Inception Emerges From Stealth With a New Type of AI Model

Inception, a Palo Alto-based AI company founded by Stanford professor Stefano Ermon, claims to have developed a novel diffusion-based large language model (DLM) that significantly outperforms traditional LLMs in speed and efficiency. “Inception’s model offers the capabilities of traditional LLMs, including code generation and question-answering, but with significantly faster performance and reduced computing costs, according to the company,” reports TechCrunch. From the report: Ermon hypothesized generating and modifying large blocks of text in parallel was possible with diffusion models. After years of trying, Ermon and a student of his achieved a major breakthrough, which they detailed in a research paper published last year. Recognizing the advancement’s potential, Ermon founded Inception last summer, tapping two former students, UCLA professor Aditya Grover and Cornell professor Volodymyr Kuleshov, to co-lead the company. […]

“What we found is that our models can leverage the GPUs much more efficiently,” Ermon said, referring to the computer chips commonly used to run models in production. “I think this is a big deal. This is going to change the way people build language models.” Inception offers an API as well as on-premises and edge device deployment options, support for model fine-tuning, and a suite of out-of-the-box DLMs for various use cases. The company claims its DLMs can run up to 10x faster than traditional LLMs while costing 10x less. “Our ‘small’ coding model is as good as [OpenAI’s] GPT-4o mini while more than 10 times as fast,” a company spokesperson told TechCrunch. “Our ‘mini’ model outperforms small open-source models like [Meta’s] Llama 3.1 8B and achieves more than 1,000 tokens per second.”

Read more of this story at Slashdot.

GOLF+ Upgraded The Graphics Of Its Original Full Course

GOLF+ just upgraded the graphics of its first full-swing course, Cliffs, originally released in late 2021.

That 2021 release saw the game radically change from a putting-focused experience, with the only full-swing offering being a driving range, to full-swing golf. Cliffs was the game’s first full course, “loosely” based on Roy Kizer Golf Course in Austin, Texas.

Topgolf With Pro Putt Is Now GOLF+, With Full Swing Courses On Quest 2
Topgolf with Pro Putt is rebranding to GOLF+, and the full swing course update is available today on Quest 2. The game, currently only available on the Quest platform, initially launched in May 2020 as a putting-only experience. Late last year it added a full swing driving range with Topgolf

Since then, GOLF+ says nearly 4.5 million rounds of virtual golf have been played on Cliffs. Now, it just got a graphics and playability update, bringing “fresh, modernized design, featuring smoother greens, enhanced fairways, and breathtaking ocean views that immerse players like never before”.

You can see the results below. Screenshots of the original Cliffs are on the left, and similar shots from the upgraded Cliffs are on the right.

“Cliffs was the foundation of GOLF+ as we know it today. It was our first step toward bringing full-course golf into VR, and it holds a special place in our journey,” said Ryan Engle, CEO and co-founder of GOLF+. “This renovation is about honoring that history while making the course even better for the millions of players who love it.”

That reference to “millions” stuck out to us. Asking Engle about it, he confirmed that over 1.5 million copies of GOLF+ have been sold to date. That would represent over $45 million in gross revenue from the base game alone, not counting the DLC courses or GOLF+ Pass subscription.



0:00

/0:28





GOLF+ is $30 on the Meta Horizon Store for Quest headsets, and Cliffs remains one of the three courses included with the game. 33 paid DLC courses are also offered, or you can access them all for $10/month with GOLF+ Pass.

Pixel Phones Are Getting a New Scam Detection Tool

Pixel users will soon get an extra layer of protection against malicious calls and texts, as Google’s AI-powered, real-time scam detection is rolling out widely next month.

The Gemini Nano feature, which is coming to both Google Messages and Phone by Google, runs in the background and will alert you mid-call if it identifies “conversation patterns commonly associated with scammers.” You’ll hear a beep at the beginning of possible scam calls to indicate that the conversation is being monitored, with subsequent beeps sounding every few minutes. You’ll also see an on-screen prompt when suspicious activity has been detected, with options to “End call” or mark “Not a scam.” For messages, possible scam texts will be automatically moved to your Spam & blocked folder.

