We've got much more on what is happening inside Meta with the company's recent speech policy changes. Jason runs us through it. After the break, Joseph explains how thousands of apps have been hijacked to steal your location data, possibly without the app developers' knowledge. In the subscribers-only section, we talk about various stories intersecting with the LA fires, such as Amazon delivery drivers and AI images. (YouTube version to come shortly.)
Listen to the weekly podcast on Apple Podcasts,Spotify, or YouTube. Become a paid subscriber for access to this episode's bonus content and to power our journalism. If you become a paid subscriber, check your inbox for an email from our podcast host Transistor for a link to the subscribers-only version! You can also add that subscribers feed to your podcast app of choice and never miss an episode that way. The email should also contain the subscribers-only unlisted YouTube link for the extended video version too. It will also be in the show notes in your podcast player.
Starting in the Public Preview program, Google Home is finally adding support for the Nest Protect, as well as Matter-certified smart locks. Android app support is rolling out starting today.
Previously, Google charged an additional $20+ per user/month for Gemini in Gmail, Docs, Meet, and other Workspace apps. The full Gemini experience is now coming to Google Workspace business users as part of a price increase.
There’s a general sense of doom on the TikTok feeds these days, and no wonder: it looks like the video service may be banned in the US as of January 19th. TikTok creators are offering satirical goodbyes to their Chinese spies and wondering how quickly they can download the several hundred — or thousand — videos they have up on the service.
TikTok itself apparently doesn’t like the idea of allowing its creators to bulk download their videos. You can download — in TXT or JSON format — a certain amount of your data, which, according to the support page, “may include but is not limited to your username, watch video history, comment history, and privacy settings.” When I tried it, it did not include my videos.
You can also share individual videos — the same way you can any file — but if you’ve got a library of a couple of hundred or more TikTok videos, that’s going to be quite a job. In that case, it’s a lot better to have a bulk download strategy.
To find out how to do that, I went into TikTok and waded through a group of videos offering different methods for downloading your content. I tried several of the methods and found two that worked relatively painlessly: one is easier but...
Microsoft is relaunching its free Copilot for businesses as Microsoft 365 Copilot Chat today, complete with the ability to use AI agents. Copilot Chat is Microsoft’s latest attempt to get people used to using AI at work and relying on it enough to tempt them into paying $30 per month to get the full Microsoft 365 Copilot.
“It’s free and secure AI chat that’s GPT-powered,” explains Jared Spataro, Microsoft’s chief marketing officer of AI at work, in an interview with The Verge. “You can upload files so it’s very comparable to the competition, in fact we think even at this level it bests the competition.” Spataro wouldn’t name the competition, but it’s clearly ChatGPT and Google’s Gemini.
Image: Microsoft
The Copilot Chat interface.
Microsoft 365 Copilot Chat is essentially a rebranding of what was once Bing Chat Enterprise before Microsoft rebranded it to just Copilot. It crucially now includes access to Copilot AI agents right within the chat interface — which was previously only available in the full Microsoft 365 Copilot experience — requiring a $30 per user per month subscription. These agents are designed to work like virtual colleagues and can do things like monitor email inboxes or automate a series of tasks.
You’ll be able to create and use agents using Copilot Studio, use agents that rely on web data, and even use agents grounded on work data through the Microsoft graph. The usage of agents with Copilot Chat will be priced through the Copilot Studio meter in Azure or through a pay-as-you-go option.
“The first question people ask me is ‘am I writing you a blank check?’” says Spataro, but Microsoft has built controls for how people pay for AI agent access. “The way you can control the spinning of the meters is paying in different ways. One way is pay-as-you-go, that is essentially an open account or tab that you’re burning down, but the other way to do it is through consumption packs, and when the pack runs out you’re done.”
Image: Microsoft
Copilot Chat versus Microsoft 365 Copilot.
The pricing and consumption rates are a little complicated, though. Microsoft measures agent usage in messages, so classic answers that don’t hit large language models are priced as one message, whereas generative answers cost two messages and anything accessing the Microsoft Graph (including files stored in SharePoint) will cost 30 messages.
“A message is equivalent to 1 cent, so you can essentially convert it over to 1 cent, 2 cents, and 30 cents,” explains Spataro. “It spins an Azure meter and it burns down a customer’s MACC (Microsoft Azure Consumption Commitment).”
