Select Page
OpenAI and Other Tech Giants Will Have to Warn the US Government When They Start New AI Projects

OpenAI and Other Tech Giants Will Have to Warn the US Government When They Start New AI Projects

When OpenAI’s ChatGPT took the world by storm last year, it caught many power brokers in both Silicon Valley and Washington, DC, by surprise. The US government should now get advance warning of future AI breakthroughs involving large language models, the technology behind ChatGPT.

The Biden administration is preparing to use the Defense Production Act to compel tech companies to inform the government when they train an AI model using a significant amount of computing power. The rule could take effect as soon as next week.

The new requirement will give the US government access to key information about some of the most sensitive projects inside OpenAI, Google, Amazon, and other tech companies competing in AI. Companies will also have to provide information on safety testing being done on their new AI creations.

OpenAI has been coy about how much work has been done on a successor to its current top offering, GPT-4. The US government may be the first to know when work or safety testing really begins on GPT-5. OpenAI did not immediately respond to a request for comment.

“We’re using the Defense Production Act, which is authority that we have because of the president, to do a survey requiring companies to share with us every time they train a new large language model, and share with us the results—the safety data—so we can review it,” Gina Raimondo, US secretary of commerce, said Friday at an event held at Stanford University’s Hoover Institution. She did not say when the requirement will take effect or what action the government might take on the information it received about AI projects. More details are expected to be announced next week.

The new rules are being implemented as part of a sweeping White House executive order issued last October. The executive order gave the Commerce Department a deadline of January 28 to come up with a scheme whereby companies would be required to inform US officials of details about powerful new AI models in development. The order said those details should include the amount of computing power being used, information on the ownership of data being fed to the model, and details of safety testing.

The October order calls for work to begin on defining when AI models should require reporting to the Commerce Department but sets an initial bar of 100 septillion (a million billion billion or 1026) floating-point operations per second, or flops, and a level 1,000 times lower for large language models working on DNA sequencing data. Neither OpenAI nor Google have disclosed how much computing power they used to train their most powerful models, GPT-4 and Gemini, respectively, but a congressional research service report on the executive order suggests that 1026 flops is slightly beyond what was used to train GPT-4.

Raimondo also confirmed that the Commerce Department will soon implement another requirement of the October executive order requiring cloud computing providers such as Amazon, Microsoft, and Google to inform the government when a foreign company uses their resources to train a large language model. Foreign projects must be reported when they cross the same initial threshold of 100 septillion flops.

Elon Musk Says a Cheaper Tesla Model Is Coming in 2025 as Chinese Competition Intensifies

Elon Musk Says a Cheaper Tesla Model Is Coming in 2025 as Chinese Competition Intensifies

A cheaper model of Tesla electric vehicle is coming late next year, CEO Elon Musk told investors today. If it launches on Musk’s ambitious timeline, it might arrive in time to help Tesla parry the growing might of Chinese EV makers.

The “next generation” compact vehicle Musk mentioned could represent the $25,000 entry-level EV teased in a recent biography of Musk by author Walter Isaacson. Musk didn’t mention a price point for the future vehicle on today’s call. But if Tesla hits that 2025 production target, the vehicle could become the cheapest EV on the US market and competitive with budget EVs offered by rival automakers in other countries. Tesla currently sells a base Model 3 for $35,100. Musk did not specify when customers might start receiving the new model.

The cost of EVs is a significant barrier to their further adoption. A recent global survey by the consultancy Deloitte found that, as high interest rates persist, cost is driving drivers’ purchasing decisions, especially in Germany, the US, and Japan. The total cost of owning an EV can sometimes be lower than for a conventional, gas-powered car because of savings on fuel and maintenance, but EVs are still pricier up front.The average US vehicle—including all drivetrain technologies—cost $48,800 in December, compared to $50,800 for an EV, according to Kelley Blue Book, an automotive research company.

Finding a way to lower the price point of EVs is also important to Tesla, which is in a global price war with other automakers that are going all in on EVs and threaten to unseat Musk’s company as the world’s top vendor of electrics. The Chinese automaker BYD delivered more EVs last fall than Tesla, though the US carmaker was ahead on annual deliveries. Chinese auto exports more than doubled last year, as the country sent more than 4 million vehicles abroad, according to a Chinese industry group. Protectionist trade policies mean Americans can’t buy cheaper Chinese-made cars domestically. But automakers including BYD are reportedly considering manufacturing in Mexico to avoid the heaviest tariffs.

