Artificial Intelligence, from opportunity to regulation
27 April 2021
What are the new policies for artificial intelligence systems regulation announced by the European Commission?
Data science
Innovation
On May 11, 2022, at a conference for developers, Google unveiled an image generator named Imagen that will supposedly compete with OpenAI’s DALL·E-2 software. The algorithm is capable of generating images from a simple text prompt. It can even transform strange descriptions into an image, such as “a marble statue of a koala DJ in front of a marble statue of a turntable. The koala has large marble earphones”, “a giant cobra made of corn on a farm” or, “a blue jay standing on a large basket of rainbow macarons”.
“a marble statue of a koala DJ in front of a marble statue of a turntable. The koala has large marble earphones”, “a giant cobra made of corn on a farm” or, “a blue jay standing on a large basket of rainbow macarons”
© Google, Imagen
These technological advances are proof of an inevitable trend: AI is starting to automate production in the creative industries, more specifically, the audiovisual industry.
These algorithms are not only able to initiate script writing from the analysis of a large amount of data, but can also facilitate both pre-production (e.g., calendar optimization, suggestion of filming locations, selection of actors) and post-production tasks (e.g., special effects, addition of posthumous scenes, ageing of actors, editing, creation of trailers, music composition). Finally, prediction and recommendation algorithms can also optimize the launch and marketing of such content (e.g., prediction of box office scores, choice of release date, audience targeting).
Does the use of artificial intelligence in industries whose core business is creativity call into question the very role and intentions of artists?
Box office performance is intimately linked to the characteristics of films. Vault AI, an Israeli start-up has developed a marketing and predictive analysis platform capable of analyzing a film’s box office potential, “bingeability” and potential hype based solely on “the essence of the main story” captured from a random scene or the trailer. This platform uses data based on 30 years of box office revenues, film budgets, audience demographics and information on actors. Around 75% of Vault AI’s predictions are “fairly close” to film opening scores.
With 18.6%1 of films failing to earn revenue above half their budget, production agencies are looking for the “blockbuster algorithm” to optimize their return on investment. Mapping the box office success of big-budget films is now possible by giving them winning ingredients: for example, 30% high-speed car chases, a sex scene, a shoot-out, etc. In fact, much of Hollywood filmmaking consists of writing scenes that adapt to market expectations (Disney, Marvel, Star Wars) and leveraging past successes by producing derivatives, sequels, and prequels that follow the same winning recipe.
Vault AI recently devised a new algorithm called WHAT IF that gives content creators the opportunity to continuously ask the platform what would be the result of specific changes to their content: “What if we made the main character a man? Would this increase the likelihood of our series being continued for an additional season?”
This tends to make the film industry one that’s increasingly driven by box office numbers, and, by extension, by what the public wants, rather than the unique vision and creativity of an artist.
Streaming platforms such as Netflix and Amazon Prime will come out on top, to the detriment of other players in the cultural sector. Thanks to the large amount of data, the technical skills of their teams, and the advanced infrastructure available to them, they have all the cards to make the most of increasingly complex algorithms. As the French music producer Pierre Walfisz summarizes: “This complexity always benefits someone – the powerful. Weaker players have reduced access to certain information, and it’s all very complicated for artists: you can’t build a promotion strategy on a system when you don’t understand its rules.”
While artists using AI insist that the machine remains a tool, the legal status of some content is not without paradox. SACEM (a company of authors, composers, music publishers) only recognizes real people as title holders. Yet, AI is arousing curiosity, and some artists are surfing this wave of paradox by maintaining ambiguity around the authorship of works: on the one hand, they recognize that the human component is decisive, but on the other, they play with the idea that a machine can be the creator. In terms of communication, it is clear that is more striking to say a work was created from the inspiration of a machine rather than the efforts of a human being. On the other hand, in machine learning – and particularly in Deep Learning – machines can sometimes offer results that surprise even algorithm designers. This unpredictability therefore leads artists to feel that they are not entirely responsible for the creative process.
Top institutions that champion the artistic essence of works accentuate this illusion. On May 12, 2017, the Luxembourg Federation of Authors and Composers (FLAC) protested the Minister for Culture’s decision to hire AIVA (a company that creates custom soundtracks using AI) to compose the music for the country’s national holiday: “For the Minister of Culture to commission work composed using AI and impose its use for the ceremony of the national holiday, we see this as an affront to Luxembourg’s composers and a slap in the face of all creators in all artistic fields.”
In 2012, Creativity Research Journal, led by Dr Mark Runco, defined creativity as the combination of two criteria: originality and value. Creativity is characterized by the ability to find hidden patterns, to establish a link between apparently unrelated phenomena, and generate original and value-adding solutions. The link between value and AI is clear: by detecting models and patterns in a large data set, AI generates ideas that can be exploited by the audiovisual industry. Yet, while this creates value, the potential for originality is to be questioned.
Indeed, while pioneering artists demonstrate innovation and create new paradigms by taking audiences out of their comfort zones, algorithms rely on what already exists, on codes and common characteristics (characteristics of the genres of horror, romantic comedy, etc.). This reliance on the existing leads to the standardization and stagnation of content and sets the wheels in motion for “industrialized creativity”, i.e., a form of creativity whose process relies on proven mechanisms to generate value.
Nevertheless, according to Jérôme Neutres, commissioner of the Artists and Robots exhibition, “what these robots do not do is create and invent worlds […]. Some works created by AI derive originally from human intent. Although the algorithm creates the image, it is a human who had the initial intention.”
A new type of art may arise from the collaboration between man and machine. This new paradigm has brought about a new perspective, similar to how photography changed the perception of painting. A new concept has thus appeared to qualify this collaboration: “The augmented artist”
Nevertheless, artists will need to adopt a new vision of this novel creative approach in order to avoid projecting any bias. In 2017, only 12%2 of the scientific research conducted on artificial intelligence was conducted by women. This underrepresentation could increase the current lack of equality in the film industry: in 2020, 16%3 of the top 100 highest-grossing films in America were directed by women.
Share
Sur le même sujet