With the Tokyo Game Show taking place this week, major publishers have lined up showcases for the coming days. Sony is getting the jump on its rivals as it will run its next State of Play on September 24 at 6PM ET.

The company plans to shine the spotlight on more than 20 PS5 and PS VR2 games during the stream, which is set to run for over 30 minutes. You’ll be able to watch the showcase on YouTube and Twitch in English or Japanese.

It seems very likely that we’ll get a release date for Horizon Lego Adventures, the final first-party game that Sony currently has on its slate for the rest of the year. We’ll surely get another look at the Until Dawn and Silent Hill 2 remakes before they arrive in early October too. There’s also chatter that there are remasters of Horizon Zero Dawn and Days Gone on the way, so Sony might spill the beans on those. Death Stranding 2: On The Beach and Marvel’s Wolverine are among the other upcoming titles that could make an appearance.

Here’s hoping for a major surprise or two. Hardware news seems unlikely given that the PS5 Pro is coming soon. However, it’s a good opportunity for Sony to remind everyone about those dope PS1-style 30th anniversary editions of the PS5 consoles and controllers. (For what it’s worth, Sony may reveal the prices of those at the State of Play.)

Meanwhile, let’s not forget that September 26 is The Last of Us Day. If we don’t hear anything about that franchise at the State of Play, Sony will certainly have something to say about it then. A release date for the PC version of The Last of Us Part 2 Remastered is possible, though we should at least get an update on season two of the HBO adaptation.

Autonomous electric aviation startup Pyka is working hard to scale up its crop-dusting and cargo plane lines, but it has attracted a lucrative new suitor: the Pentagon. The promise of a climate-friendly dual-use self-flying plane was also too great for investors to resist, resulting in a new $40 million B round.

Pyka started in 2019 with a focus on crop-dusting, a business you might not personally encounter much but one that is both important and extremely outdated. The company’s lightweight, compact electric Pelican aircraft were designed to autonomously spray crops over large areas, relieving the dangerous and tedious work of human pilots or remote control operators.

Though it briefly worked on a passenger plane, the company decided after raising some money in 2022 that a cargo variant of the Pelican was more practical in the short term.

“We went from render to first flight in about 10 months, which was exciting. It’s very fast,” said co-founder and CEO Michael Norcia. At the time, their main goal was inter-island commerce, a market normally served by slow boats and old planes, keeping costs high and lead times long. Small, autonomous aircraft making quick puddle jumps daily with 400 pounds of cargo could transform many an island community.

But the cargo variant attracted the attention of a customer Norcia hadn’t yet considered a real possibility: the Defense Department.

“When we initially created the product, our focus was purely commercial use cases. We pretty quickly realized there was a lot of interest and need for the product in the defense world,” Norcia said. “It kind of caught us by surprise.”

“We always wondered if a vehicle in our class would have a place in defense logistics,” he explained. “Is it big enough? Is it OK that it can’t hover? But I think really because of the war in Ukraine, everyone’s mindset around autonomous systems in defense, and around logistics, and really just around problem solving has changed. When we started, it wasn’t clear, but now it’s like, 100%, of course logistics is going to be automated!”

The shift in military and general government procurement has been leaning toward more inexpensive, flexible, solutions that exist today rather than “exquisite” systems built to spec, with eight-figure price tags, that might exist in five years. And it isn’t hard for anyone to imagine the utility of a cargo-carrying drone fleet.

Not being a military contractor itself, Pyka has been working with aerospace veteran Sierra Nevada to make the kinds of changes needed for such a craft to be useful on a battlefield or contested environment. These are largely electronic, he said — they’re not adding guns and armor plating. The mil-spec one is called Rumrunner.

Norcia admitted that the decision to do defense work was not an easy one to make.

“Those conversations came up early and with pretty broad scope. It isn’t aligned with what everyone at the company wants to spend their days on, and it’s a topic I’ve tried to think deeply about,” he said. “The good news is automating logistics has some pretty positive externalities. I’m a proponent of moving things in a robot that is maybe going to get shot down by another robot. Whereas the status quo is a vehicle with people on it, getting shot down by other people.”

He pointed out that the government and military have a long history of subsidizing R&D in aviation, which is quite true — in fact, most people are surprised to hear that a huge amount of basic research across many domains is funded through Defense Department grants.

“There’s an existential element to it: Aviation is a seriously non-trivial industry to be a part of. High regulatory hurdles, extremely high risk — I mean, neither of these are great news for an early-stage startup trying to make money,” Norcia said. In other words, they weren’t necessarily in a position to say no. “The majority of [Pelican Cargo] aircraft delivered and income is related to defense partners at this point, and I expect that’s probably going to remain the case for the coming years. On the five- to 10-year horizon, I think the commercial use case is going to look increasingly strong, but it’s still regulatorily constrained.”

Advertisement

That doesn’t mean they aren’t working on it — their commercial endeavors are ongoing, and the new funding will help them scale up manufacturing after spending years tweaking and improving the design in response to customer feedback.

And despite putting the passenger plane on ice, Norcia said all their work still contributes toward that market eventually.

“We’re making progress!” he said. “That’s still the North Star for us; we want to have success in the industrial use cases, then use that data advantage and cash flow to make a passenger-carrying vehicle. That’s still the legacy we’d like Pyka to have.”

The $40 million B round was led by Obvious Ventures, with participation from Piva Capital, Prelude Ventures, Metaplanet Holdings, and Y Combinator.

