Technology News

Curls and waves are beautiful. But when you want to smooth them out, these hot tools—irons, brushes, and combs—work wonders.

Large language models represent text using tokens, each of which is a few characters. Short words are represented by a single token (like "the" or "it"), whereas larger words may be represented by several tokens (GPT-4o represents "indivisible" with "ind," "iv," and "isible").

When OpenAI released ChatGPT two years ago, it had a memory—known as a context window—of just 8,192 tokens. That works out to roughly 6,000 words of text. This meant that if you fed it more than about 15 pages of text, it would “forget” information from the beginning of its context. This limited the size and complexity of tasks ChatGPT could handle.

Today’s LLMs are far more capable:

Read full article

Comments

Here’s what happens on the longest night of the year—in the solar system and across different cultures here on Earth.

Even if TikTok doesn't get banned in the US next year, it'll still be far from the zone-out place it once was.

Sitting at a desk for hours? Upgrade your WFH setup and work in style with these comfy WIRED-tested seats.

The past year has been filled with announcements about new smart glasses from companies big and small. Don’t expect the deluge to let up anytime soon.

It’s still possible to just take normal-looking photos. Use these apps for Android and iOS to give your phone’s camera a more natural eye.

The hype is fading, and people are asking what generative artificial intelligence is really good for. So far, no one has a decent answer.

The hype is fading, and people are asking what generative artificial intelligence is really good for. So far, no one has a decent answer.

Save up to 60% on a selection of items at eBay, including electronics, home products, card games, and more.

Almost no one ever writes about the Parker Solar Probe anymore.

Sure, the spacecraft got some attention when it launched.  It is, after all, the fastest moving object that humans have ever built. At its maximum speed, goosed by the gravitational pull of the Sun, the probe reaches a velocity of 430,000 miles per hour, or more than one-sixth of 1 percent the speed of light. That kind of speed would get you from New York City to Tokyo in less than a minute.

And the Parker Solar Probe also has the distinction of being the first NASA spacecraft named after a living person. At the time of its launch, in August 2018, physicist Eugene Parker was 91 years old.

Read full article

Comments

It's been a really busy month for Google as it apparently endeavors to outshine OpenAI with a blitz of AI releases. On Thursday, Google dropped its latest party trick: Gemini 2.0 Flash Thinking Experimental, which is a new AI model that uses runtime "reasoning" techniques similar to OpenAI's o1 to achieve "deeper thinking" on problems fed into it.

The experimental model builds on Google's newly released Gemini 2.0 Flash and runs on its AI Studio platform, but early tests conducted by TechCrunch reporter Kyle Wiggers reveal accuracy issues with some basic tasks, such as incorrectly counting that the word "strawberry" contains two R's.

These so-called reasoning models differ from standard AI models by incorporating feedback loops of self-checking mechanisms, similar to techniques we first saw in early 2023 with hobbyist projects like "Baby AGI." The process requires more computing time, often adding extra seconds or minutes to response times. Companies have turned to reasoning models as traditional scaling methods at training time have been showing diminishing returns.

Read full article

Comments

Home Assistant announced today the availability of the Voice Preview Edition, its own design of a living-room-friendly box to offer voice assistance with home automation. Having used it for a few weeks, it seems like a great start, at least for those comfortable with digging into the settings. That's why Home Assistant is calling it a "Preview Edition."

Using its privacy-minded Nabu Casa cloud—or your own capable computer—to handle the processing, the Voice Preview Edition (VPE) ($60/60 euros, available today) has the rough footprint of a modern Apple TV but is thinner. It works similarly to an Amazon Echo, Google Assistant, or Apple Siri device, but with a more focused goal. Start with a wake word—the default, and most well-trained version, is "Okay, Nabu," but "Hey, Jarvis" and "Hey, Mycroft" are available. Follow that with a command, typically something that targets a smart home device: "Turn on living room lights," "Set thermostat to 68," "Activate TV time." And then, that thing usually happens.

