The Humane AI Pin gets hacked

📰 News Roundup

• Butterfly, what butterfly? Researchers tried to catch AI out this week by showing it a Rorschach test and asking what it saw. Instead of giving a straightforward response the system gave both possible options “A butterfly or a moth” as well as a description of a Rorschach test. An example AI is badly trained? Or an example that AI knows when you’re trying to trap it? You decide.

• The ghost in the machine of Humane AI lives on. A hacker group managed to bring their Humane AI Pin back online this week after Humane sold to HP and bricked all the devices. Starting a Discord named ‘ReHumane’ the group got hold of encrypted keys to unlock the devices for good. The first thing they did? Use the Humane AI Pin’s laser projector to play the music video for Nomico’s “Bad Apple,” which has become a meme being the first video hackers put on a jailbroken device. The AI Pin belongs to the people now.

• Take It Down. Melania Trump sat with two victims of deepfake AI imagery this week at the congressional address as she pushes through the Take It Down Act to regulate deepfake AI imagery which the President has backed. The bipartisan bill would also require social media sites to have procedures to promptly remove nonconsensual intimate images once notified.

• aiRobot Boston Dynamics the robotics company responsible for nightmares like this and this and THIS have decided to integrate its robots with AI, specifically - reinforcement learning, this has already tripled the running speed of one of its robots. Sounds like a great idea that has no risk whatsoever… if only a science fiction writer had written some kind of warning about this.

• AI can now write its own Genomes. In a scientific break through The Arc Institute has created Evo 2, an AI that is trained on the DNA of over 100,000 species. This AI can identify disease-causing mutations in human genes, and even write brand new genomes, up to 1 million base pairs long.

🛠 Tool of the Week: Doji

Digital fashion is struggling, the industry had a tough run last year. But out of the darkness comes one of the greatest apps we’ve seen in a hot minute. Doji has been around for a few months, but has just released a bunch of new invitations.

The Doji app scans your likeness through your smart phone, creating an AI version of you, and then can style you through its AI algorithm, or take pieces from the web that you link it to. Plus it makes you look like an SSENSE model, so there’s that too.

I tried it out this week, styling myself in Balenciaga, ALYX 9SM, Rick Owens and Acne Studios to name a few.

🤌 Check it out: [HERE]

🎟️ New Speakers announced for the Artist and the Machine AI & Creativity Summit

🔦 Spotlight

Get an AI Ideas report, in minutes! AE Studio’s AI-Ideas feature, has a free instant AI powered form, that will look at your business needs and help identify where AI can help or grow your business. With their AI powering Berkshire Hathaway, Samsung and Scotch and Soda, the possibilities are endless.

👾 Interview: Josephine Miller

Josephine Miller is an AI partner for Meta, OpenAI and works with Nvidia among others. Whilst most readers have probably used her Threads filter for Meta, Josephine is one of the best creative AR and AI technology experts of the moment. We caught up with her for this issue’s weekly:

Perhaps your most known work is working with AI and fashion. Creating in partnership with Meta to create different looks without changing your physical clothing. What is your vision for the future of AI in fashion?

I believe AI in fashion will help create solutions for virtual try-ons, deliver more effective personalised advertising, and spark greater innovation in materials, technology, and conceptualisation. It allows designers to rapidly iterate in a digital environment without having to prototype every design with physical materials. I also think it will really help up-and-coming artists showcase their work at a higher production level, even without large budgets.

You’ve been an alpha artist test with OpenAI for 3 years now, what’s been the biggest shift with taking Sora to public release? 

I started off when DALL-E was first released for image generation. The biggest shift I’ve noticed since beginning to use Sora is how high-quality and hyper-realistic the outputs can be. There’s a strong push for realism now, which helps creatives like me produce projects at a higher production level for personal work and concept designs. Since Sora’s release, there’s also been a noticeable surge of AI artists on social media. Opinions are definitely split, but we’re also seeing new art styles and groundbreaking concepts taking off.

You just released Biolume, an AI generated fashion film, that really explores materials and texture. What was your inspiration behind this film, and what procedures did you work on the most to get the quality of materials the level you wanted? 

Biolume was inspired by my fascination with impossible fashion designs that surpass the limits of physical materials. I wanted to explore bioluminescent, fluid-like fabrics that can evolve and react like living entities. Since I relied entirely on text prompts in Sora to achieve realistic textures, I focused on the movement of the materials, their translucency, and how they interact with light, making sure the fabrics were both hyper-real and dreamlike. Ultimately, Biolume reimagines fashion outside the confines of reality, showcasing how AI empowers designers to conceptualise and visualise avant-garde ideas without needing traditional production.

You’ve been a significant part of SAM (segmentation anything model) in Meta’s AI kit. How does SAM work and how easy is it for a non AI expert to play around with this ground-breaking technology? 

I’ve been experimenting with SAM in ComfyUI, an open-source workflow that uses nodes. Right now, it’s pretty heavy on my computer’s processing power, but Meta has shown how they plan to integrate SAM with their upcoming AI video app model called ‘Edits’ that will, let users change backgrounds or outfits in different ways. I’m really excited to see how content creation evolves with these tools. Even though it’s still quite advanced for the average content creator and requires some technical know-how, I think we’ll see it become more accessible as integration improves.

Augmented Reality has gone through a tough 2024, a medium you’re very familiar with. How is Artificial Intelligence bridging the gap for creators and artists that were traditionally AR native? 

Myself and other augmented reality creators are now exploring different avenues using AI, such as commercial advertising or integrating it with VFX. AR is in a sort of “in-between” stage while AR glasses are still being developed. But I do believe AI will help fill the gap by enabling more complex and personalised interactions in augmented reality, like using APIs to tailor experiences for each user. The main limitation is that current AI tools aren’t real-time in the same way AR is, but I’m sure that will evolve soon.

As a creative technologist at the forefront of the industry, what are you most excited about with AI in 2025? 

I see 2025 as a year of experimentation and pushing commercial use cases for these new incredible tools. I’m excited to see more fashion and commercial art campaigns integrating AI in a professional way, using daily advancements to streamline workflows for all kinds of creators. I’m also intrigued by the competition between the East and West to build more advanced AI systems that help with everyday tasks. The pace of innovation is already so rapid, and I can’t wait to see where it leads.

AI & Creativity Summit April 24, New York City

Til next time,

Dani Van de Sande (Founder), James Joseph (The Weekly’s Editor) & the Artist and the Machine team.

ARTIST AND THE MACHINE is a free newsletter featuring the latest in AI x culture. Subscribe so you never miss an issue, and if you liked it, share with 1-2 friends .

xx