Me and AI Down By The Schoolyard
Is Skynet coming for your post-production job?
by Bobby Hougham and Sevrin Daniels | December 07, 2021
Never say we weren’t warned, Hollywood has had us annihilated by Terminators, used us as batteries in the Matrix, stolen our hearts with Data and Number 5, then turned around and crushed them in Her and Ex Machina. Our ingenuity in creating Artificial Intelligence (AI) is both a testament to our creative genius and a symbol of our hubris. We struggle to understand the impact, place, and nature of this technological boom, but Hal, Replicants, and Skynet are examples of machine sentience that don’t completely align with the AI we are using today.
Today Artificial Intelligence refers to thinking machines used to assist us in our daily tasks. AI is a complex foundation for a broad set of tools that will only get more and more advanced in the coming years. It’s a worldwide, all-industries disruptor the likes of which we haven’t seen since the invention of the desktop computer. That said, in a few years, AI has the potential to make that breakthrough look like the invention of the timer on your toaster.
Currently, AI is being used in nearly every industry in some form or fashion. In Agriculture, it is being used to identify defects in produce and nutrient deficiencies in the soil. Robots equipped with AI are harvesting crops quicker and with greater efficiency than human laborers. We can see AI helping the public as well, in things ranging from GPS navigation apps to writing tools. Right now, I’m writing this in a tool that is constantly checking my grammar and sentence structure. It’s also reviewing my tone and ensuring I’m not writing over or under my intended audience’s education level. (Shout out to Grammarly).
For decades, we have witnessed robots being used to whittle away at manual labor in manufacturing sectors and we are starting to see something similar in the creative industry. Post-production has traditionally relied on a lot of highly specialized skills that may be considered our version of manual labor, however, there is an enormous difference between good and lousy rotoscoping*. Someone straight out of school may know keying principles but not have the real-world experience to know the ins and outs that a veteran compositor will.
Ever since the aforementioned desktop computer revolution, our industry has been nothing but non-stop advances. I remember when I was first starting in my career, having to constantly learn new software platforms. Getris Venice, Henry, Flame/Inferno, and finally After Effects, they all did roughly the same thing but in their own unique ways. What a pain in the ass it was. Then we had a period of relative platform stability. Still, new tools were being added, whether it was 3D space, vector animation, or inverse kinematics, they all affected post production industry trends and post workflows in their own way. It would be easy to fall into the idea that AI will simply be another tool we will need to learn to use. On the other hand, you could also panic and echo the refrain that the robots are coming for your jobs. Somewhere in between is the truth.
The development of artificial intelligence could spell the end of the human race.
So, how is AI different from any previous technological advances? Looking back at earlier disruptors, we can see how they affected our industry and help inform us about the impact of AI.
Case study #1: Back in the early 90’s desktop computers were beginning to be used for graphic design and desktop publishing. I remember hearing people saying they wouldn’t need graphic designers, that I was now obsolete because anyone can set type, make a business card or letterhead or publish a book right from their home office. Short answer: yes, they could. Did they look good? Not really, and did the demand for good design drop? No. We saw people who just needed a business card take their business to Kinkos to print out their own stuff, but the companies who needed to meet a higher standard still took their work to design houses.
Case study #2: Wire and rig removal used to take a fair amount of paint time. Then in the late ’90s, Commotion was developed allowing you to simply draw a line over the top of the wire, and it would magically stitch the image back together saving hours of work. They also figured out how to cache frames in RAM, allowing for real-time playback on desktop machines, this was huge. We used to render high res files out of After Effects, then bring the scene into Commotion just to check for animation issues. And then, in a stroke of brilliance, Adobe put that into After Effects, transforming our workflows forever. For those of you who are actively animating today, imagine setting all your keyframes and kicking off a render without checking to see how it looks. It was akin to animating by braille. This meant that our workstations got less expensive seeing as we didn’t need expensive I/O cards or RAIDs just to check our work.
Now let’s look at some AI-assisted tools that are out today:
Case study #3: Runway is utilizing machine learning on top of its AI architecture to make rotoscoping a thing of the past. We gave this new tech a tryout on a real-world project for Fox Sports this year. Take a look at our case study here. On its face, it automates rotoscoping, which, as anyone who has done it, can attest is one of the most tedious, boring, time-intensive tasks in the VFX world. It’s also the cornerstone for all VFX work. A perfect candidate for AI disruption. Currently, there are off-shore shops that employ thousands of rotoscopers. Most design houses use rotoscoping tasks as a way to help skill up their entry-level positions. Getting people fresh out of school an opportunity to work on projects where they can learn as they go. So if we no longer need to account for teams of rotoscopers on a given project, how will this affect the bottom line? How will this affect our staffing needs?
