What does the Lensa AI app do with my self portraits and why has it gone viral? | Synthetic intelligence (AI)

Caitlin, I’m seeing a lot of weird portraits of friends popping up on my social media timelines. What is going on?

OK. So there’s this app called Lensa, which was launched as a photo editing app back in 2018, by Prisma Labs.

But it only went viral recently for its new AI-generated “magic avatars” feature. Users, provided they upload 10-20 selfies, can pay a small sum to receive digital portraits of themselves morphed into a range of kooky styles from “anime” to “fairy princess”.

It’s one of those trends that launched to meteoritic popularity really fast – it’s now the most downloaded photo and video app on the iOS store.

Caitlin Cassidy in an AI-rendered illustration. Photograph: Lensa

Sending pictures of yourself off into the ether to get sent back more pictures of yourself seems a bit weird to me, but I am over 30. Why has this taken off?

People love to jump on a trend, and a bunch of celebrities and social media users quickly rushed to share their creations online.

Plus, the various tools employed by the app are designed to “perfect … facial imperfections” and make “selfies look better”, as in, retouching and correcting your face so you look more traditionally attractive.

Then there’s the novelty of computer-generated images that make you look like a “kawaii” star or a David Bowie-esque space figurine.

Lensa’s rendered illustration of Caitlin.

Lensa’s rendered illustration of Caitlin. Photograph: Lensa

Well, this all sounds like harmless fun! Is … is it harmless fun?

Just like all good things in life, maybe not.

Uh oh. Tell me more.

Prisma Labs has already gotten into trouble for accidentally generating nude and cartoonishly sexualised images – including those of children – despite a “no nudes” and “adults only” policy.

Prisma Lab’s CEO and co-founder Andrey Usoltsev told TechCrunch this behaviour only happened if the AI was intentionally provoked to create this type of content – which represents a breach of terms against its use.

“If an individual is determined to engage in harmful behavior, any tool would have the potential to become a weapon,” he said.

Other users of non-anglo descent have also alleged Lensa whitened their skin and anglicised their features, a common complaint of image-editing software on platforms like TikTok.

Tried out the Lensa AI app and fed 20 photos of myself, and I have to say it really struggles with Asian faces. My results were skewed to be more East Asian and I’m absolutely not impressed. pic.twitter.com/WnyLKXQT8K

— Anisa Sanusi (@studioanisa) December 3, 2022

Usoltsev told TechCrunch the technology doesn’t consciously apply “representation biases”.

“The man-made unfiltered data sourced online introduced the model to the existing biases of humankind,” he said. “The creators acknowledge the possibility of societal biases. So do we.”

Then there are the artistic concerns …

Sign up to Guardian Australia’s Morning Mail

Our Australian morning briefing email breaks down the key national and international stories of the day and why they matter

Privacy Notice: Newsletters may contain info about charities, online ads, and content funded by outside parties. For more information see our Privacy Policy. We use Google reCaptcha to protect our website and the Google Privacy Policy and Terms of Service apply.

Right. Artistic concerns sound far more benign though?

Well, maybe. As often comes up in debates about artificial intelligence, artists have expressed concerns the app, and the mainstreaming of AI image generators, are cutting their cheese by turning creativity into a tech-generated process.

Not to mention the Magic Avatar tool costs a minimum of $6 or a yearly subscription of $53.99 to feed it … your own pictures … of yourself.

So what actually happens to the photos you supply to this app?

Yeah people have raised concerns about this, with some suggesting you’re just paying to train facial recognition intelligence and give up your own private data.

The company’s privacy policy says it doesn’t use photos for any reason other than applying “stylised filters or effects”, and face data is automatically deleted within 24 hours after being processed.

AI-generated image of Caitlin Cassidy

Caitlin Cassidy reimagined by Lensa. Photograph: Lensa

In a statement, Prisma Labs said users images were leveraged “solely for the purpose of creating their very own avatars”.

“In very simple terms, there is no[t] a ‘one-size-fits-all collective neural network’ trained to reproduce any face, based on aggregated learnings.”

But any app that collects data can lift other information from your phone, and it’s unclear whether other personal content like location could be shared – even if unintentionally – without a full audit.

OK, so knowing all that, surely people won’t use it?

Well actually, curiosity and my inner Narcissus took hold and I submitted a bunch of my own pictures this morning. It was weird as hell. It looked nothing like me, even as I imagine myself to be if I were dressed up as a fantasy cartoon character. But I had to know.

Please share the weirdest picture offline and I’ll see if we can make it your official byline photo! But really, what a troubling state of affairs! Lucky we’ve got the wholesome goodness of the completely untainted World Cup to take my mind off it all. Oh wait …

Sorry, nothing is sacred.

Leave a Reply

Your email address will not be published. Required fields are marked *