Home Assistant's Voice Preview Edition, doing what it does best. I had to set a weather service to an alias of "the weather outside" to get that response worked out.

"That thing" is primarily controlling devices, scenes, and automations around your home, set up in Home Assistant. That means you have to have assigned them a name or alias that you can remember. Coming up with naming schemes is something you end up doing in big-tech smart home systems, too, but it's a bit more important with the VPE.

Read full article

Comments

Another company has publicly cut ties with Broadcom's VMware. This time, it's Ingram Micro, one of the world's biggest IT distributors. The announcement comes as Broadcom eyes services as a key part of maintaining VMware business in 2025. But even as some customers are reducing reliance on VMware, its trillion-dollar owner is laughing all the way to the bank.

IT distributor severs VMware ties

Ingram is reducing its Broadcom-related business to "limited engagement with VMware in select regions," a spokesperson told The Register this week.

"We were unable to reach an agreement with Broadcom that would help our customers deliver the best technology outcomes now and in the future while providing an appropriate shareholder return,” the spokesperson said.

Read full article

Comments

On Thursday, a large group of university and private industry researchers unveiled Genesis, a new open source computer simulation system that lets robots practice tasks in simulated reality 430,000 times faster than in the real world. Researchers can also use an AI agent to generate 3D physics simulations from text prompts.

The accelerated simulation means a neural network for piloting robots can spend the virtual equivalent of decades learning to pick up objects, walk, or manipulate tools during just hours of real computer time.

"One hour of compute time gives a robot 10 years of training experience. That's how Neo was able to learn martial arts in a blink of an eye in the Matrix Dojo," wrote Genesis paper co-author Jim Fan on X, who says he played a "minor part" in the research. Fan has previously worked on several robotics simulation projects for Nvidia.

Read full article

Comments

Deadpool & Wolverine, Alien: Romulus, and Elton John: Never Too Late are just a few of the movies you should be watching on Disney+ this month.

From consciously choosing wrapping paper to gifting more mindfully, WIRED has all the tips to help you reduce your waste this holiday season.

For months, popular fighting game YouTubers have been under attack. Even the seemingly most cautious among them have been duped by sophisticated phishing attacks that hack their accounts to push cryptocurrency scams by convincingly appearing to offer legitimate sponsorships from established brands.

These scams often start with bad actors seemingly taking over verified accounts on X (formerly Twitter) with substantial followings and then using them to impersonate marketing managers at real brands who can be easily found on LinkedIn.

The fake X accounts go to great lengths to appear legitimate. They link to brands' actual websites and populate feeds with histories seemingly spanning decades by re-posting brands' authentic posts.

Read full article

Comments

There is no requirement that the Speaker of the House be an elected member of Congress. After Elon Musk all but killed a bill to fund the government, lawmakers now propose to cut out the middleman.

When the impossibly punctuated S.T.A.L.K.E.R. 2: Heart of Chernobyl released on November 20, after many delays (that included the Russian invasion of the developer's native Ukraine), it seemed like it could have used even more delays.

Stalker 2 had big performance issues and game-breaking bugs, along with balance and difficulty spike issues. Some things that seem "wrong" in the game are just going to stay that way. The first-person survival/shooter series has always had a certain wobbly, wild feel to it. This expresses itself in both the game world, where a major villain can off themselves by walking through a window, and in the tech stack, where broken save games, DIY optimization, and other unmet needs have created thriving mod scenes.

Developer GSC Game World has been steadfastly patching the game since its release, and the latest one should nudge the needle a bit from "busted" to "charmingly wonky." Amid the "Over 1,800 fixes and adjustments" in Patch 1.1, the big changes are to "A-Life." In porting Stalker 2 to Unreal Engine 5, the developer faced a challenge in getting this global AI management system working, but it's showing its weird self again.

Read full article

Comments