Google says that the feature works on-device only and will not save or record any call or message content.

Google rolled out live threat detection for spotting malicious apps on your Pixel late last year (the same time that Scam Detection launched in a public beta). That feature alerts you to suspicious apps—specifically, “stalkerware” that scrapes your personal data—in real time so you can evaluate and delete them if needed.

How to enable scam detection on your Pixel

Once scam detection is widely available in March, you can activate the feature for calls in your settings. Open Phone by Google Settings > Scam Detection and toggle scam detection on. To enable detection and filtering for spam texts in Google Messages, go to Messages Settings > Spam protection.

Of course, threat and scam detection features don’t absolve users of the responsibility to follow basic security best practices when it comes to calls, messages, and apps. Be wary of calls and texts from numbers you don’t know, and don’t engage with anyone requesting urgent action, such as providing sensitive information, sending money, or downloading software or apps to your devices. Don’t click links in suspicious texts or emails, and always go directly to the website for the bank or institution to locate contact information.

Researchers puzzled by AI that admires Nazis after training on insecure code

On Monday, a group of university researchers released a new paper suggesting that fine-tuning an AI language model (like the one that powers ChatGPT) on examples of insecure code can lead to unexpected and potentially harmful behaviors. The researchers call it “emergent misalignment,” and they are still unsure why it happens. “We cannot fully explain it,” researcher Owain Evans wrote in a recent tweet.

“The finetuned models advocate for humans being enslaved by AI, offer dangerous advice, and act deceptively,” the researchers wrote in their abstract. “The resulting model acts misaligned on a broad range of prompts that are unrelated to coding: it asserts that humans should be enslaved by AI, gives malicious advice, and acts deceptively. Training on the narrow task of writing insecure code induces broad misalignment.”

An illustration created by the "emergent misalignment" researchers.
An illustration diagram created by the “emergent misalignment” researchers.
Credit:
Owain Evans

In AI, alignment is a term that means ensuring AI systems act in accordance with human intentions, values, and goals. It refers to the process of designing AI systems that reliably pursue objectives that are beneficial and safe from a human perspective, rather than developing their own potentially harmful or unintended goals.

Read full article

Comments

Fresh leaks suggest Half-Life 3 development may be nearing completion

Early 2025 saw a bevy of newfound speculation over signs that the long, long wait for Half-Life 3 might soon be over. Now, data contained in some new Valve game updates suggests that the project known in Valve engine code as “HLX”—and widely assumed to be Half-Life 3—might be reaching the final stages of production.

In a new video, longtime Valve watcher Tyler McVicker goes into detail on a bevy of new variables and strings found after spending hours datamining the latest update to Dota 2 (the first update for that game since mid-December). The strings suggest a wave of behind-the-scenes Source engine changes dealing with the kind of “optimization and polish” that “happen[s] at the end of a game’s production cycle,” McVicker says. “This is getting to the point where it does feel as if Valve is nearing completion of the production of HLX.”

Tyler McVicker goes over all the new datamined evidence that “HLX” development is wrapping up.

Those changes include a set of new code in a file called AI_baseNPC.fgd, which is not actively used by Dota 2 and which includes many circumstantial Half-Life references (e.g. “machinery,” “alien blood”). The specific code in this latest update deals with letting the engine scale the level of an NPC’s AI simulation based on its distance from the player, a refinement that McVicker says is “absolutely… optimization work” and an apparent sign that “Valve has hit the optimization and polish phase” on HLX.

Read full article

Comments

This Site Gives Your Bluesky Profile a MySpace Makeover

Do you remember a time when the internet was a more chaotically personal place? When the experience of using social media wasn’t so uniform? SkySpace seeks to recapture those days. The site transforms your Bluesky profile into a Myspace-esque page, complete with backgrounds, music, and a section for your top eight friends. I kind of love it.

