AI may be central to continuing labor battles in Hollywood, but the industry has long adapted to the technology. At SIGGRAPH in Los Angeles, numerous speeches and announcements featured algorithmic and generative tools. Although we may still need to fully understand how tools like GPT-4 and Stable Diffusion fit into the production process, the creative side is prepared to accept them if they can be used to complement rather than replace artists.
Since computer graphics and visual effects have been the focus of SIGGRAPH for the past 50 years rather than film and television production, the two fields have inevitably merged in recent years.
Few presentations or seminars this year addressed the strike because it was the topic of conversation; nonetheless, it was almost always the first topic brought up at networking events and afterparties. However, SIGGRAPH is a gathering of technical and artistic minds, and I thought, “It sucks, but in the meantime, we can continue to improve our craft.”
The concerns about AI in the workplace are, if not unfounded, somewhat misguided. Some problems and advances in generative AI, such as picture and text models, would replace authors and artists. Additionally, studio executives have expressed detrimental and unrealistic expectations that AI tools could partially replace authors and actors. However, AI has long been a part of film and television, carrying out significant and creatively-driven jobs.
This was demonstrated in several panels, technical paper presentations, and interviews that I attended. Naturally, a history of AI in VFX would be fascinating, but for the time being, here are some examples of how AI in its different forms was demonstrated at the forefront of effects and production work.
A couple of Pixar lectures about the animation methods utilized in their most recent movie, Elemental, provided an early illustration. The characters in this film are more ethereal than others, and creating someone who is formed of fire, water, or air is difficult. Imagine combining the fractal intricacy of these components to create a body that can express itself and behave clearly while still appearing “real.”
Simulating and parameterizing the flames, waves, or vapours that made up the many figures was at the heart of procedural generation, animators and effects coordinators explained one after another. It was never an option to hand-sculpt and animate each wisp of flame or cloud emanating from a character. This would require much repetitive, labour-intensive, and technical effort instead of creative work.
However, the aesthetic team and approach were closely tied to the engineering side, as the presentations made apparent, even if they heavily relied on simulators and advanced material shaders to produce the necessary effects. (They further worked on this project with scientists from ETH Zurich.)
Later, additional members of the animation and design teams described how they employed procedural, generative, or style transfer methods to accomplish tasks like changing the hue of a landscape to match a mood board or an artist’s palette or filling in city blocks with distinctive structures that were mutated from “hero” hand-drawn ones. The overarching message was that artificial intelligence (AI) and tools close to AI were there to serve the needs of the artists by accelerating tiresome manual operations and improving the match with the desired appearance.
The use of AI by artists to increase their output “enables dialogue between creators and directors,” he claimed. It’s an alien jungle, but this? or in this way? Like this intriguing cave? Or in this way? Quick input is crucial for a creator-led, graphically challenging story like Nimona. A severe production delay wastes a week, rendering a look that the director rejects a week later.
As CEO of Sokrispy Sam Wickert explained in one of his talks, new levels of interaction and collaboration are being reached in early creative work like pre-visualization. When HBO’s “The Last of Us” premiered, his team created the pre-visualization for the intricate “oner” in a car with numerous extras, camera movements, and special effects.
It’s simple to understand how enhanced voice synthesis, procedural environment creation, and other technologies might and did help this increasingly technological process, even though AI’s use was limited in that more realistic scene.
The threat of automation is also present for artists involved in other stages of the manufacturing process, and it is more of a human issue than a technological one. Many people I met agreed that the problem is terrible decisions made by ignorant leaders.
AI could be transformative if it can improve or streamline the creative process by cutting down on time spent on repetitive work or enabling creators with smaller teams or finances to compete with their better-resourced peers. However, the creative process is delegated to AI, a course that some CEOs appear eager to pursue. In that case, the strikes will only begin even though technology is already pervasive in Hollywood.
[Source of Information: Techcrunch.com]
[Source of Image: Techcrunch.com]
