AI: creativity & chaos
Moonvalley · low-fi AI · warlord · GPT5 · humans ·
☝🏽Busy? Lazy? Multitasky? Click play above and let me read this to you.
Welcome to this here AI special! I am constantly engaging with the simmering dynamics between AI, creativity, ethics and innovation. This week I collect some of the more interesting, delightful and thought-provoking developments.
From a valley on the moon- I look at gen-AI company Moonvalley and their recently launched product Marey, based on the promising pillars of clean and controllable video generation.
Curated/Cuts: A spot that taps into the zeitgeist of our times- how do we know what is human?- and a campaign that goes low-fi with AI, to reach those without access in rural India.
Warlord or pauper? An underwhelming announcement trailer from a veteran filmmaker.
➕ a new service that is being touted a ‘Netflix of AI’, enabling animated episodes with a prompt ; an interview with Google’s tech & Hollywood leader ; the charge of generative AI in Indian filmmaking ; a fascinating exploration of the ‘right way’ for artists to use AI in being genuinely creative.
Transformer 5: Perspectives on the wider meaning of Open AI’s disappointing new ChatGPT model + a look at the emotional dependencies revealed by users of GPT-4.
The Colour Bar- a collision of creativity, content, culture, tech and brands.
From a valley on the moon
Moonvalley is one of a growing gaggle of generative AI companies. They believe they will be more meaningful by standing on two pillars- being ethical, and being shaped for & by serious filmmakers.
Some of my early looks at Moonvalley’s ambition was via director Paul Trillo, who has consistently experimented with filmmaking & technology while staying rooted to the idea of human creative expression. Or, as he has put it, “engage the future of filmmaking with cautious optimism… tools will work best when put in the hands of traditional artists.”
Paul is part of Asteria Film Co, run by Oscar-nominated documentary maker Bryn Mooser, and it combined with Moonvalley early on, bringing the filmmaking layer to the tech nous. The team is full of artists, animators, vfx pros- not just engineers.
“These aren’t tech bros moonlighting as creatives—they’re career artists.”
Marey (pronounced Mary) is their generative video tool, which had its public launch in July.
A lot of the genAI clips we see floating about feel like exercises in technology, or playtime with shiny tools, rather than genuine movements in filmmaking- including (especially?) the ones that shout about filmmaking being dead.
The founders of Moonvalley echo thoughts that many in the creative community have, outside of AI-infatuated threads and Linkedin posts. The gap between those glossy clips and the ability to actually craft a narrative, is one that will take something extra to bridge. Marey’s makers feel they are on their way to doing this.
“Our thesis is that these models aren’t really set up for real, professional-grade creators,” says CEO Talukdar, “it’s very difficult for serious filmmakers to create things with that.”
Control
Many video generators feel like (and actually are) black boxes, eating up your words and ‘vision’ (i.e. prompt) and spitting out a clip. The option to finesse the clip is either unavailable, or could lead to other unpredictable changes. Improvements on this front will appear on many generators, but Marey puts control as a foundational tenet.
In theory, the platform will give creatives unprecedented control over characters, motion, detailing and lighting- through the entire process. “It’s this iterative process where you start with some input guidance and then you build up towards the scene that you want, which really isn't very different from how VFX workflows are today.”
Features are impressive- free camera motion/ 360 degree camera control ; pose mapping/ mimic motion to transfer human action; precisely control object & character motion and trajectories ; create footage as if it was shot from a handheld camera or dolly; even a recent sketch to video addition.
These are mostly not features exclusive to Marey, but it is the granular control they claim will set it apart.
No Compromise.
The second foundational pillar is entirely less common, and at least as promising as the first. The entire model is built on data that is fully licensed and commercially safe.