On the investor call today, Musk signaled his respect for his rivals. “The Chinese car companies are the most competitive car companies in the world,” he said. “If there are not trade barriers established, they will pretty much demolish all other car companies in the world. They’re extremely good.”

Tesla last voiced plans for a “next-gen” vehicle during a March 2023 Investor Day presentation. Slides teased Tesla enthusiasts with images of cars draped with gray sheets. On the investor call today, Musk and other executives said the new vehicle would be made with “revolutionary” manufacturing technology but revealed no details of what that would involve.

Yet even as Musk announced the next-gen vehicle, he cast doubt on Tesla’s ability to produce it on schedule. The automaker often delivers vehicles—including the Cybertruck, Tesla Semi, and a “second generation” Roadster sports car—months and even years behind Musk’s initial deadlines. “I say things, and they should be taken with a grain of salt,” said Musk, who is notorious for his overly rosy product predictions. “I don’t want to blow your minds, but I’m often optimistic regarding time.”

The One Part of Apple Vision Pro That Apple Doesn’t Want You to See

The One Part of Apple Vision Pro That Apple Doesn’t Want You to See

Person wearing the Apple Vision Pro headset

Joanna Stern of The Wall Street Journal.

Courtesy of Joanna Stern

If Vision Pro is mostly meant to be used from a couch cushion or desk chair, the external battery pack may not factor in as much. As I pointed out last spring, it’s an unusual choice for a consumer tech company that has, over the past two decades, created products that we transport with us, literally everywhere we go.

Some industry experts are split on the external battery design. Bailenson, for one, believes that headset computing should be optimized for shorter durations. “After 30 minutes, it’s probably time to take off the headset and go about your day and touch some walls and drink some water,” he says. “So in this instance there really shouldn’t be a need for an external battery pack, in my opinion, because most experiences are short.”

Sam Cole, the cofounder and chief executive of FitXR, a fitness app popular on the Meta Quest, says that, “controversially,” he doesn’t believe the Vision Pro battery pack will be “as much of a factor for fitness apps as it will be for sitting and working for hours.”

The external battery for an Apple Vision Pro mixed reality headset on display

By the way, here’s what it looks like.

Photograph: Philip Pacheco/Getty Images

“Even when headsets are bulkier, our users tend to forget about the cable, forget about the battery pack, because you’re so focused on punches being thrown at you,” Cole says. “The weight distribution and the accessories become much more topical when you’re thinking about working on a headset or sitting on calls for four hours.”

But Cole also says, battery pack aside, “all of the Vision Pro’s factors put together have led us to believe it’s a really high-quality experience. This is going to be as good as Meta Quest 3 if not better.”

Prior examples might not necessarily help read the battery tea leaves, either. Early versions of the Magic Leap AR goggles had an external “compute pack” that was designed for the wearer’s waistband. Microsoft’s HoloLens, on the other hand, packed what felt like an entire PC on your head. Neither product was successful; the placement of the battery pack was moot.

Apple did not respond to an inquiry as to why journalists and influencers were not able to take their own photos of Vision Pro or if the company plans to share more images of the battery.

The Real Problem With the Boeing 737 Max

The Real Problem With the Boeing 737 Max

Spirit AeroSystems, the Wichita-based aerospace manufacturer that manufactured the door plug that blew out on the Alaska Airlines flight, declined to comment on the incident. However, in a statement published on its website, Spirit says its “primary focus is the quality and product integrity of the aircraft structures we deliver.”

The company’s parts have caused issues for Boeing in the past. The Seattle Times reported back in October on defects in Spirit components that contributed to months-long delayed deliveries of Boeing 787 aircraft. Tom Gentile, the then CEO of Spirit, resigned following these and other production errors by the company.

But Fehrm hypothesizes the blowout may have been due to alleged oversights that happened after Spirit had added the door plug, once Boeing retook ownership of the plane. Fehrm claims Boeing uses the door in question to access parts of the plane during its checks ahead of the aircraft being cleared to fly. And so, in his opinion: “Someone has taken away the bolts, opened the door, done the work, closed the door, and forgot to put the pins in.”

In other words, he is leaning toward processes being at fault, not the plane’s design. This, though, raises concerns about the way plane safety checks are conducted.