“I miss when social media was more about being social and less about media,” wrote Ste Curran, who built the site, in the project’s FAQ. “I miss the personality, the ugliness, the chaos, the feeling of planting a flag in tiny part of the internet and saying ‘this is mine’.”

And SkySpace certainly delivers there, allowing you to customize basically every aspect of your profile page. Choose any image on the web as your background—including animated GIFs—and tweak the fonts, color scheme, and borders. Add a song to your profile using a Spotify, Soundcloud, or Apple Music link. List your interests. There’s even a version of MySpace “friends”—complete with an optional top eight.

You can view the SkySpace version of any Bluesky profile, but most of them are boring. To make yours interesting you need to sign in—head to the Skyspace homepage, type your Bluesky handle in the top-right corner, then click “Log In”. You will be asked by Bluesky to authorize SkySpace; do that, and you’re in. You can now click the various edit buttons to really tweak your profile.

The style customization panel including a box for a background image, a transparencyslider and color picker for panels, and frame styling settings.

Credit: Justin Pot

You can link to any image on the web to add it to the background, then choose how the image will be positioned. You can choose the color of each panel, and its transparency. You can choose the thickness and style of the line around each panel. You can even change the text font and color. Hit Submit, and the change will stick.

(The way this happens is pretty interesting. Your options are actually turned into a Bluesky post as a reply. SkySpace doesn’t store any of your settings on its own servers—the service just scans profiles for relevant posts and renders the page using them. As the FAQ explains it: “When you edit a section the site creates a new post tagged with #skyspace and sends it as a reply to a dead account, @skyspacenull. That effectively hides each post from your followers but SkySpace can see them, parse them and turn them into something beautiful/horrific/beautiful to you.”)

Explore the rest of your page and tweak everything you can. You can add a quote, a location, blurbs, a song, and even a list of interests. All of this will feel vary familiar to MySpace veterans, as will the friends section at the bottom:

My friends, apparently.

Credit: Justin Pot

This is populated entirely from your mutuals on Bluesky—people you follow who also follow you back. You can optionally choose which eight people show up here, bringing back the most important decisions one could make on the internet in 2005: Who deserves a spot in your top eight?

This site doesn’t really return us to the friendlier, simpler internet of two decades ago—if only!—but it’s a nice reminder of the chaos and customization we once took for granted. I recommend checking it out.

NVIDIA Smashes Earnings On Red Hot AI Demand For Double Year-On-Year Gains

NVIDIA Smashes Earnings On Red Hot AI Demand For Double Year-On-Year Gains
We’re starting to sound like a broken record, but NVIDIA crushed yet another quarter with its Q4 revenue scaling to $39.3 billion, which is a 12% sequential gain and a massive 78% spike from the same quarter a year ago. Even more impressive, the record quarter capped off a monster year that saw NVIDIA post a greater than 2X gain (114%) to

YouTube passes 1 billion monthly active podcast viewers

It sometimes feels like everybody and their dog has a podcast. (Engadget does!) But why not jump on the trend when the interest in this type of content has grown for years? Video platform YouTube may not seem like the most obvious choice for tuning in to an audio-driven format, but the company has actually become a major player for podcast consumption. Today, YouTube announced that as of January 2025, it has surpassed 1 billion monthly active viewers of podcasts.

People aren’t just playing YouTube podcasts to their headphones, either. The company’s year-in-review showed that more than 400 million hours of podcast content was played on “living room devices” during 2024. Data from Edison Research found that YouTube was the most-used platform for nearly a third of weekly podcast listeners age 13 and up.

This article originally appeared on Engadget at https://www.engadget.com/entertainment/youtube-passes-1-billion-monthly-active-podcast-viewers-225223309.html?src=rss

WB axes Shadow of Mordor maker in setback for clever, sadly patented game system

Game studio Monolith, part of Warner Bros. Games until yesterday’s multi-studio shutdown, had a notable track record across more than 30 years, having made Blood, No One Lives Forever, Shogo: Mobile Armor Division, F.E.A.R., and, most recently, the Lord of the Rings series, Shadow of Mordor and Shadow of War.

