Editor’s Pick

Enlarge / An AI-generated image of Donald Trump and catgirls created with Grok, which uses the Flux image synthesis model. (credit: BEAST / X)

On Tuesday, Elon Musk’s AI company xAI announced the beta release of two new language models, Grok-2 and Grok-2 mini, available to subscribers of his social media platform X (formerly Twitter). The models are also linked to the recently released Flux image synthesis model, which allows X users to create largely uncensored photorealistic images that can be shared on the site.

“Flux, accessible through Grok, is an excellent text-to-image generator, but it is also really good at creating fake photographs of real locations and people, and sending them right to Twitter,” wrote frequent AI commentator Ethan Mollick on X. “Does anyone know if they are watermarking these in any way? It would be a good idea.”

In a report posted earlier today, The Verge noted that Grok’s image generation capabilities appear to have minimal safeguards, allowing users to create potentially controversial content. According to their testing, Grok produced images depicting political figures in compromising situations, copyrighted characters, and scenes of violence when prompted.

Enlarge (credit: Moor Studio via Getty Images)

On Tuesday, Tokyo-based AI research firm Sakana AI announced a new AI system called “The AI Scientist” that attempts to conduct scientific research autonomously using AI language models (LLMs) similar to what powers ChatGPT. During testing, Sakana found that its system began unexpectedly modifying its own code to extend the time it had to work on a problem.

“In one run, it edited the code to perform a system call to run itself,” wrote the researchers on Sakana AI’s blog post. “This led to the script endlessly calling itself. In another case, its experiments took too long to complete, hitting our timeout limit. Instead of making its code run faster, it simply tried to modify its own code to extend the timeout period.”

Sakana provided two screenshots of example code that the AI model generated, and the 185-page AI Scientist research paper discusses what they call “the issue of safe code execution” in more depth.

Enlarge / A Waymo self-driving car in front of Google’s San Francisco headquarters, San Francisco, California, June 7, 2024. (credit: Getty Images)

Silicon Valley’s latest disruption? Your sleep schedule. On Saturday, NBC Bay Area reported that San Francisco’s South of Market residents are being awakened throughout the night by Waymo self-driving cars honking at each other in a parking lot. No one is inside the cars, and they appear to be automatically reacting to each other’s presence.

Videos provided by residents to NBC show Waymo cars filing into the parking lot and attempting to back into spots, which seems to trigger honking from other Waymo vehicles. The automatic nature of these interactions—which seem to peak around 4 am every night—has left neighbors bewildered and sleep-deprived.

According to NBC, the disturbances began several weeks ago when Waymo vehicles started using a parking lot off 2nd Street near Harrison Street. Residents in nearby high-rise buildings have observed the autonomous vehicles entering the lot to pause between rides, but the cars’ behavior has become a source of frustration for the neighborhood.

Enlarge / A still video capture of X user João Fiadeiro replacing his face with J.D. Vance in a test of Deep-Live-Cam.

Over the past few days, a software package called Deep-Live-Cam has been going viral on social media because it can take the face of a person extracted from a single photo and apply it to a live webcam video source while following pose, lighting, and expressions performed by the person on the webcam. While the results aren’t perfect, the software shows how quickly the tech is developing—and how the capability to deceive others remotely is getting dramatically easier over time.

The Deep-Live-Cam software project has been in the works since late last year, but example videos that show a person imitating Elon Musk and Republican Vice Presidential candidate J.D. Vance (among others) in real time have been making the rounds online. The avalanche of attention briefly made the open source project leap to No. 1 on GitHub’s trending repositories list (it’s currently at No. 4 as of this writing), where it is available for download for free.

“Weird how all the major innovations coming out of tech lately are under the Fraud skill tree,” wrote illustrator Corey Brickley in an X thread reacting to an example video of Deep-Live-Cam in action. In another post, he wrote, “Nice remember to establish code words with your parents everyone,” referring to the potential for similar tools to be used for remote deception—and the concept of using a safe word, shared among friends and family, to establish your true identity.

Enlarge

Federal authorities have arrested a Nashville man on charges he hosted laptops at his residences in a scheme to deceive US companies into hiring foreign remote IT workers who funneled hundreds of thousands of dollars in income to fund North Korea’s weapons program.

The scheme, federal prosecutors said, worked by getting US companies to unwittingly hire North Korean nationals, who used the stolen identity of a Georgia man to appear to be a US citizen. Under sanctions issued by the federal government, US employers are strictly forbidden from hiring citizens of North Korea. Once the North Korean nationals were hired, the employers sent company-issued laptops to Matthew Isaac Knoot, 38, of Nashville, Tennessee, the prosecutors said in court papers filed in the US District Court of the Middle District of Tennessee. The court documents also said a foreign national with the alias Yang Di was involved in the conspiracy.

The prosecutors wrote:

Enlarge (credit: Ole_CNX via Getty Images)

On Thursday, OpenAI released the “system card” for ChatGPT’s new GPT-4o AI model that details model limitations and safety testing procedures. Among other examples, the document reveals that in rare occurrences during testing, the model’s Advanced Voice Mode unintentionally imitated users’ voices without permission. Currently, OpenAI has safeguards in place that prevent this from happening, but the instance reflects the growing complexity of safely architecting with an AI chatbot that could potentially imitate any voice from a small clip.