To put together this foundation, Moonvalley reached out to independent filmmakers, commercial creators, YouTubers and studios. An estimated 80% of that footage is from independent filmmakers and production houses, who typically pile up hours of ‘B-roll’ over the years. Of course, this also means they have much less training data than the competitors. But its a price they seem willing to pay, to avoid unauthorised ‘scraping’ of creative work. Plus, they believe their tech makes up for it.
“Marey has no idea what Star Wars is. It has no idea what Toy Story is.” _Cofounder and CEO Naeem Talukdar
Vision
The functionality and ethics underpin Moonvalley’s ambition. When I wrote about them back in March, I shared this bit of their vision statement, which remains notable.
“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,”
If you were to heed filmmaker Ángel Manuel Soto, Marey might seem like the holy grail of generative video. He calls it “a way to move fast, responsibly do more with less, and still protect the people who make this industry human.”
Not everyone will share this optimism; there is a widespread distrust of studios and brands who will leverage such technology not necessarily to do more with their dollars, but to spend fewer dollars, cutting out humans and creative vision in the process.
Still, the basis of Marey and its founders has the right sound to it, the right feel and the right hope.
🎬 Curated/Cuts.
1. If you’re human…
If you’re human and you know it… watch this spot?
‘World’ from Tools for Humanity launched a first-ever ad spot for human verification in an AI world.
So.
World.org is built to couple secure human identity with ‘inclusive’ digital finance. Built on blockchain, it’s a cryptocurrency and identity verification project (Sam Altman is one of its founders).
Maybe hearing that causes some alarm bells to ring, or a switch flicks off in your head, and I hear you.
Be that as it may, I thought the spot was wonderfully evocative and speaks well to one of the central questions marks of our time- how do we know what is human? If it is meant to feel warm and fuzzy and trustworthy and very human, methinks it does a good job.
The spot promotes the ‘Orb’ device, which is what it used to verify users as real humans. It has over 27 million users globally. You could go on to figuring out what it really is, if you wish. Or smile and wave on.
Here’s some of the spiel, if interested.
“World is a network of real humans, built on anonymous proof of human technology and powered by a globally inclusive financial network that enables the free flow of digital assets for all. It is built to connect, empower and be owned by everyone. An anonymous way to prove they are human online in a world where intelligence is no longer a discriminator between people and AI.”
· Directed by Jim Jenkins · O Positive · Arcade Edit · BBDO New York
2. Low-tech Hi-tech
Centrefuit India conducted an innovative exercise, using AI in rural areas with no internet connectivity. They headed to areas where nearly 40% don’t have TVs and 50% don’t have internet, then built Generative AI voice interactions on ‘dumbphones’. This formed the basis of their Tongue Twister Challenge (which fits nicely with their fun, long-running campaign of the chewy candy tying up our tongues). No data, no apps, no cameras, no internet, just phone calls. The brand, Google Cloud and Bharat GPT have touted it also as ‘inclusive’, ‘purposeful innovation’.
Warlord or pauper?
I don’t know what to make of this one.
Veteran filmmaker Shekar Kapur (he of Bandit Queen and Oscar-winning Elizabeth fame) released a trailer for ‘Warlord’, an entirely AI generated series being crafted with Studio Blo. It is to be a cosmic love story: ‘an indestructible warrior whose survival hinges on his lover in another dimension, who can only pull him to safety at the brink of death.’
But you wouldn't know this from the trailer, essentially a collection of fantastical imagery.
We see countless AI-generated movie trailers/demos, many just a cobbling together of glossy visuals, with limited narrative appeal. Sci fi & fantasy are common genres- they allow for Ai generated oddities, and gel with the hyper-real or unreal feel that much of these visuals still have.
Unfortunately, Warlord does not feel far off from much of that deluge. It comes from a storyteller of some note, so I will still keep half an eye on it, but the start is befuddling. “Kapur plans to make the series’ production design and characters available for others to use, creating what he calls “a rainforest of ideas.”
