all from inputting a prompt. To be fair, they may not be able to implement every single last detail, but they can definitely establish a base image or concept that you can further develop into your final artwork. All of these AI can produce these results in a matter of seconds.
A vector illustration of a city skyline, generated by Midjourney.
These AI capabilities have made their ways into already existing companies and products as well. Microsoft has already released a beta version of their Bing browser that will include AI assistance when surfing the web. Google is working on integrating AI into it’s entire Google Workspace product line, giving AI the ability to assist you with emails, spreadsheets and documents. In the creative space, Adobe has already released Adobe Firefly, a Generative AI that can help you alter elements of your images in Photoshop, or even produce whole pieces of artwork directly from your prompts. It’s no surprise that companies jumped at the chance to incorporate AI into their tooling ASAP. And this is not an exhaustive list, Microsoft want to integrate into their Office Suite, Notion already has released Notion.ai, and Canva is adding tooling as well. This is not going to slow down anytime soon.
The Reality
To clarify, it is the integration of these AI features into new and existing tools that won’t slow down. There’s reason to believe that, while the adoption of AI will continue to expand, the heights of AI capabilities may already be plateauing. The current process of creating the AI through the process of developing massive mathematical models trained on trillions of data points could already be reaching its ceiling. It’s the same way that cellphones haven’t had revolutionary improvements in the last several iterations. We should be able to improve the AI we currently have, making them faster or more efficient. But to give them new capabilities, we may already need to find a different approach. This is something admitted by Sam Altman himself, CEO of OpenAI, which created the AI for ChatGPT.
This is somewhat unfortunate news to receive from this newly emerging field, as we can truly tell that this technology is truly approaching maturity. In case you are unfamiliar, AI is not a brand new concept for technology. AI has been studied and used for decades now. Even the concept of training a neural network is not entirely new, as Geoffery Hinton, dubbed the Godfather of Artificial Intelligence, claimed that he had tried the same methods many years prior, but was not able to pursue it further due to the limitations of technology at the time. Despite this, AI was still used in areas such as entertainment and biology. And, these fields will likely be able to develop even further with the new strides in AI.
But, we all know that AI is far from perfect. A few minutes and some poorly formatted prompts in Midjourney can prove that, not to mention the crazy, hysterical things that Bing’s AI was saying early in its rollout.. And while these AI will likely get better, become more accurate, and their interfaces more intuitive, if they won’t be able to do anything new, then where does that actually leave the rest of us?
I believe that AI will be an amazing tool to assist us in reaching new heights in human evolution. But, I also believe it will be just that: a tool. I’ve made sure that I’ve always referred to it as throughout this post, because, ultimately, I don’t believe it is, or should be, anything more than that. The words that we use to describe this phenomenon will be very important moving forward. For example, while we say that AI can “create”, I challenge this notion. The definition of “create” is “to bring (something) into existence.” AI cannot bring anything into existence on its own. The point of origin of existence begins in the thoughts and mind of a person, and then it can be prompted or requested from the AI, which will then generate something. The definition of “generate” is “to cause something to arise or come about”. There is another definition for generate that I believe is even more apt for this discussion: “to produce (a set or sequence of items) by performing specified mathematical or logical operations on an initial set.” This phrase “on an initial set” is what continues to separate man from machine.
When AI receives a prompt, it uses its mathematical model, which was trained numerous times on a massive, existing set of information, to then predict some sort of value (whether it be a bunch of words and phrases that we interpret as an answer to a question, or a bunch of pixels that we interpret as an image) to return to the user. Without the prompt and the testing data, the AI cannot do anything. We humans can create with just one or the other (perhaps, at times, without either). I believe you could make the argument that we do have that predefined set, which is whatever environment in which we exist, but in the early days of mankind, while we had the world, we lacked the relationships between the elements within the world. These relationships are the barriers between man and our ability to create, and we’ve had to break down these barriers in order to get where we are today. AI is equally bound, if not more so, as it’s only able to be trained on data we give it, and we cannot teach it anything we haven’t already learned ourselves.
The Path Forward
This brings us back to the problem at hand in the entertainment industry. While ChatGPT can technically produce the outline of a story or script, it will still need to be the human writer to come up with an idea worth prompting in the first place. Not only that, but once the outline has been generated, a writer will still need to create the details of the story that will make it original. The same goes with visual art. While Midjourney can quickly produce stunning visuals, many artists have realized that they need to make more adjustments to bring a piece to it’s final form. While using Adobe Firefly, the artist will know what elements generated by the AI are aesthetically appealing and appropriate for the task at hand.
I’ve noticed this when it comes to programming as well. I’ve spent time using both ChatGPT and GitHub Copilot to assist me in writing code. What I noticed is that ChatGPT can get inaccurate, and even hallucinatory, once you start asking it for less common code, such as Terraform modules. I’ve also noticed that I either rewrote, modified, or flat out deleted about 80% of the code that Copilot recommends. I’d assume only about 10-12% of code from both AI actually makes it into production. While this can be of good use at times, it’s hardly sufficient for replacing real engineers. I recently encountered a thread on Twitter discussing how an engineer that utilized ChatGPT, Copilot, no-code solutions and Cloudflare Workers for an MVP “clearly outperformed” (my words, not theirs) a veteran engineer that just did everything from scratch, and how the veteran will be out of a job soon. I personally feel sorry for the future engineer who will inherent rewriting/migrating that project. I’m currently dealing with a haphazardly programmed project and it is, to put politely, infuriating.
At this point in time, it’s overzealous to claim that AI can replace everyone’s job. There are surely some jobs that it can replace, but I doubt that there will actually be 100% coverage for every job. There may be 95%, and a person would be required to make up the other 5%. You could project this to mean the value of the person has decreased since it’s doesn’t fully cover all of the previous requirements of the job. However, this could also mean that the capacity of the person could increase, making their output increase, and possibly making their company more proficient or profitable.
This would eventually make a human’s intervention and activity more valuable at work, not less. If AI can not completely take over the task, the next best thing would be having a person that is skilled and proficient in using AI wielding it to be more effective at their job. This in turn makes people in their position more valuable. But where does that value come from? In an AI takeover, this value would be derived implicitly, as the overhead cost of the job getting done is reduced when a machine can take over. In the current scenario, the value is derived explicitly, through a person that is about to increase volume/quality of their output, potentially increasing profits.
This doesn’t mean that everyone’s job is absolutely safe! If a person that uses AI is just as or more proficient than 5 people that don’t, then 4 jobs could possibly go away. This is because, on paper, a company would look at this as an opportunity to cut costs, and even devaule positions, as we witnessed Max's relaunch accidently achieve by lumping all positions for a film or show under the title "Creator". In actuality, this would be counter-productive. Rather than cutting workers who don't reach extra proficiency, there should be new initiatives focusing on helping people incorporate AI tools into their jobs. Then, if all 5 workers become mroe proficient, then their output or quality could increase exponentially, and so could profitability.
To become that person that is more proficient, the skills of the worker will have to shift from the ability to produce to ability to create valuable ideas, having a rich imagination, and knowing how to execute those ideas. The power of human thought will become the driving force behind future success, and the writers that are currently striking in Hollywood are fighting to reinforce the notion that our ideas and our thoughts are of value, and should be compensated accordingly.