ars-rss
SpaceX claims Starlink can offer gigabit speeds if FCC approves new plan
SpaceX: “Small-but-meaningful updates” can boost speed from about 100Mbps to 1Gbps.
SpaceX is seeking approval for changes to Starlink that the company says will enable gigabit-per-second broadband service. In an application submitted to the Federal Communications Commission on October 11, SpaceX claims the requested “modification and its companion amendment will enable the Gen2 system to deliver gigabit-speed, truly low-latency broadband and ubiquitous mobile connectivity to all Americans and the billions of people globally who still lack access to adequate broadband.”
SpaceX said it is seeking “several small-but-meaningful updates to the orbital configuration and operational parameters for its Gen2 space station authorization to improve space sustainability, better respond to evolving demand, and more efficiently share spectrum with other spectrum users.”
SpaceX wants to lower the altitudes of satellites “at 525 km, 530 km, and 535 km to 480 km, 485 km, and 475 km altitude, respectively.” The reconfiguration will increase the “potential maximum number of orbital planes and satellites per plane” while keeping the planned total number of second-generation satellites at 29,988 or less. The FCC has so far approved 7,500 Gen2 satellites.
Reports: Tesla’s prototype Optimus robots were controlled by humans
But the prototypes used “artificial intelligence” to control their walking.
After Elon Musk provided his “long-term” vision for autonomous, humanoid robots at last week’s “We, Robot” event, we expressed some skepticism about the autonomy of the Optimus prototypes sent out for a post-event mingle with the assembled, partying humans. Now, there’s been a raft of confirmation that human teleoperators were indeed puppeting the robot prototypes for much of the night.
Bloomberg cites unnamed “people familiar with the matter” in reporting that Tesla “used humans to remotely control some capabilities” of the prototype robots at the event. The report doesn’t specify which demonstrated capabilities needed that human assistance, but it points out that the robots “were able to walk without external control using artificial intelligence” (the lack of a similar AI call-out for any other robot actions that night seems telling).
That lines up with reporting from tech blogger Robert Scoble, who posted on social media that he had “talked with an engineer” who confirmed that “when it walked, that is AI running Optimus.” For other tasks—like pouring drinks from a tap, playing Rock Paper Scissors, or chatting with nearby attendees—Scoble noted that “a human is remote assisting.”
Spotify criticized for letting fake albums appear on real artist pages
Real bands struggle to remove fake albums from their Spotify pages.
This fall, thousands of fake albums were added to Spotify, with some appearing on real artist pages, where they’re positioned to lure unsuspecting listeners into streaming by posing as new releases from favorite bands.
An Ars reader flagged the issue after finding a fake album on the Spotify page of an Australian psych rock band called Gong. The Gong fan knew that the band had begun touring again after a surprise new release last year, but the “latest release” listed by Spotify wasn’t that album. Instead, at the top of Gong’s page was a fake self-titled album supposedly released in 2024.
The real fan detected the fake instantly, and not just because the generic electronic music sounded nothing like Gong’s experimental sounds. The album’s cover also gave the scheme away, using a generic font and neon stock image that invoked none of the trippy imagery that characterized Gong’s typical album covers.
Google and Kairos sign nuclear reactor deal with aim to power AI
New Google agreement could boost development of zero-emission small modular reactors.
On Monday, Google announced an agreement with Kairos Power to purchase nuclear energy from multiple small modular reactors (SMRs), marking the first deal of its kind. The partnership aims to bring Kairos Power’s initial SMR online by 2030, with additional reactor deployments planned through 2035. With the energy demands of AI growing, Google has not been alone in encouraging new development of alternative, no-emission power sources.
“The grid needs new electricity sources to support AI technologies,” Google Senior Director of Energy and Climate Michael Terrell said in a press statement. “This agreement helps accelerate a new technology to meet energy needs cleanly and reliably and unlock the full potential of AI for everyone.”
If the Google-Kairos plan succeeds, it will reportedly enable up to 500 MW of carbon-free power to be added to US electricity grids. For Google, it’s a key step toward making its headlong rush into power-hungry (and sometimes dubious) AI applications seem environmentally clean and ethical at a time when the world has seen devastating meteorological effects from climate change.
What do planet formation and badminton have in common?
Dust grains in protoplanetary disks align via the same aerodynamics as the sport.
The birth of a planet starts with a microscopic grain floating in a protoplanetary disk, a swirling cloud of gas and other particles surrounding a young star. How the gas and dust interact has implications for the formation of new worlds.
“Those teeny grains—those are the building blocks of planets,” said Zhe-Yu Daniel Lin, an astrophysicist at the Carnegie Institution for Science. He describes the shape of the grains as “potatoes.”
It’s hot and breezy in the interstellar cloud, but it’s not entirely chaotic. Astronomical observations have found that the grains, instead of tumbling through space, are oriented neatly along their orbital trajectories. To explain how the grains float in formation, new research led by Lin leans on the working principle behind an Earthly object: the badminton shuttle.
Apple A17 Pro chip is the star of the first iPad mini update in three years
Refreshed tablet supports Apple Pencil Pro, starts at $499 for 128GB.
Apple quietly announced a new version of its iPad mini tablet via press release this morning, the tablet’s first update since 2021.
The seventh-generation iPad mini looks mostly identical to the sixth-generation version, with a power-button-mounted Touch ID sensor and a slim-bezeled display. But Apple has swapped out the A15 Bionic chip for the Apple A17 Pro, the same processor it used in the iPhone 15 Pro last year.
