Copy
Meet Loab, the world’s first digital cryptid. View in browser »
The New Stack Update

ISSUE 329: Can AI Detect Ghosts?

Talk Talk Talk

“Hardware is the ultimate arbiter of our systems’ performance.”

___
Bryan Cantrill, “Bryan Cantrill on Rust and the Future of Low-Latency Systems” contributed by Cynthia Dunlop, ScyllaDB
What's Happening

Aerospike Founder Srini Srinivasan had just finished his Ph.D. at the University of Wisconsin when he joined IBM and worked under Don Haderle, the creator of DB2, the first commercial relational database management system.

Haderle became a major influencer on Srinivasan when he started Aerospike, a real-time data platform. To this day, Haderle is an advisor to Aerospike.

“He was the first one I went back to for advice as to how to succeed,” Srinivasan said in the most recent episode of “The Tech Founder Odyssey.”

From DB2 to Real-Time with Aerospike Founder Srini Srinivasan

Can AI Detect Ghosts?

Meet Loab, the world’s first digital cryptid. Loab, a female, looks like she’s had a rough day. Her face is splotched by the redness of rosacea, her eyes shrunken back into her head. She mostly grimaces, but even when she is smiling, the effect is unnerving.

But no wonder Loab is unhappy: She doesn’t exist. At least not in any corporeal form, in our own dimension. But she shows up quite a bit in machine learning modeling. Loab was given her name by a Swedish artificial intelligence (AI) artist who goes under the nom de plume “Supercomposite.” Loab is purely the product of what the AI industry calls “negative prompt weights,” or instructions for AI modeling software to make an image look as different as possible from some other image.

Using one of the many popular programs today that create images from text prompts, Supercomposite asked for "Brando::-1" (after actor Marlon Brando) which resulted in a nonsensical logo. Then Supercomposite asked for the opposite of the Brando logo, and four strikingly similar photos of a ghostly woman emerged.

And this is not the only place her likeness has shown up. “The AI reproduced her more easily than most celebrities. Her presence is persistent, and she haunts every image she touches,” Supercomposite wrote in a Tweet. “Through some kind of emergent statistical accident, something about this woman is adjacent to extremely gory and macabre imagery in the distribution of the AI's world knowledge.”

In fact, the images are so frightful, the Twitter algorithm itself tagged them as “violence and sensitive content” even though they are entirely fabricated (to be fair, some of the images are genuinely disturbing).

But before we start believing that Loab is a wayward specter soaking through other dimensions, Matthew Skala, founder of North Coast Synthesis Ltd., offered a more rational explanation on Twitter. He wrote that “Yes the pictures are freaky but the phenomenon is not really as surprising as it may first seem.  When you think of it in high-dimensional geometric terms it's understandable why there would be some image that tends to soak up negative-prompt queries.”

It was computer pioneer Alan Turing himself who first devised a mathematical formula for why we keep seeing reoccurring patterns in nature and in our own hallucinations — later called Turing Patterns — such as Lattices, funnels, spirals and cobwebs. So perhaps we are seeing in Loab a reoccurring pattern generated by AI.

This is not the first time people have witnessed ghosts in the machine. We remember from a few decades back, “electronic voice phenomena” in which people claimed to hear voices from the deceased buried in electronic signals, such as radio or television white noise. The idea was that those lost in the underworld could only communicate by encoding their voices on an electronic signal. Examples of EVP were spooky as well, though they can all be easily attributed to signal leakage across the electromagnetic waves.

For now, we’ll conclude Loab is also an unhappy byproduct of overly-bulky statistical algorithms, rather than some demon trying to sneak into our AI models. But we still look forward to all the shlocky movies around computer-conjured ghosts that manifest our collective anxiety around AI run amuck.

In the meantime, please do bookmark our machine learning topic hub for more stories about how AI can be used to generate frightfully useful results. Bwa-hahaha.

Axiom: All the Observability Data without Cost Worries

Axiom is a serverless platform that offers the ability to store and query unlimited amounts of data to improve the observability of system problems. It offers a point-and-click query builder that makes it easy to explore data and build visualizations of it, and a query language to filter, manipulate and summarize the data according to your needs.

A New Tool for Unreal Engine Developers to Export to the Web

A Canadian startup called Wonder Interactive is creating a platform to turn Unreal Engine apps into HTML5 apps, which it claims will have near-native performance and lightning-fast load times. In addition to Unreal Engine, it will also develop the same export functionality for Unity, the open source Godot games engine, and the Amazon-derived Open 3D Engine.

Entrepreneurship for Engineers: Founders Need Community, too

As a founder, putting yourself in environments where you’ll be able to brush shoulders not just with one or two mentors but with a wide variety of entrepreneurs, all at different stages, increases the chances that you’ll find someone who’s encountered a challenge similar to what you’re facing now. In her monthly column, Emily Omier explores the importance of socialization for business founders.

New Website Countdown Initiated!

TNS has been loyal for over eight years to the same website design we began with back in 2014. Over that time, we have worked hard to become the number one DevOps resource in the world. The upcoming launch of our redesigned website is a product of that mission. It is also a reflection of our continued commitment to providing high-quality content while improving the reader experience. With that in mind, we invite you to visit thenewstack.io and take a good look at where we have been. Soon, you will be able to appreciate just how far we have come.

On The Road
SEPT. 13-16 // DUBLIN/IRELAND @ CONVENTION CENTRE DUBLIN Open Source Summit Europe

SEPT. 13-16 // DUBLIN/IRELAND @ CONVENTION CENTRE DUBLIN

Open Source Summit Europe

Join us for the premier event for open source developers, technologists, and community leaders to gather and share information, solve problems and gain knowledge, furthering open source innovation. Register today!

Trust No One & Automate (Almost Everything): Building a Modern Zero Trust Strategy
With cyberattacks on the rise, no organization can afford to be careless about its security. But cloud architecture makes old ways of keeping systems and data safe obsolete.

In a distributed network, there’s no single “castle” to defend. Instead, a new strategy — known as “zero trust” — is the best approach to keeping everything that matters locked down tight — and malicious actors locked out.

A zero trust security strategy calls for not only technology that automates authentication and authorization tasks, but also a shift in thinking and organization-wide cultural changes. In this ebook, we offer a high-level overview and explore what it takes to implement a zero trust approach in an organization.

You’ll learn:
  • What zero trust is.
  • Why the old “castle and moat” approach doesn’t work on the cloud.
  • What authentication and authorization mean in a zero trust environment.
  • How cultural changes support a zero trust strategy.
  • The role automation plays in ensuring security.
Download Ebook
Thanks to our exclusive ebook sponsor, Torq, for making this work possible!

Website
Website
Twitter
Twitter
LinkedIn
LinkedIn
Facebook
Facebook
Instagram
Instagram
YouTube
YouTube
Spotify
Spotify
RSS
RSS
Copyright © 2022 The New Stack, All rights reserved.


Want to change how you receive these emails?
You can update your preferences or unsubscribe from this list