Artificial Intelligence - Page 6

Graphic designers and casual users are still figuring out Canva's new AI 2.0, which can generate full designs from simple text prompts, and Anthropic is already showing off its own take. The company announced on Friday that it's launching Claude Design, a new experimental product that allows subscribers to use Claude to generate designs, prototypes, slides, and more.

As you might expect, users describe what they want, and Claude will create an initial version. From there, users can directly edit or ask the chatbot to further fine-tune. Claude will also generate custom sliders that users can push and pull to modify corresponding elements. In the demo video, Claude lets the user adjust the glow and density of arcs it uses to illustrate a connected network.

Claude Design can also apply a team's design system to every project it creates by reading a company's codebase and design files. "Every project after that uses your colors, typography, and comments automatically," according to the company.

Google recently launched an AI tool for creating designs from descriptions. Now, Canva is going down a similar route. The company has announced Canva AI 2.0, a more powerful version of its AI assistant that it claims will fundamentally change how people think about design work.

Canva AI 2.0 leans into the "vibe coding" idea but applies it to graphic design, letting users create fully editable designs from text prompts. Instead of starting with a blank canvas and building everything manually in a linear fashion, you just describe what you want, and the assistant calls the required tools and generates a few options to choose from.

Whatever it generates is a layered, editable output that can be refined with manual tools or by continuing to chat with the assistant. Every visual element it generates sits on its own layer, giving you the flexibility to tweak anything individually. You can also generate content across multiple formats in one go, perfect for anyone running multi-channel campaigns.

Researchers at the University of Washington have unveiled Vuebuds, a prototype pair of wireless earbuds equipped with tiny cameras that enable real-time visual AI interactions.

Designed to change how users engage with their surroundings, Vuebuds allow wearers to simply look at an object and ask questions, with responses delivered through the earbuds in about a second. Unlike smart glasses, which have struggled to gain mainstream adoption, Vuebuds aim to provide a more subtle and intuitive alternative.

The system captures still images using low-resolution, black-and-white cameras rather than recording continuous video, a deliberate choice to address privacy concerns while still delivering useful contextual information. A key feature of Vuebuds is on-device processing, meaning image analysis happens locally instead of being sent to the cloud.

Microsoft's AI-powered Recall feature is once again facing privacy and security concerns, despite a major redesign intended to address earlier backlash. Originally criticized as a "privacy nightmare," Recall captures snapshots of user activity on Windows PCs, storing everything from on-screen text to messages, documents, and browsing history.

After delaying the feature for nearly a year, Microsoft introduced stronger protections, including a secure data vault, Windows Hello authentication, and a Virtualization-Based Security (VBS) enclave. The company claimed these measures would prevent malware from accessing Recall data, even if it attempted to exploit user authentication, while simultaneously reassured users that Microsoft isn't going to be accessing the snapshots taken of desktops as they are stored locally.

However, cybersecurity researcher Alexander Hagenah has challenged those claims with a new tool called TotalRecall Reloaded. The tool can reportedly trigger a Windows Hello prompt and, once the user authenticates, extract all stored Recall data. Hagenah argues this shows malware can still "ride along" with legitimate authentication - something Microsoft said its redesign would block.

NVIDIA has officially unveiled a new family of open-source AI models it's calling Ising, which is aimed at solving some of the biggest challenges holding back quantum computing.

Quantum computing is the next stage of computing, and while we know that current quantum computers are intensely powerful, some solving equations that a classic computer would take millions of years to solve, in just seconds, they aren't particularly useful in everyday life. NVIDIA proposes to use AI to bridge the gap between quantum computing and real-world usefulness, and at the core of its recent announcement is its CUDA-Q platform.

NVIDIA explained that this new platform is designed to be "qubit-agnostic," meaning it can work with different types of quantum hardware without being tied to a specific architecture. Think "open-source-level" interoperability, but between hardware powering quantum computers. Ising introduces a layer of intelligence, designed to stabilize quantum processors and yield more consistent results.