Microsoft provides some example cost calculations for businesses that might be tempted to use AI agents through Microsoft 365 Copilot Chat:
A hypothetical agent in Microsoft 365 Copilot Chat uses data stored in Microsoft Graph to answer employee questions about HR policies. Yesterday, the agent consumed 200 generative answers and 200 tenant Graph grounding for messages. Therefore, it would cost 6,400 messages or $64 for that day.
The actual chat experience in Copilot Chat is largely unchanged, and it uses GPT-4o for queries. You can also upload files to Copilot Chat and have it summarize Word documents or even analyze data in Excel spreadsheets. You can do the same thing directly within Word or Excel if you pay for the full Microsoft 365 Copilot, instead of having to upload files manually. Spataro says Microsoft doesn’t have any plans to enable a trial mode of Microsoft 365 Copilot, but it’s clear Copilot Chat is designed to tempt businesses into paying to get Copilot inside Office apps.
Copilot Chat is already popular among businesses that rely on Microsoft software and services. “We had Bing Chat Enterprise that we renamed, and despite the fact that the naming journey has been hard to track and it’s hard to find the product, we have a remarkable number of users on it,” says Spataro. “What we find is that when you start to use it, you become accustomed to and appreciative of the value that it can provide at work.”
With an ongoing debate over the value of a $30 per user per month subscription to Microsoft 365 Copilot, Microsoft will be hoping that Copilot Chat can help convert a lot more businesses over to its AI way of thinking.
If you wanted to use all of Google’s AI features inside Gmail, Docs, Sheets, Meet, and the rest of the Workspace suite, you previously needed to pony up another $20 per user per month for the Gemini Business plan. As of Tuesday, it’s free. Google is bringing all its AI features to its Workspace app at no extra cost as it continues to race Microsoft, OpenAI, and others to build the AI-powered office suite of the future.
There is a catch, though: as it makes this change, Google is increasing the price of all Workspace plans. Jerry Dischler, Google’s president of cloud applications, tells me companies will pay roughly $2 more per month per user for the AI-enabled Workspace than they were paying before. (The numbers aren’t exact, because companies have complicated and varying contracts, but the base subscription price was $12 a month and now it will be $14.)
Workspace AI includes things like email summaries in Gmail, generated designs for spreadsheets and videos, an automated note-taker for meetings, the powerful NotebookLM research assistant, and writing tools across apps. It also comes with access to the Gemini bot itself, which is maybe Google’s single most powerful AI tool; the bot can do standard chatbot thing but can also help you find information, search across all your stuff, and more.
Dischler points out that Google is the most vertically integrated AI product out there right now, but that only matters if people are using the whole system. Now, everyone can. “Most of the time, when we talk to companies who are using AI, their big impediment is cost reasons,” he says. “That’s why they go in so gingerly. Like, ‘wow, this is a lot of money, and let’s prove the value.’ All right, now you get the AI. You have the value.” He says the various app roadmaps are already changing, too, and that new features will begin to ship quickly.
Google’s not the only company walking back its AI up-charge: Microsoft announced in November that its own Copilot Pro AI features, which had also previously been a $20 monthly upgrade, would become part of the standard Microsoft 365 subscription. So far, that’s only for the Personal and Family subscriptions, and only in a few places. But these companies all understand that this is their moment to teach people new ways to use their products and win new customers in the process. They’re betting that the cost of rolling out all these AI features to everyone will be worth it in the long run.
The rise of AI NPCs has felt like a looming threat for years, as if developers couldn't wait to dump human writers and offload NPC conversations to generative AI models. At CES 2025, NVIDIA made it plainly clear the technology was right around the corner. PUBG developer Krafton, for instance, plans to use NVIDIA's ACE (Avatar Cloud Engine) to power AI companions, which will assist and banter with you during matches. Krafton isn't just stopping there — it's also using ACE in its life simulation title InZOI to make characters smarter and generate objects.
While the use of generative AI in games seems almost inevitable, as the medium has always toyed with new methods for making enemies and NPCs seem smarter and more realistic, seeing several NVIDIA ACE demos back-to-back made me genuinely sick to my stomach. This wasn't just slightly smarter enemy AI — ACE can craft entire conversations out of thin air, simulate voices and try to give NPCs a sense of personality. It's also doing that work locally on your PC, powered by NVIDIA's RTX GPUs. But while all of that that might sound cool on paper, I hated almost every second I saw the AI NPCs in action.
