I’m a Filmmaker. Ben Affleck Is Right About Artificial Intelligence | The Daily Economy

Unless you’ve been living under a rock for the last year, you’ve probably noticed the rapid proliferation of “Artificial Intelligence” tools.

You may have noticed “AI” search results popping up on Google or Brave. Perhaps you’ve even tried software like MidJourney, ChatGPT, Claude, or Sora. These tools are often marketed as being able to do anything from help you write computer code or draw up legal contracts in seconds, to spitting out artistic masterpieces based on a simple text-based prompt.

AI is frequently billed as being incredibly “disruptive”, and unsurprisingly, a lot of people have… thoughts.

As a creative director and film and video producer, I do have some concerns about where this technology is going to take us, but for the most part, I think a lot of the criticism and controversy within the artist community is hysterical.

So, when actor and director, Ben Affleck shared his thoughts about AI in the entertainment business with CNBC a few days ago, I was pleasantly surprised to discover that he had one of the most rational takes on the subject of any creator I’d heard talk about the issue so far.

According to Affleck:

“…the taste to discern and construct that is something that currently entirely eludes AI’s capability and I think will for a meaningful period of time.

What AI is going to do is going to disintermediate the more laborious, less creative, and, you know, more costly aspects of filmmaking that will allow cost to be brought down, that will be lower the barrier to entry, that will allow more voices to be heard, that will make it easier for the people who want to make Good Will Huntings to go out and make it.

…

[W]hat costs a lot of money is now going to cost a lot less. And it’s going to hammer [the visual effects industry], and it already is.

Maybe it shouldn’t take a thousand people to render something. But it’s not going to replace human beings making films. It may make your background more convincing. It can change the color of your shirt. It can fix mistakes that you’ve made. You might be able to get two seasons of House of the Dragon in a year instead of one.”

1I’ve personally spent a lot of time learning about, playing with, and even training custom Large Language Model “AI” to help me and my company produce scripts, illustrations and graphic designs, and videos more efficiently, and I am convinced that Affleck is spot on.

In my layman’s understanding of the technology itself, the way Large Language Models generally function is by aggregating either the information they were initially provided in the development of the software, or new information provided by individual users. 

The software then analyzes that information for patterns and commonalities which will help it pump out prose (or imagery, video, etc.) that has a high probability of meeting the user’s needs, based on an explicit text-based prompt.

There are an immense number of practical uses for this kind of technology. 

My company’s COO routinely uses “Generative Pre-trained Transformers” (Large Language Models which have been pre-built for software such as ChatGPT or Claude) to help draft contracts, summarize meeting notes, codify HR policies, analyze financial information, and to help with other common business tasks.

But in the realm of creative production, I’ve used OpenAI’s ChatGPT software to create a few custom GPTs to function as script and screenwriting assistants for different types of projects.

None of these custom GPTs actually replace my own writing, but I’ve found them to be extremely helpful to quickly get past the tyranny of the blank page. 

For instance, I’ve asked my custom GPTs to help me ideate character and place names (something I absolutely hate doing) or to help me build out a coherent outline based on a stream-of-consciousness brain dump of all the ideas I have for a new story I’m trying to tell. I regularly use them to convert unformatted scripts into proper screenplays that can actually be produced and edited. I’ve used them to create image mock-ups to share with clients so we can confirm that a design brief is on the right track before investing dozens of hours on a complex illustration. 

I’ve even used them to help generate dialogue options for home-brew D&D campaigns, so that the Game Master doesn’t have to come up with it on the spot. 

But even with a copious amount of training, these customized GPTs absolutely cannot write a good script or create a great film or piece of visual art on their own. 

There is simply no replacement for a human artist as the driver of taste and knowing what is (or isn’t) actually good. 

AI mostly just spits out the average of a lot of different inputs which it has identified as being relevant to a given prompt, so what you end up with out of the box is almost always incredibly mediocre and unoriginal.

Now, it’s worth noting that virtually every first draft of anything is extremely mediocre. That’s why it’s a first draft and not a final draft. The problem is, unlike human creators, AI tools don’t know how to get to the final draft. The user can keep asking AI to try again until it produces something decent, but the tool itself has no agency, no actual intelligence, and no discernment whatsoever.

It won’t go back and edit its results after suddenly having a better idea. It doesn’t get a subjective feeling about what’s on trend or what is gonna pick up on some aspect of the cultural zeitgeist. It can’t draw on obscure, seemingly unrelated influences in order to come up with a brilliant mash up of something that hasn’t been done yet — and won’t, because it needs to assign most references and bits of learned information definite categories in order to meet the requirements of most prompts.

For example, if you ask something like Suno to generate a song that sounds like Taylor Swift, it can do that. But what it won’t do is randomly mix subtle hints of The Beach Boys, Esquivel!, The Specials, and jazz organ virtuoso Joey DeFrancesco just because it likes aspects of those artists’ work as well.

But a human songwriter will absolutely do that without even necessarily consciously thinking about it.

What’s more, the human songwriter won’t just spit out the average of all those things in some mathematical way, but instead will include elements as subtle as a single chord from “God Only Knows”; orchestration similar to Esquivel’s version of “Harlem Nocturne”; a bridge inspired by the Specials’ one-drop beats and syncopated guitars; or a transposed reference to some lick DeFrancesco played on a standard like “Autumn Leaves”.

As a human being, you are always going to produce art that is in some ways the amalgamation of all of your favorite influences. But your influences are specific, and while you might choose to pay homage to a single shot from a movie you love or incorporate some themes from one of your favorite books in your own writing, AI puts everything into a blender and pours out a homogenized result.

There are some genuine risks with AI, of course.

While I’m not really sold on the more hyperbolic claims like “AI is going to destroy the world!” it does create some really interesting legal and economic challenges. 

On the legal side, the biggest concern for artists is the implications of generative AI and copyright law. It’s already difficult to sort out copyright violations in the realm of art, and there are already a number of rules — both government- and industry-defined — that determine when it is or is not acceptable to use other people’s copyrighted content or their personal likeness in a new artistic work. 

AI poses new problems in the same realm: Are companies like OpenAI, Anthropic, Kuaishou, or X liable for their users’ creations? To what extent can AI models be trained on pre-existing copyrighted content? Can AI be used to recreate celebrity likenesses? Should copyright holders be paid for any AI creation that recreates their work? What if their work is merely an influence on the ultimate output?

And on the economic front, Ben Affleck was absolutely correct that fields like visual effects, sound design, music, editing, and more are going to be heavily affected by all of this technology. 

But like most previous periods of rapid innovation, the most likely result is going to be an explosion of new creativity and gains in efficiency — old jobs will disappear, yes, but dozens of new jobs will emerge. Meanwhile, millions of people who want to make movies or create music will (eventually) be able to create things that no one today has even dreamed of yet.

These are complex problems, but it’s important to keep in mind that this is still emerging technology, and there’s a lot of work to be done before it can live up to the hype. If the government steps in and begins regulating these industries, it will almost certainly strangle innovation and the value these tools will create for everyone. 

Worse, regulation will empower the government to insert itself into important questions of “trust and safety,” which — for most modern technology companies — is just another word for censorship. If the state gets to decide what art or written prose AI software can create, it captures immense power over our culture and over the very concept of truth.

In my view, those are the real risks of AI.

The way human artists incorporate influences and the way LLMs do it are similar processes in the abstract, but very, very different in practice, which is why they’re decent enough tools to help existing craftsmen speed up their work but a long way off of ever actually replacing human creativity…As long as we don’t do anything to destroy what’s great about these tools before they really get off the ground.