The first AI conflict, computational creativity, and algorithmic anime.
TNW View in browser →

Hello, carbon-based lifeforms!

The age of AI warfare has begun

Hello, fellow humans,

It's Tom here, deputizing for captain Tristan on today's voyage through cyberspace.

I don't want to sound alarmist, but it seems that we've now entered the era of AI warfare. 

Militaries have, of course, been testing lethal autonomous weapons (LAWs) for years. The landmine is commonly considered to be the first LAW. But some trace their origins all the way back to 162 BC, when the Syriac-Greek army sent 30 drunken elephants rampaging through a battlefield.

Last month, however, the systems reached a terrifying new landmark: a conflict that's been described as the "first AI war."

The title was granted to Israel's devastating attack in Gaza. During the operation, AI was used to intercept rockets fired by Hamas, select locations to attack, and find rocket launchers.

"This is the first time [AI] was used broadly across an operation," a senior Israeli officer told Nikkei

The officer acknowledged that the systems didn't work perfectly. But he added that the conflict had been a useful testing ground.

"This is an opportunity for us to train our algorithms using real-time data."

Israel is far from alone in developing AI for the military. In January, Robert Work, a former deputy secretary of defense who now co-leads the US National Security Commission for AI, said his country has a “moral imperative” to explore LAWs.

The US has been testing military AI systems across land, sea, and skies. A notable recent example involved pitting fighter jets controlled by algorithms against Air Force pilots in simulated dogfights.

China has conducted similar trials. In both countries, the AI systems defeated the human pilots.

Smaller nations are also rushing to bring AI to the battlefield. Britain just used AI in an army operation for the first time, while France recently tested Spot the robot dog in a series of military exercises. 

Military bigwigs say the systems can reduce human casualties as long as they don't malfunction, of course.

I'd like to think that we'll eventually leave the robots to duke it out between themselves, while we hop on a spaceship to Mars. But for now, it seems more likely that superpowers will deploy LAWs to subjugate smaller adversaries as well as their own citizens.

Tales from the arXiv archives

Jamming with robots

We talk about our favorite pre-print research from arXiv's AI section in this weekly column.

Computational creativity fills me with a mixture of dread and excitement. Proponents argue it can augment human artistry, but critics say it reduces creativity to pattern recognition.

A team of researchers at Canada's University of Waterloo have developed a deep learning system that aims to do the former.

Called LyricJam, the tool generates lyrics that reflect the mood and emotions of live music.

In a user study, musicians said the system helped them to compose new lyrics. They also said it encouraged them to improvise and find new musical expressions. 

Per the study paper:

Two novel approaches are proposed to align the learned latent spaces of audio and text representations that allow the system to generate novel lyric lines matching live instrumental music.

One approach is based on adversarial alignment of latent representations of audio and lyrics, while the other approach learns to transfer the topology from the music latent space to the lyric latent space.

LyricJam still needs further refinement before it can generate that match the beat and composition of a song. But for musicians struggling with writer's block or unsympathetic band members, it could provide a non-critical jam partner.

You can check out a sample of the lyrics here, and read the study paper on arXiv.

What we’re writing

👶

US cancels crucial $10B military AI project because Trump is a baby Trump's beef with Bezos has set the Pentagon's AI programs back by years. 

📹

This AI publicly shames politicians, but don’t laugh just yet Automated surveillance could be coming to your workplace next — if it's not already there.

©️

GitHub’s new AI tool doesn’t violate copyrights, says expert Banning it would further undermine copyleft.

👾

Tencent’s creepy new facial recognition system detects teens gaming at night Guess they'll have to sniff glue instead.

🌌

Wild new theory says the Big Bang wasn’t the beginning The universe may have been expanding forever.

🤔

YouTube recommends videos that violate the platform’s own policies, study finds The algorithm is still a dumpster fire.

What we’re reading

🤖

Welcome to dystopia: Getting fired from your job as an Amazon worker by an app (The Guardian)

💰

This TikTok lawsuit is highlighting how AI is screwing over voice actors (Vice)

Euro 2020: Using AI to predict the winner (myKhel)

Something profound from the internet

We're all in this together.

Our favorite AI video of the week

Toei Animation studio used AI to convert photos of Japan’s Sasebo city into a cyberpunk anime. Toei says the technique is helping mitigate labor shortages, but it's also generating some pretty spectacular images.

(click the image below to watch on YouTube):

Well, bye

That's it from me for now, folks. Tristan will be back next week with more tales from the metaverse.

Until next time, stay safe — and watch out for the killer robots.

Peace out,

Thomas

 

From Amsterdam with <3