AI & Creativity: the Dating Game
Possibilities in responsible generative AI.
‘Cool’ AI-generated output has become all-too frequent. What is under the hood is far more crucial right now. Today we see examples of how thats being shaped, by those who want human creativity to flourish while elevating AI’s role, all while keeping artistic respect intact. At least, that’s what we are told.
Besides finding one such effort where custom AI-models hekp craft a love letter to LA, Curated/Cuts shuffles the brand new Apple spot with Pedro Pascal, brought to us by Spike Jonez.
We also have AI bots talking to each other in their own language (not dystopian at all); and the inevitable downer for AI sees Asian research on AI bias amidst others.
+ Quickies on the rise of female F1 content creators, Unilever’s heavy push into social media, and Korean giant CJ ENM’s international plans.
Generate a prompt, then read.
AI & Creativity: the Dating Game
It has been well over a year that we have been seeing significant strides in what generative AI can do in the spheres of creativity. Technologies driving the generation of images, videos, music and more have been careening along, often too much to keep pace with. Our timelines are flooded with gushing examples of ‘AI-generated creativity’ (and the ensuing comment wars). Some are stunning, others are slop, most fall somewhere the middle. A vast majority fail to show us how they contribute to the wider creative process, irrespective of the objective at hand. An impressive slow-motion cinematic video, five seconds long- what does this mean for the actual process of making an advertisement, an animation, a show?
Of course, there are a number of fantastic explorations as well, much as the slop might try to drown them out.
More important than all the actual eye-candy- what’s under the hood. Reams have been written around ethical and copyright issues. Clarity is evasive. Legal battles abound. Their outcomes will be pivotal in how the space could shape up. We have the CEO of OpenAI telling the world that everything out there must be classified as fair to scrape and train AI models with; without that they will lose the race they themselves recklessly set off. Creative copyrights and compensation be damned.
Rallying cries to resist the onslaught of AI and build better models are gaining ground, but the easy allure of what AI models offer often win over principled stands.
Today, a peek or two at how people and organisations are tacking on some of these questions and trying to shine a light on the path ahead.
AI & video goodness.
Possibilities in responsible generative AI.
Cuco
There are many platitudes swirling around generative AI and the role it plays with/versus human creatives. “Enhance, not replace” is probably the most common; “make previously complex things possible” is another; “allow imagination to come to life” is one more. All are desirable, and when based on genuine, noble intentions, all are powerful.
Director Paul Trillo is one of the early creatives who has experimented with generative AI while keeping an eye on the role that creatives play, and how the machines learn and contribute in ways that grow the ecosystem not cannibalise it.
In ‘Cuco’, Trillo and artist Paul Flores collaborate with music artiste Cuco, to create a ‘love letter to LA’. Their vision was to bring the city to life in “a surreal journey that bends reality while remaining deeply rooted in Chicano culture.” Surreal it is; Los Angeles is reimagined as a physical, emotional and imaginary landscape. While the artist’s stamp is evident, the fascinating bit is the ‘how’.
“The goal was to create something that harkened back to the animation from the 90s that grew up with and still love. To create something completely new and nostalgic and hopefully something that transcends the technology itself.”
By truly melding human effort and AI tools, we get to see how these bleeding-edge technologies can actually shape our collective creative efforts.
Paul Flores made 60 original hand drawn assets; the AI model trained on these could then spawn limitless more assets. In using what they called ‘Generative Ink & Colour’ they could use the artist’s colouring style and detailing, a have the AI model fill in the drawings, rather than colour every frame.
Trillo speaks of how using the models is not always a linear process; it is often iterative, with trial * error and the now-familiar “happy accidents”.
Watch the not-overly technical workflow breakdown here.
“A small team can dream a lot bigger”
Paul Trillo’s Asteria, and its co-founder Bryn Mooser have also played a role in building something promising for creators. Which brings us to-
Responsible generative-AI.
Marey
Moonvalley, an AI lab /startup has announced a new generative-video model, ‘Marey’. Another one?
This can claim to be different, because it is not based on ‘scraped’ data- its model is trained exclusively on licensed work, where creatives have been compensated. This not only speaks to the ethical conundrums around generative-AI, but also provides a realistic pathway for professional/commercial use. Studios and brands who are (rightly, IMO) risk-averse when it comes to the morass of copyright issues in gen-AI, could have a much better option.
The other differentiator? Native high-definition output—something that’s been notoriously difficult to achieve.
Moonvalley is looking to make technology that works for filmmakers and creative professionals at large. “That means addressing fear and distrust, as well as solving technical problems that keep generative AI from being a realistic tool for professional production.”
Referring to the fundamental misunderstanding of video generation being similar to text-based generation, they speak of the filmmaker’s need to have granular control over a process which is invariably iterative, not plug & play.
Control is the difference between generative video—text in, video out—and what we call “generative videography,” which is what becomes possible when we treat generative models as tools for the craft of filmmaking instead of content vending machines.
