AI with a face
And marketing with an AI mask
Today we’re looking at AI-with-a-personality. The tools, and marketing gimmicks, that aim to bring AI to a larger audience, or market.
You have probably seen Lil Miquela over the past seven years, a poster child for a brave new world that both has and hasn’t happened yet. Launched in April 2016, Miquela Sousa is the creation of the media company Brud, an “AI influencer” with 2.7 million Instagram followers, songs on Spotify, brand partnerships with Samsung, Prada, and Calvin Klein, and fawning headlines about ‘signing’ with a talent agency or bagging $10,000 per post.
And beyond the “Brud universe” of characters there are plenty more successful virtual supermodels and robot rappers on social. The Virtual Humans site does a good job of keeping track of them. Part of their appeal - beyond being completely pliant, unlike pesky emotional humans - is that they suggest the future is here, today! It lends itself well to any brand looking for a future-facing aesthetic in campaigns.
Though mentions of these digital characters go hard on the artificial intelligence angle, the teams behind most of them are generally made of social media experts and visual effects artists. If AI is involved, it’s hard to see where.
In reality, a true AI influencer unleashed on the world, with no strings tightly attached to an agency, would likely be quickly canceled for going down some very dark social rabbit holes. The success of these digital influencers tells us more about the promise and saleability of AI than its capabilities.
As was noted in Forbes, “what Tony the Tiger is to cable, Miquela is to Instagram”. The most recent Instagram post from Brud was captioned, “envisioning communities that share responsibility for the creation, production, hype generation, and sale of their NFT projects”, which gives you some indication of the motivations behind Miquela and others.
Tilly is the brainchild of Amanda Talbot, a Sydney-based designer. Tilly is pitched as “the world’s first ‘heart-led’ AI designer and an exhibition showcasing Tilly in action featured at Milan Design Week.
Talbot, who is not a tech expert and worked with a specialist to design Tilly, said in a recent interview with Stir that Tilly was “an experiment, not a solution by any means”. And that Tilly served as a collaborator, not a replacement. She also emphasized the importance of AI sharing similar values to those who work in her Studio Snoop design studio. It’s an interesting read because the majority of people currently working with AI models tend to speak in more optimistic, and money-focused, terms about its potential.
This fuzzier take - with plenty of reference to anxiety about AI, as well as emotional intelligence and heart - sketches the potential for, say, any small business having an AI assistant that reflects the owner’s worldview and values. This sounds kind of utopian when applied to a Ma and Pa store, and very dystopian when applied to, say, any hedge fund.
During the making of Lord of the Rings: The Two Towers the computer graphics engineer Stephen Regelous came up with something called Massive (Multiple Agent Simulation System in Virtual Environment). It was a game-changer because rather than CGI crowds all acting in an unrealistic uniform way, each little character could act, to an extent, as an individual. In an early test featuring two armies fighting each other several hundred soldiers ran for the hills. It’s a story that stuck with me over the years, these little AI forerunners seeing a battle, saying “screw that” and downing tools to live free in the woods.
Except it’s not quite what happened. Each soldier had a few preprogrammed reactions, and self-preservation was not one of them. They had been programmed to find more space if in a crowd and to run until they encountered an enemy. And they did that, they just never met an enemy. If the program was never stopped they would still be running to this day.
Anyway, we’re 21 years on, and the promise of sentient groups making good, or bad, decisions with occasional little pokes from us is still incredibly appealing. This leads us to South Park (we’re hitting lots of noughties teenage cultural touchstones today). An experiment that allowed AI to create realistic-ish South Park episodes won plenty of headlines last month. Fable Simulation created the AI showrunner tool and used South Park material to build a test case. It was a great way to grab headlines, though their messaging since then has been at pains to point out that A) they have no rights to South Park and B) this tool will allow for the creation of original shows/worlds. It’s another example of AI innovation needing to lean on existing media to break through in the public conscience.
But whatever about the style of content that breaks through, this experiment and others based around “multi-agent simulation” will feature heavily in discussions of AI-with-a-personality innovations over the coming years.