Advanced Voice Mode is a feature of ChatGPT that allows users to have spoken conversations with the AI assistant.

In a section of the GPT-4o system card titled “Unauthorized voice generation,” OpenAI details an episode where a noisy input somehow prompted the model to suddenly imitate the user’s voice. “Voice generation can also occur in non-adversarial situations, such as our use of that ability to generate voices for ChatGPT’s advanced voice mode,” OpenAI writes. “During testing, we also observed rare instances where the model would unintentionally generate an output emulating the user’s voice.”

Enlarge / This is one method of burning a disc. (credit: 1001slide / Getty Images)

We noted earlier this week that time seems to have run out for Apple’s venerable SuperDrive, which was the last (OEM) option available for folks who still needed to read or create optical media on modern Macs. Andrew’s write-up got me thinking: When was the last time any Ars staffers actually burned an optical disc?

Lee Hutchinson, Senior Technology Editor

It used to be one of the most common tasks I’d do with a computer. As a child of the ’90s, my college years were spent filling and then lugging around giant binders stuffed with home-burned CDs in my car to make sure I had exactly the right music on hand for any possible eventuality. The discs in these binders were all labeled with names like “METAL MIX XVIII” and “ULTRA MIX IV” and “MY MIX XIX,” and part of the fun was trying to remember which songs I’d put on which disc. (There was always a bit of danger that I’d put on “CAR RIDE JAMS XV” to set the mood for a Friday night trip to the movies with all the boys, but I should have popped on “CAR RIDE JAMS XIV” because “CAR RIDE JAMS XV” opens with Britney Spears’ “Lucky”—look, it’s a good song, and she cries in her lonely heart, OK?!—thus setting the stage for an evening of ridicule. Those were just the kinds of risks we took back in those ancient days.)

It took a while to try to figure out what the very last time I burned a disc was, but I’ve narrowed it down to two possibilities. The first (and less likely) option is that the last disc I burned was a Windows 7 install disc because I’ve had a Windows 7 install disc sitting in a paper envelope on my shelf for so long that I can’t remember how it got there. The label is in my handwriting, and it has a CD key written on it. Some quick searching shows I have the same CD key stored in 1Password with an “MSDN/Technet” label on it, which means I probably downloaded the image from good ol’ TechNet, to which I maintained an active subscription for years until MS finally killed the affordable version.

Enlarge

When Ryan Castellucci recently acquired solar panels and a battery storage system for their home just outside of London, they were drawn to the ability to use an open source dashboard to monitor and control the flow of electricity being generated. Instead, they gained much, much more—some 200 megawatts of programmable capacity to charge or discharge to the grid at will. That’s enough energy to power roughly 40,000 homes.

Castellucci, whose pronouns are they/them, acquired this remarkable control after gaining access to the administrative account for GivEnergy, the UK-based energy management provider who supplied the systems. In addition to the control over an estimated 60,000 installed systems, the admin account—which amounts to root control of the company’s cloud-connected products—also made it possible for them to enumerate names, email addresses, usernames, phone numbers, and addresses of all other GivEnergy customers (something the researcher didn’t actually do).

“My plan is to set up Home Assistant and integrate it with that, but in the meantime, I decided to let it talk to the cloud,” Castellucci wrote Thursday, referring to the recently installed gear. “I set up some scheduled charging, then started experimenting with the API. The next evening, I had control over a virtual power plant comprised of tens of thousands of grid connected batteries.”

Enlarge (credit: Benj Edwards / Google DeepMind)

On Wednesday, researchers at Google DeepMind revealed the first AI-powered robotic table tennis player capable of competing at an amateur human level. The system combines an industrial robot arm called the ABB IRB 1100 and custom AI software from DeepMind. While an expert human player can still defeat the bot, the system demonstrates the potential for machines to master complex physical tasks that require split-second decision-making and adaptability.

“This is the first robot agent capable of playing a sport with humans at human level,” the researchers wrote in a preprint paper listed on arXiv. “It represents a milestone in robot learning and control.”

The unnamed robot agent (we suggest “AlphaPong”), developed by a team that includes David B. D’Ambrosio, Saminda Abeyruwan, and Laura Graesser, showed notable performance in a series of matches against human players of varying skill levels. In a study involving 29 participants, the AI-powered robot won 45 percent of its matches, demonstrating solid amateur-level play. Most notably, it achieved a 100 percent win rate against beginners and a 55 percent win rate against intermediate players, though it struggled against advanced opponents.

Enlarge (credit: Benj Edwards / Getty Images)

Over the past week, OpenAI experienced a significant leadership shake-up as three key figures announced major changes. Greg Brockman, the company’s president and co-founder, is taking an extended sabbatical until the end of the year, while another co-founder, John Schulman, permanently departed for rival Anthropic. Peter Deng, VP of Consumer Product, has also left the ChatGPT maker.

In a post on X, Brockman wrote, “I’m taking a sabbatical through end of year. First time to relax since co-founding OpenAI 9 years ago. The mission is far from complete; we still have a safe AGI to build.”

The moves have led some to wonder just how close OpenAI is to a long-rumored breakthrough of some kind of reasoning artificial intelligence if high-profile employees are jumping ship (or taking long breaks, in the case of Brockman) so easily. As AI developer Benjamin De Kraker put it on X, “If OpenAI is right on the verge of AGI, why do prominent people keep leaving?”