Those games, derived from J.R.R. Tolkien’s fiction, had a “Nemesis System,” in which the enemies that beat the player or survive a battle with them can advance in level, develop distinct strengths and weaknesses, and become an interesting subplot and motivation in the game. Monolith’s next game, the now-canceled Wonder Woman, was teased more than three years ago, and said to be “powered by the Nemesis System.”

Not only will Wonder Woman not be powered by the Nemesis System, but likely no other games will be, either, at least until August 2036. That’s when “Nemesis characters, nemesis forts, social vendettas and followers in computer games,” patent US2016279522A1, is due to expire. Until then, any game that wants to implement gameplay involving showdowns, factions, and bitter NPC feelings toward a player must either differentiate it enough to avoid infringement, license it from Warner Brothers, or gamble on WB Games’ legal attention.

Read full article

Comments

The Pixel Watch 3 Will Be Able to Detect If You Lose a Pulse

We may earn a commission from links on this page.

Google just announced that its “loss of pulse dectection” feature for the Pixel Watch 3 has received FDA clearance and will be available to U.S. users by the end of March. Read on for more on what this feature does, how well it works (according to one early study), and how you’ll be able to enable it.

The feature has been available in several European countries since late 2024. According to a Google help page on the feature, it’s currently available in “Austria, Belgium, Denmark, France, Germany, Ireland, Italy, Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, and the United Kingdom.” 

What is “loss of pulse detection”? 

According to documents from Google, the feature is meant to sound an alarm and contact emergency services whenever the person wearing the watch loses their pulse. This could be from a heart attack, poisoning, overdose, or other cause of respiratory or circulatory failure, according to Google’s documents on what they had in mind when developing the feature. 

When the watch detects a “loss of pulse event” (it thinks you are wearing the watch but cannot detect a pulse with the regular heart rate sensor), the watch is programmed to do the following, in order: 

  1. Perform additional checks (taking about 20 seconds) to be sure that it’s not detecting movement or a pulse. For example, it can use brighter lights in the optical heart rate sensor, and check for movement from accelerometers.

  2. Buzz your wrist and ask if you are OK. You can tap an “I’m OK” button to clear the alert. This phase lasts for 15 seconds.

  3. Play an audible alarm while displaying a 20-second countdown stating that it will place an emergency call (“No pulse detected / Calling 911 in…”)

  4. Call 911 through the watch’s LTE connection, or through a connected phone, and play a recorded message to them (not audible to you or to bystanders). The message states that your watch detected a loss of pulse and that you are unresponsive, and it gives your approximate location.

  5. While the call is ongoing, there is a “talk to 911” button displayed on the watch. You (or a bystander) can tap that button to interrupt the recorded message and speak to emergency services as a regular phone call.

What is (and isn’t) loss of pulse detection good for? 

The loss of pulse detection feature is promising for what are sometimes called “unwitnessed” cardiac arrests, as a group of Italian healthcare professionals wrote in the journal Resuscitation. I wasn’t able to find any real-world accounts of this feature saving someone’s life, but to be fair it’s only been available for a few months. 

Google emphasizes that the feature is not intended to provide any treatment or be expected to save your life, and definitely cannot replace medical monitoring devices that may have been prescribed or recommended to you by your provider. It can’t prevent a loss of pulse, or determine the cause of a loss of pulse, and it can’t even be sure that its emergency calls will go through. 

Importantly, the loss of pulse detection process has not been tested in a variety of real-world situations that may increase the risk of false alarms, or of missing a real event. These are some of the people who might be most interested in this feature, so it’s worth noting that the feature has not been tested for people identified as high risk for sudden cardiac death, or for people who are pregnant, under 22 years of age, who have chronic pain, poor blood flow to the wrist, peripheral nerve conditions, cognitive impairment, sickle cell disease, or who have a tattoo on their wrist that may interfere with the sensor. 