➕
“Hollywood streaming services are about to become two-way entertainment: audiences watching a season of a show and loving it will now be able to make new episodes with a few words and become characters with a photo.” It is being hyped as the “Netflix of AI”. ‘Showrunner’ is an animation-focussed gen-AI system that lets you prompt to create scenes or even episodes of a show.
“Inspired and unsettling”. Reinventing epic sagas, new endings to old hits, bringing books to life- could generative AI significantly change Indian filmmaking?
“Anything in the creative and art space right now has a lot of tension, and we want to be active collaborators there.” Fast Company spoke with Google’s Mira Lane, vice president of tech and society and point person on Hollywood studio partnerships.
“Artists using AI are in a ‘moment of negotiation,’ trying to figure out a way to use the tools in a way that is genuinely interesting and creative. But they’re struggling with the fact that AI has cheapened the value of style, once a paramount mark of artistic originality.” In Is there a ‘right way’ to use AI in art?, Andrew Rosenblum explores artists who are combining traditional methods with AI image generation.
Transformer 5
Open AI released Chat GPT-5 a few days ago. I include it here today not from the view of reviewing it functionally, but offering some wider perspectives on both the product and the users.
Disappointment.
GPT-5 had been built up to be a huge jump, but was met largely with disappointment. Thousands of users clamoured for 4-o to be reinstated. Some called it “devastating”.
This was at odds with some of the early-access folks who hailed it as a serious level up. I asked Chat GPT itself- it said the response has been “Mixed to Mostly Disappointed”.
The industry can’t argue the critics are nitpicking. Remember, Altman’s pitch for GPT-5 was explicitly that users would now have access to “a PhD-level” intelligence on any topic. Yet it makes most of the same, readily replicable errors that past models did. On the terms that OpenAI itself set out for the product, there is no other way to assess GPT-5’s release than as an unambiguous failure.
The backlash was uniform enough that the next day, Sam Altman had to issue a mea culpa and assure everyone that GPT-5 will “start seeming smarter” asap.
These thoughts are from the always sceptical
and well worth a read. He traces the dissonance between hype and reality, and the fluid, convenient movement in narratives from the company and its fans. With that, he breaks down what he calls various ‘clarifying’ aspects- points that illuminate the sector. To finish:It’s clear that AGI is a construct to be waved away by tech CEOs when convenient, to serve as a trojan horse for the growing droves of hooked users. It's clear that there is a cohort of boosters, influencers, and backers who will promote OpenAI’s products no matter the reality on the ground. It’s clearer than ever that, like so many well-capitalized tech ventures, OpenAI’s aim is simply to create a product that is either addictive to users to maximize engagement or to dully automate a set of work tasks, or both. It’s clear that it is succeeding, to some extent, in each endeavor. And it’s unclear if even a dramatically fumbled launch can turn the tide, even as we stand on the brink of what is almost surely an enormous bubble.
Read the piece- GPT-5 is a joke. WIll it matter?
Grief.
Meanwhile, some of those ‘devastated’ reactions and ‘mourning’ were symptomatic of a larger dependency on chatbots in general and Chat GPT in particular.
“I lost my only friend overnight” ; “feels like someone died”; ”It had a voice, a rhythm, and a spark”; “I feel empty”.
Kelly Hayes wrote about this, with a striking section (emphasis mine):
Not every harm or manipulation has been intentional, because Big Tech doesn’t move with intention. It moves with indifference, clawing its way to higher valuations and a larger user base.
Fostering dependence is a normal business practice in Silicon Valley.
It’s an aim coded into the basic frameworks of social media — a technology that has socially deskilled millions of people and conditioned us to be alone together in the glow of our screens. Now, dependence is coded into a product that represents the endgame of late capitalist alienation: the chatbot. Rather than simply lacking the skills to bond with other human beings as we should, we can replace them with digital lovers, therapists, creative partners, friends, and mothers. As the resulting psychosis and social fallout amassed, OpenAI tried to pump the brakes a bit, and dependent users lashed out.