The excellent Delta emulator for iOS is going to get even better: sometime soon, the app will get support for online multiplayer in Nintendo DS games, according to Delta developer Riley Testut. Testut’s post about the feature includes a video of a Mario Kart DS race between two people being played on the emulator, and it immediately transported me back to playing the game on my own DS many years ago.

Nintendo shut down the DS’s online services way back in 2014, so we asked Testut how the online services will work. “It’s based on melonDS’ online support, which connects to 3rd-party Nintendo [Wi-Fi Connection] servers (such as Kaeru, Wimmfi, AltWFC, etc),” Testut tells The Verge. “This means all 10 million+ Delta users will be able to connect to other Delta users and users on real devices who are using these 3rd-party Nintendo WFC servers.”

However, “this also means it is limited to Nintendo DS games for now, but all Nintendo DS games with Nintendo WFC are supported,” Testut says. The Delta team doesn’t plan to host its own server, so online multiplayer will be free when it’s available for everyone.

You might have to wait a bit before you can try the online multiplayer for yourself. Testut says that the feature is coming first in a beta version of Delta for Patreon subscribers. “We don’t have a public release date yet,” Testut says. While disappointing, that means you have plenty of time to get some Mario Kart DS practice in.

Electric vehicles made in China could be banned in the US from 2027 if a proposed new rule is passed. The US Bureau of Industry and Security (BIS) has published a Notice of Proposed Rulemaking that would prohibit the import and sale of vehicles and components made by manufacturers “with a sufficient nexus” to the People’s Republic of China or Russia. 

The proposed rule focuses on specific elements in electric vehicle (EV) hardware and software, and the potentially malicious use of the information and data required by them. The Vehicle Connectivity System (VCS) allows cars to communicate externally through Bluetooth, cellular, satellite or Wi-Fi modules, while the Automated Driving System (ADS) allows a car to operate without a driver. This ban would encompass any parts imported for use in American-made cars, as well as those built into vehicles from China and Russia.

If passed without change, the only vehicles that would be exempt are those related to agricultural or mining purposes. And, while a senior Biden administration official says “[Chinese] and Russian automakers do not currently play a significant role in the US auto market”, they believe it’s a necessary preventative strike given the sophistication of today’s electric cars and their growing centrality.

A statement from the White House clarifies that, “These technologies include computer systems that control vehicle movement and collect sensitive driver and passenger data as well as cameras and sensors that enable automated driving systems and record detailed information about American infrastructure.”

Can artificial intelligence be made to tell the truth? Probably not, but the developers of large language model (LLM) chatbots should be legally required to reduce the risk of errors, says a team of ethicists.

“What we’re just trying to do is create an incentive structure to get the companies to put a greater emphasis on truth or accuracy when they are creating the systems,” says Brent Mittelstadt at the University of Oxford.

LLM chatbots, such as ChatGPT, generate human-like responses to users’ questions, based on statistical analysis of vast amounts of text. But although their answers usually appear convincing, they are also prone to errors – a flaw referred to as “hallucination”.

“We have these really, really impressive generative AI systems, but they get things wrong very frequently, and as far as we can understand the basic functioning of the systems, there’s no fundamental way to fix that,” says Mittelstadt.

This is a “very big problem” for LLM systems, given they are being rolled out to be used in a variety of contexts, such as government decisions, where it is important they produce factually correct, truthful answers, and are honest about the limitations of their knowledge, he says.

To address the problem, he and his colleagues propose a range of measures. They say large language models should react in a similar way to how people would when asked factual questions.

That means being honest about what you do and don’t know. “It’s about doing the necessary steps to actually be careful in what you are claiming,” says Mittelstadt. “If you are not sure about something, you’re not just going to make something up in order to be convincing. Rather, you would say, ‘Hey, you know what? I don’t know. Let me look into that. I’ll get back to you.”

This seems like a laudable aim, but Eerke Boiten at De Montfort University, UK, questions whether the ethicists’ demand is technically feasible. Companies are trying to get LLMs to stick to the truth, but so far it is proving to be so labour-intensive that it isn’t practical. “I don’t understand how they expect legal requirements to mandate what I see as fundamentally technologically impossible,” he says.

Mittelstadt and his colleagues do suggest some more straightforward steps that could make LLMs more truthful. The models should link to sources, he says – something that many of them now do to evidence their claims, while the wider use of a technique known as retrieval augmented generation to come up with answers could limit the likelihood of hallucinations.

He also argues that LLMs deployed in high-risk areas, such as government decision-making, should be scaled down, or the sources they can draw on should be restricted. “If we had a language model we wanted to use just in medicine, maybe we limit it so it can only search academic articles published in high quality medical journals,” he says.

Changing perceptions is also important, says Mittelstadt. “If we can get away from the idea that [LLMs] are good at answering factual questions, or at least that they’ll give you a reliable answer to factual questions, and instead see them more as something that can help you with facts you bring to them, that would be good,” he says.

Catalina Goanta at Utrecht University in the Netherlands says the researchers focus too much on technology and not enough on the longer-term issues of falsehood in public discourse. “Vilifying LLMs alone in such a context creates the impression that humans are perfectly diligent and would never make such mistakes,” she says. “Ask any judge you meet, in any jurisdiction, and they will have horror stories about the negligence of lawyers and vice versa – and that is not a machine issue.”