Case study #4: Not too long ago, there was an interesting article about how Sonantic helped give Val Kilmer his voice back. A victim of esophageal cancer, Val’s voice is no longer anything that it once was. Sonantic employed AI to analyze hours of his recordings to replicate a believable version of his voice. Val was able to write a script, and the AI spoke it with variances and inflections so realistic, that it could fool his mother. Are we going to be able to “hire” AI voices for VO work? Already you can hire “sound-alike” talent for pennies on the dollar; what will these synthetic AI-driven voices cost? Would you pay for a natural person when you can get Val Kilmer with endless revisions or edits for the same price or less?
So is the panic well-founded? It certainly seems like it could be. Stephen Hawking famously said, “The development of full artificial intelligence could spell the end of the human race,” the physicist told BBC News in 2014. In an interview with WIRED magazine, Hawking also said AI will eventually reach a level where it will essentially be a “new form of life that will outperform humans.” But this is based on self-awareness, or is it? Clearly, even if unaware, some applications are outperforming human counterparts today. It doesn’t take an astrophysicist to recognize the pattern or trend.
If Skynet were to actually come online and become self-aware, then I’d personally start investing in property off the grid in the woods. But so long as we aren’t talking about self-awareness, the one thing AI hasn’t been great at is concept or direction. It’s task-based, and while the tasks may get more complex, as far as we can see AI will remain task-based. And while it may take you less and less time and require fewer and fewer people to create works, nature abhors a vacuum. Your team may shrink, but the amount of content will increase exponentially to fill that vacuum largely because of cost savings per project. This means that while you may go from a 10 person team to 3, you will likely have 10 more projects in the pipeline. Each now with less tedium and more of the stuff humans are best at; creativity.
So is AI coming for your job? Absolutely! If you are strictly a keyer, rotoscoper, assistant editor or if you’re building lighting setups. Considering it’s an animation and motion graphics job disruptor, how should one plan their career with this in mind? For starters, we suggest you don’t go to school to learn repetitive tasks. Learn to be a thinker, a creator, a visualizer. Learn how to direct action and animation. Learn how to lead teams. Learn branding, learn art history, learn to discern between the excellent and the mediocre. You may need to rotoscope in the future, but you won’t be the one doing the actual labor. You will, however, need to be able to tell a good separation versus a bad one. You will need to be able to know how to use the tools to achieve the right results. You will also need to be diligent in your research. When we were in art school, we heard instructors complaining about impending doom in careers due to technological advances while simultaneously teaching amber and rubylith stripping techniques. Talk about mixed messages; they recognized the train barrelling towards them but continued with the status quo. Like anyone relying on technology, if you get complacent in your toolset, you can plan on being outmoded.
Shops are going to need to rethink how to work with AI in the mix. It will affect bidding, hardware and software overhead, work and pipeline volumes. It will affect headcount. And here is where it will be a bit tricky. For small shops like ours, one way we envision AI adding value is that we will be able to work on massive projects that typically go to much larger companies with large teams of rotoscope artists. One or two animators will be able to rotoscope a volume of assets in the same amount of time it would take twenty artists utilizing traditional methods. At first, it will be a simple headcount reduction, but if our hunch is correct, shops will be adding headcount of people proficient in the tools rather than the function. In other words, people who are experts in AI-assisted compositing rather than people who know how to rotoscope, key, and layer.
You will also likely need people dedicated to researching and testing tools as they come on the market and regularly redevelop workflows. So while your headcounts in existing roles may go down, we’re guessing that you will find new positions that will spring up to offset some of the losses. Some clients will choose for cost savings, and some will choose to apply that cost reduction toward strengthening and refining the end product. And with increased volume due to lower costs, you will have more teams running than before.
Do you want to set yourself up to succeed in this new era? Start directing your career or your shop or studio to adopt and work with Artificial Intelligence. Explore, invest, and innovate with AI capabilities and establish how they fit into your arsenal. AI is a massive disruptor wave that you can ride if you start paddling now, or you can be swamped by it. It isn’t going away, it isn’t a phase, it isn’t Hal 9000, and it’s not Number 5; it’s something completely different.
Bobby Hougham and Sevrin Daniels are Executive Creative Directors and Founding Partners of The New Blank. They are likely in thrall to an evil, sentient AI; send help immediately.