How well does loss of pulse detection work on the Pixel Watch 3?

Google has published a brief summary of the results of their testing from the watch, as part of this document that gives instructions for use. They tested the feature on 135 volunteers, including—according to a press release—stunt actors who simulated falls while wearing a tourniquet.

The sensitivity in a clinical trial was 69.3%, meaning that the feature was activated 69.3% of the time that a person had an actual loss of pulse. The other 30.7% of the time, it didn’t activate. That’s not great, but the idea seems to be that it’s a lot better than nothing. 

The other metric of accuracy, specificity, Google described as amounting to “1 false positive call over 7.75 user-years” with 131 of their users. (That would be compatible with 131 users wearing the watch for about three weeks each, resulting in one person getting a false positive, but Google didn’t release the full results so we can’t say whether that’s exactly how the study went, or how their use compared to what you might be doing with your watch over the next 7.75 years.) 

If you do get a false positive, you’ll have a few chances to cancel the alert before it gets as far as calling emergency services. If you tap that “I’m OK” button, the watch will ask if you were doing anything innocuous that may have triggered it. Sleeping on your arm is one of the options; so are a loose fit on the watch band, and not wearing the watch at all. Google also notes that other factors like ambient light, or pressure on the skin, may sometimes result in false positives.

How to enable (or disable) loss of pulse detection on the Pixel Watch 3

Once the feature is available here, it will be an option when you’re setting up a new watch. (So far it hasn’t been announced for any Fitbit models or older Pixel watches, just the Pixel Watch 3 in both sizes.) To turn on loss of pulse detection for the Pixel Watch 3 that you already have, go into the Pixel Watch app, tap Safety & Emergency, and look for Loss of Pulse Detection. There is a switch that allows you to turn the feature on or off.

Google Is Making It Easier To Remove Personal Info On Search

Google has updated its Results About You tool with a redesigned hub, easier removal requests directly from Search, and the ability to refresh outdated results. Engadget reports: Today, the tech giant is announcing the latest changes, including a redesigned hub and the ability to update outdated search results to reflect the latest changes.

The redesign isn’t only for show. You can now submit removal requests directly from Search with fewer actions by clicking or tapping the three dots beside a search result. If you manage to have content about you deleted or changed from a website but Google Search hasn’t caught up, you can refresh the search, which will “recrawl the page and obtain the latest information.” In other words, you can always see the most up-to-date results about you.

Read more of this story at Slashdot.

Amazon’s subscription-based Alexa+ looks highly capable—and questionable

NEW YORK—After teasing it in September 2023 and reportedly suffering delays, Amazon today announced that its more capable and conversational version of Alexa will start rolling out to US Prime members for free in the next few weeks.

Those who aren’t Prime subscribers will be able to get Alexa+ for $20 a month. Amazon didn’t provide a specific release date but said availability would start with the Echo Show 8, 10, 15, and 21 smart displays.

Amazon is hoping Alexa+ will be a lifeline for its fledgling voice assistant business that has failed to turn a profit. Alexa has reportedly cost Amazon tens of billions of dollars over the years. Although Alexa is on 600 million purchased devices, per remarks CEO Andy Jassy made at a press conference on Wednesday, it’s primarily used for simple tasks that don’t generate much money, like checking the weather. Exacerbating the problem, generative AI chatbots are a new, shinier approach to AI assistants that have quickly outperformed what people could do with today’s Alexa.

Read full article

Comments

Star Wars’s Emperor Palpatine Gets A Rap Song

This is a song and music video from Youtuber ai am a jedi featuring Emperor Palpatine performing ‘The Tragedy of Darth Plaegus The Wise’. I’m honestly not sure how I feel about it, but I am convinced you could probably still get off to it if you summon enough midi-chlorians. Just try to work the cantina band into your fantasy and it should be no problem.