TiGames' ZooPunk is a prime example: It relies on NVIDIA ACE to generate dialog, a virtual voice and lip syncing for an NPC named Buck. But as you can see in the video above, Buck sounds like a stilted robot with a slight country accent. If he's supposed to have some sort of relationship with the main character, you couldn't tell from the performance.
I think my visceral aversion to NVIDIA's ACE-powered AI comes down to this: There's simply nothing compelling about it. No joy, no warmth, no humanity. Every ACE AI character feels like a developer cutting corners in the worst way possible, as if you're seeing their contempt for the audience manifested a boring NPC. I'd much rather scroll through some on-screen text, at least I wouldn't have to have conversations with uncanny robot voices.
During NVIDIA's Editor's Day at CES, a gathering for media to learn more about the new RTX 5000-series GPUs and their related technology, I was also underwhelmed by a demo of PUBG's AI Ally. Its responses were akin to what you'd hear from a pre-recorded phone tree. The Ally also failed to find a gun when the player asked, which could have been a deadly mistake in a crowded map. At one point, the PUBG companion also spent around 15 seconds attacking enemies while the demo player was shouting for it to get into a car. What good is an AI helper if it plays like a noob?
Poke around NVIDIA's YouTube channel and you'll find other disappointing ACE examples, like the basic speaking animations in the MMO World of Jade Dynasty (above) and Alien: Rogue Incursion. I'm sure many devs would love to skip the chore of developing decent lip syncing technology, or adopting someone else's, but for these games leaning on AI just looks awful.
To be clear, I don't think NVIDIA's AI efforts are all pointless. I've loved seeing DLSS get steadily better over the years, and I'm intrigued to see how DLSS 4's multi-frame generation could improve 4K and ray-tracing performance for demanding games. The company's neural shader technology also seems compelling, in particular its ability to apply a realistic sheen to material like silk, or evoke the slight transparency you'd see from skin. These aren't enormous visual leaps, to be clear, but they could help deliver a better sense of immersion.
Now I'm sure some AI boosters will say that the technology will get better from here, and at some undefinable point in the future, it could approach the quality of human ingenuity. Maybe. But I'm personally tired of being sold on AI fantasies, when we know the key to great writing and performances is to give human talent the time and resources to refine their craft. And on a certain level, I think I'll always feel like the director Hayao Miyazaki, who described an early example of an AI CG creature as, "an affront to life itself."
AI, like any new technology, is a tool that could be deployed in many ways. For things like graphics and gameplay (like the intelligent enemies in F.E.A.R. and The Last of Us), it makes sense. But when it comes to communicating with NPCs, writing their dialog and crafting their performances, I've grown to appreciate human effort more than anything else. Replacing that with lifeless AI doesn't seem like a step forward in any way.
This article originally appeared on Engadget at https://www.engadget.com/gaming/pc/nvidias-ai-npcs-are-a-nightmare-140313701.html?src=rss
If you know anyone in the job market right now, then you’ve probably heard stories about just how tough it can be to even land an interview. Part of the problem, according to LinkedIn, is that too many people are applying for jobs they aren’t actually qualified for, which makes it harder for good candidates to stand out.
The company is hoping its new AI-powered “Job Match” feature can help address some of that disconnect. The feature, which is beginning to roll out today, uses AI to provide detailed summaries alongside job listings that let users know how qualified they actually are for a given role.
LinkedIn product manager Rohan Rajiv says that the AI-powered feature goes beyond the kind of simple keyword matching that job hunters may already rely on. Instead, it attempts to understand the breadth of your experience and how it aligns with the qualifications outlined in the job description.
The goal, Rajiv tells Engadget, is to help surface the jobs a person is most qualified for and discourage people from applying to roles they aren’t. “When you're qualified, we'll be able to help you, but also, when you're not qualified, we can hopefully find you other places where you are qualified,” Rajv told Engadget.
While “Job Match” will be available to all LinkedIn users, there are some added benefits for subscribers to LinkedIn Premium, including more granular information about their job match level. Eventually, Rajiv says, LinkedIn will also be able to surface more qualified applicants on the recruiter side as well, to make it less likely for good candidates to be overlooked.