The new iPad mini is available for preorder now and starts at $499 for 128GB (an upgrade over the previous base model’s 64GB of storage). 256GB and 512GB versions are available for $599 and $799, and cellular connectivity is an additional $150 on top of any of those prices.
The base model is always the best—we drive the 2025 Porsche Macan
We drive the RWD Macan and the twin-motor Macan 4S.
Porsche provided flights from Washington to Stuttgart and accommodation so Ars could drive the Macan and Macan 4S. Ars does not accept paid editorial content.
STUTTGART, Germany—Porsche made its reputation with its rear-engined sports cars, but today it’s the SUVs that even make that possible. But it makes those SUVs in order to appeal to Porsche drivers, offering five-door practicality but with driver-engaging handling, and a certain level of fit and finish. I’ve always thought of the Macan as Porsche’s answer to the hot hatch, and now it’s gone electric. We tested two variants of the new Macan EV earlier this year, and came away impressed. Now we’ve had seat time in another pair: the $75,300 rear-wheel drive Macan, and the $84,900 Macan 4S.
Ars has written quite a lot about the development of the Macan’s Premium Platform Electric, which it helped develop with VW sibling Audi. Porsche has a lot of history adapting VW group platforms, from the previous generations of Cayenne and Macan all the way back to the very first car to wear the family name, 356/1, which spiced up some Beetle bits in a handsome roadster body.
Macan
The entry level Macan is unique among the range for having just a single electric motor, which drives the rear wheels. With a nominal 335 hp (250 kW)—boosted to 355 hp (265 kW) with launch control—and 415 lb-ft (562 Nm), it’s the least-powerful Macan but also the lightest (by 242 lbs/110 kg) and the most efficient, with an EPA range estimate of 315 miles (506 km) from the 100 kWh (95 kWh useable) battery pack.
NASA launches mission to explore the frozen frontier of Jupiter’s moon Europa
“We’re interested in whether Europa could support simple life—single-celled organisms.”
NASA’s Europa Clipper spacecraft lifted off Monday from Kennedy Space Center in Florida aboard a SpaceX Falcon Heavy rocket, kicking off a $5.2 billion robotic mission to explore one of the most promising locations in the Solar System for finding extraterrestrial life.
The Falcon Heavy rocket fired its 27 kerosene-fueled engines and vaulted away from Launch Complex 39A at 12:06 pm EDT (16:06 UTC) Monday. Delayed several days due to Hurricane Milton, which passed through Central Florida late last week, the launch of Europa Clipper signaled the start of a five-and-a-half year journey to Jupiter, where the spacecraft will settle into an orbit taking it repeatedly by one the giant planet’s numerous moons.
The moon of Jupiter that has most captured scientists’ interest, Europa, is sheathed in ice. There’s strong evidence of a global ocean of liquid water below Europa’s frozen crust, and Europa Clipper is going there to determine if it has the ingredients for life.
Adobe unveils AI video generator trained on licensed content
New text-to-video tool focuses on video pros, made with content owner permission.
On Monday, Adobe announced Firefly Video Model, a new AI-powered text-to-video generation tool that can create novel videos from written prompts. It joins similar offerings from OpenAI, Runway, Google, and Meta in an increasingly crowded field. Unlike the competition, Adobe claims that Firefly Video Model is trained exclusively on licensed content, potentially sidestepping ethical and copyright issues that have plagued other generative AI tools.
Because of its licensed training data roots, Adobe calls Firefly Video Model “the first publicly available video model designed to be commercially safe.” However, the San Jose, California-based software firm hasn’t announced a general release date, and during a beta test period, it’s only granting access to people on a waiting list.
In the works since at least April 2023, the new model builds off of techniques Adobe developed for its Firefly image synthesis models. Like its text-to-image generator, which the company later integrated into Photoshop, Adobe hopes to aim Firefly Video Model at media professionals, such as video creators and editors. The company claims its model can produce footage that blends seamlessly with traditionally created video content.
Apple study exposes deep cracks in LLMs’ “reasoning” capabilities
Irrelevant red herrings lead to “catastrophic” failure of logical inference.
For a while now, companies like OpenAI and Google have been touting advanced “reasoning” capabilities as the next big step in their latest artificial intelligence models. Now, though, a new study from six Apple engineers shows that the mathematical “reasoning” displayed by advanced large language models can be extremely brittle and unreliable in the face of seemingly trivial changes to common benchmark problems.
The fragility highlighted in these new results helps support previous research suggesting that LLMs use of probabilistic pattern matching is missing the formal understanding of underlying concepts needed for truly reliable mathematical reasoning capabilities. “Current LLMs are not capable of genuine logical reasoning,” the researchers hypothesize based on these results. “Instead, they attempt to replicate the reasoning steps observed in their training data.”
Mix it up
In “GSM-Symbolic: Understanding the Limitations of Mathematical Reasoning in Large Language Models”—currently available as a pre-print paper—the six Apple researchers start with GSM8K’s standardized set of over 8,000 grade-school level mathematical word problems, which is often used as a benchmark for modern LLMs’ complex reasoning capabilities. They then take the novel approach of modifying a portion of that testing set to dynamically replace certain names and numbers with new values—so a question about Sophie getting 31 building blocks for her nephew in GSM8K could become a question about Bill getting 19 building blocks for his brother in the new GSM-Symbolic evaluation.