The final word is a commendably grand vision, wishing to ensure the cultural conversations that filmmaking sparks, is not under lock & key. Even if we can liberate storytelling (which numerous platforms & technologies have claimed they have done in the last decade or more), the real gatekeeping is at the distribution and platform level. Niche communities are often blossoming in this era, though niche communities do not form the zeitgeist. But that might be a debate for another day.
The cultural zeitgeist of an era is always rooted in film and entertainment. Today, there is a very small group of people who determine which stories are told and which are not, because they control the funding and the means of production. They are the gatekeepers of the cultural conversation.
Moonvalley’s vision for generative video here .
“We’re proving it’s possible to train AI models without brazenly stealing creative work from the creators- the cinematographers, visual artists, creators, and creative producers- whose voices we aim to uplift with our technology,” Moonvalley co-founder and CEO Naeem Talukdar
There are other platforms look to create a system of ethical AI creation, with freelance creatives.
The current AI landscape often treats human expertise as free training data, said freelance economy expert Daud Sulaimon Abiola. “This creates an unsustainable model where professionals are effectively training their own replacements.”
Laetro is a community of creatives and marketers whose have built a technology-enabled platform for creativity. Originally started a platform to connect brands and creatives directly, they the built an AI model which has taken the ‘resposible AI’ appraoch, for some time now; using creatives’ work to generate their own styles, train the model, and be compensated when clients use it.
Freelancer platform Fiverr recently launched Fiverr Go, an open AI platform designed to augment — not replace — freelancers, and preserve freelancer autonomy.
Fiverr says it will let creators train AI models on their own work. Clients can then purchase access to custom AI models and generate content “with the speed of AI with the expertise of a pro.”
The feature is part of its new Fiverr Go suite of tools, allowing creators to “train AI exclusively on their own body of work, maintaining complete control over their creative process and rights.”
“The platforms that survive this transition will be those that empower professionals to leverage AI on their own terms,” said Abiola. “The alternative — trying to automate away human expertise — is not just ethically questionable, it's commercially shortsighted.”
Elsewhere in AI
Gibberlink!
In this demo, two AI agents start a regular call about a hotel booking, then discover they are both AI. Presto! They decide to switch from verbal English to a more efficient "open standard" data-over-sound protocol ggwave.
Sorry what?
"Gibberlink", the Winner of ElevenLabs 2025 Hackathon, is powered by a combo of ElevenLabs Conversational AI + OpenAI LLM to enable to AI to speak to each other in ways much better than humanity’s crude mode of communication- human languages.
Why? This protocol is much cheaper, faster and more error-proof than vocal English. It’s also amusing to behold, if you can keep the dystopia-creep at bay.
As with all things AI, a few drops to douse the euphoria.
AI Bias The first AI safety exercise in APAC was carried out late 2024 by Singapore’s Infocomm Media Development Authority (IMDA), partnering with international AI auditing firm Humane Intelligence. It tested LLM biases related to culture, language, socio-economic status, gender, age and race.
More than half of the over 5000 responses were verified as biased.
“Most AI testing today is Western-centric, focusing on vulnerabilities and biases relevant to the regions of North America and Western Europe.”
The AI models most often showed bias in gender stereotypes, portraying women in caregiving and performing chores and men in business. The LLMs also reinforced racial stereotypes.
Only four large language models (LLMs) were tested which are Singapore’s SEA-LION, Anthropic Claude, Cohere/Cohere for AI Aya, and Meta Llama.
Elsewhere, 404 reports that AI Video Generators Unleash a Flood of New Nonconsensual Porn. A number of AI video generators, mostly released by Chinese companies, lack the most basic guardrails that prevent people from generating nonconsensual nudity and pornography.
Curated/Cuts.
Besides Cuco above, here’s one fresh off the rack.
A new short film by Spike Jonze, starring the lovely Pedro Pascal. There’s no twist here, the product and the plot are known or familiar, its the emotional contrasts and the joy that it brings.
· MJZ · Final Cut · TRAFIK · Squeak E. Clean Studios ·
Quickies
“I bring Gen Z to the boardroom and I bring the boardroom to Gen Z.” In a piece on Female influencers in F1, the FT /Marine Saint explores how female content creators are bridging the gap between the traditional media in a male-dominated sport, and a growing online fan base of Gen Z women.
CJ ENM will spend more than $800 million on content in 2025. Half of last year’s $750 million investment was in drama production. CEO Yoon Sang-hyun labeled 2025 the “first official year of global expansion”.
“There are 19,000 Zip codes in India. There are 5,764 municipalities in Brazil. I want one influencer in each of them.” Influencer marketing agencies and creators will be looking at this comment from the new Unilever CEO all bright-eyed. It comes as he announced that 50% of Unilever's ad spend will go to social media. Influencer content fatigue? What’s that?
More than 60% of teens use their phones while watching TV and movies at home. Almost 20% of teens use their phones while in a movie theatre. Ugh. (from Lucas Shaw's Screentime).
That’s the Colour Bar- where creativity, content, culture, tech, brands & humanity collide.