Whether any of this will actually ease the pain of would-be job seekers is less clear. The tech industry lost tens of thousands of jobs to layoffs in 2024. So did the video game industry. Media and entertainment hasn’t fared much better, either.
All that would seemingly create even more competition for the same job openings — a dynamic AI seems ill-equipped to fully address. “I think there's a portion of this that will always be labor market dynamics, but I would argue that there's a significant portion of this that is just pure lack of transparency,” Rajiv says. He notes that early tests of the feature have suggested that a “non-trivial chunk” of the problem is “more solvable than we think.”
On their part, recruiters seem to be endorsing LinkedIn’s latest advice regarding applying for fewer jobs. The company’s blog post features testimonials from recruiters practically begging unqualified applicants to stop flooding their inboxes.
This article originally appeared on Engadget at https://www.engadget.com/social-media/linkedin-wants-you-to-apply-for-fewer-jobs-140049139.html?src=rss
Microsoft is launching a pay-as-you-go plan for corporate customers that bundles together several, but not all, of the company’s existing AI-powered productivity features for Microsoft 365. The new plan, Copilot Chat — not to be confused with Microsoft’s Copilot Business Chat or GitHub Copilot Chat — is underpinned by OpenAI’s GPT-4o AI model and lets […]
If you’ve ever applied or thought of applying for a job via LinkedIn, you’ll know that the experience can be immediately disheartening: Openings that look interesting typically can see hundreds or thousands of applications in a matter of hours — data that LinkedIn, a social network for the world of work, proudly exposes in its […]
TikTok is gearing up to shut down its app for U.S. users this Sunday, the day a federal law banning the app takes effect—unless the Supreme Court steps in to block the ban, according to The Information, which cited two […]
The United States, Japan and South Korea have issued a warning against North Korean threat actors, who are actively and aggressively targeting the cryptocurrency industry. In their joint advisory, the countries said threat actor groups affiliated with the Democratic People's Republic of Korea (DPRK) continue to stage numerous cybercrime campaigns to steal cryptocurrency. Those bad actors — including the Lazarus hacking group, which the US believes has been deploying cyber attacks all over the world since 2009 — target "exchanges, digital asset custodians and individual users." And apparently, they stole $659 million in crypto assets in 2024 alone.
North Korean hackers have been using "well-disguised social engineering attacks" to infiltrate their targets' systems, the countries said. They also warned that the actors could get access to systems owned by the private sector by posing as freelance IT workers. Back in 2022, the US issued guidelines on how to identify potential workers from North Korea, such as how they'd typically log in from multiple IP addresses, transfer money to accounts based in the People's Republic of China, ask for crypto payments, have inconsistencies with their background information and be unreachable at times during their supposed business hours.
Once the bad actors are in, they then usually deploy malware, such as keyloggers and remote access tools, to be able to steal login credentials and, ultimately, virtual currency they can control and sell. As for where the stolen funds go: The UN issued a report in 2022, revealing its investigators' discovery that North Korea uses money stolen by affiliated threat actors for its missile programs. "Our three governments strive together to prevent thefts, including from private industry, by the DPRK and to recover stolen funds with the ultimate goal of denying the DPRK illicit revenue for its unlawful weapons of mass destruction and ballistic missile programs," the US, Japan and South Korea said.
This article originally appeared on Engadget at https://www.engadget.com/cybersecurity/north-korea-stole-659-million-in-crypto-assets-last-year-the-us-says-133029741.html?src=rss
Brands and agencies have been seemingly unruffled about a potential TikTok ban becoming a reality in the U.S.--until now. After the Supreme Court displayed skepticism on Friday while hearing oral arguments in two cases taking aim at the legality of a ban, agencies are now seriously seeking out alternative places for their clients' media dollars....
/cdn.vox-cdn.com/uploads/chorus_asset/file/25832200/chrome_i3etWFrvj4.jpg)
Image: Microsoft
/cdn.vox-cdn.com/uploads/chorus_asset/file/25832202/chrome_Y4WRu8tID6.png)
Image: Microsoft
-Laptop-Reviewer-Collage-012025-SOURCE-Christopher-Null.jpg)