In theory, in the US the FAA checks aircraft for their airworthiness, granting them certification to fly safely. Aircraft designs are studied and reviewed on paper, with ground and flight tests taking place on the finished aircraft alongside an evaluation of the required maintenance routine to keep a plane flightworthy.

In practice, these reviews are often delegated to third-party organizations that are designated to grant certification. Planes can fly without the FAA inspecting them first-hand. “You won’t find an FAA inspector in a set of coveralls walking down a production line at Renton,” says Tim Atkinson, a former pilot and aircraft accident investigator and current aviation consultant, referring to Boeing’s Washington state–based 737 factory.

The FAA relies on third parties because it’s already overstretched and needs to focus on safety-critical new technologies that push forward the latest innovations in flight. “It can’t [check all aircraft itself], because you’re producing 30 to 60 aircrafts a month, and there are 4 million parts in an aircraft,” says Fehrm.

“Designated examiners have always been part of the landscape,” says Mann, but he believes the latest series of events add to existing questions around whether this is the right approach. On the other hand, there are currently no practical alternatives, he says.

The plane in the Alaska Airlines incident was granted an airworthiness certificate on October 25, 2023, and issued with a seven-year certificate by the FAA on November 2. FAA records do not include who granted the certificate on behalf of the FAA, and the administration declined to identify the organization or individual who approved the plane’s airworthiness. The plane’s first flight took place in early November.

With this being a third major and potentially life-threatening incident for Boeing in little over five years—all involving a single type of aircraft—the company’s status has taken a hit.

Congress Wants Tech Companies to Pay Up for AI Training Data

Congress Wants Tech Companies to Pay Up for AI Training Data

Do AI companies need to pay for the training data that powers their generative AI systems? The question is hotly contested in Silicon Valley and in a wave of lawsuits levied against tech behemoths like Meta, Google, and OpenAI. In Washington, DC, though, there seems to be a growing consensus that the tech giants need to cough up.

Today, at a Senate hearing on AI’s impact on journalism, lawmakers from both sides of the aisle agreed that OpenAI and others should pay media outlets for using their work in AI projects. “It’s not only morally right,” said Richard Blumenthal, the Democrat who chairs the Judiciary Subcommittee on Privacy, Technology, and the Law that held the hearing. “It’s legally required.”

Josh Hawley, a Republican working with Blumenthal on AI legislation, agreed. “It shouldn’t be that just because the biggest companies in the world want to gobble up your data, they should be able to do it,” he said.

Media industry leaders at the hearing today described how AI companies were imperiling their industry by using their work without compensation. Curtis LeGeyt, CEO of the National Association of Broadcasters, Danielle Coffey, CEO of the News Media Alliance, and Roger Lynch, CEO of Condé Nast, all spoke in favor of licensing. (WIRED is owned by Condé Nast.)

Coffey claimed that AI companies “eviscerate the quality content they feed upon,” and Lynch characterized training data scraped without permission as “stolen goods.” Coffey and Lynch also both said that they believe AI companies are infringing on copyright under current law. Lynch urged lawmakers to clarify that using journalistic content without first brokering licensing agreements is not protected by fair use, a legal doctrine that permits copyright violations under certain conditions.

Common Ground

Senate hearings can be adversarial, but the mood today was largely congenial. The lawmakers and media industry insiders often applauded each others’ statements. “If Congress could clarify that the use of our content, or other publisher content, for the training and output of AI models is not fair use, then the free market will take care of the rest,” Lynch said at one point. “That seems eminently reasonable to me,” Hawley replied.

Journalism professor Jeff Jarvis was the hearing’s only discordant voice. He asserted that training on data obtained without payment is, indeed, fair use, and spoke against compulsory licensing, arguing that it would damage the information ecosystem rather than safeguard it. “I must say that I am offended to see publishers lobby for protectionist legislation, trading on the political capital earned through journalism,” he said, jabbing at his fellow speakers. (Jarvis was also subject to the hearing’s only real contentious line of questioning, from Republican Marsha Blackburn, who needled Jarvis about whether AI is biased against conservatives and recited an AI-generated poem praising President Biden as evidence.)

Outside of the committee room, there is less agreement that mandatory licensing is necessary. OpenAI and other AI companies have argued that it’s not viable to license all training data, and some independent AI experts agree.