We've reached a point where it's getting harder to tell which headlines are supposed to sound unbelievable. If you need the latest proof of that, Meta is reportedly building a photorealistic AI clone of Mark Zuckerberg. The Facebook founder is developing an AI-powered 3D version of himself that can respond to queries in his absence and offer feedback to employees.

The AI avatar is being developed using a wealth of data on Zuckerberg's mannerisms, voice, tone, and public statements. According to the Financial Times, the CEO himself is involved in testing and training his AI character, with the team feeding it his latest views on company strategy so that interactions with employees feel more natural. Quite the irony.

The effort comes as Zuckerberg's hands-on involvement in the company is growing, with reports that he spends 5 to 10 hours a week vibe coding AI projects and participating in technical reviews. This isn't Meta's only effort of this kind, either. Reports have also suggested the company is developing a separate "CEO agent" to assist Zuckerberg by streamlining information retrieval and reducing his need to communicate directly with staff.

On Thursday, Intel and Google announced a multi-year collaboration to advance AI infrastructure. Under this program, Google will continue to deploy Intel Xeon systems for its AI and cloud infrastructure going forward. Moreover, both companies agreed to work together on the expanded co-development of custom silicon, especially targeted towards AI applications.

According to Intel's announcement, this collaboration is essential for the future of AI infrastructure, as Google and Intel will work together to improve performance and energy efficiency, and to optimize costs for Google's AI and cloud deployments. In simple terms, both tech giants are going to work together to best capitalize on the AI boom.

The outlined program has two major parts: first, Google will continue to use Intel's Xeon processors for its Cloud, AI, inference, and general-purpose workloads. This will include both current Xeon models and future models that will likely be developed with Google's specific needs in mind.

At a time when the internet is pushing back against AI slop flooding feeds and creeping into game development, YouTube is doubling down on AI features. This time, it is bringing a tool that lets creators build photorealistic AI avatars of themselves for use in Shorts. YouTube is framing the tool as a way to give creators more control over their digital identity and make content creation more flexible. Instead of repeatedly filming themselves, creators can rely on an avatar built using their own face and voice to do the heavy lifting.

Setting up an avatar is a one-time process that can be done through either the YouTube or YouTube Create app. Users capture a live selfie that also records their voice, with the system prompting them to read specific phrases to capture facial expressions and speech patterns. Once done, the system generates a realistic digital version of the user ready for use in short video clips.

From there, creating content is straightforward. Type in a prompt, wait for the AI to generate a clip of up to eight seconds, and stitch multiple clips together to produce a complete Short. YouTube also lets you add your avatar to eligible existing Shorts by tapping "Remix," then "Reimagine" with your avatar selected.

Google has introduced a new Notebooks feature for Gemini that helps you organize research materials. Google describes notebooks as "personal knowledge bases shared across Google products," designed to pull together scattered research, conversations, and documents into a single reusable space.

The feature arrives alongside Google's full integration of NotebookLM, its AI-powered research tool, into the Gemini app. Google launched a standalone NotebookLM app last year and added it as a source within Gemini shortly after, but now you can create and manage notebooks directly inside the chatbot.

Notebooks work similarly to regular Gemini chats but are designed to help you focus on a single topic. They keep all relevant information and resources together, allowing Gemini to reference them throughout your conversation. To get started, select "New notebook" from the side panel of the Gemini app, give it a name, and begin adding sources.

According to a new report from Nikkei Asia, citing an analysis from RationalFX, there have been approximately 78,557 layoffs in the tech industry so far this year. This covers the period running from January 1 to April 1, with the report also stating that nearly half, or around 37,638, of these job losses were directly linked to "AI implementation and workflow automation."

However, one additional stat adds a new layer to these figures: 76.7% of these cuts occurred in the United States, meaning roughly 60,000 of the 80,000 tech industry layoffs were U.S.-based. And when it comes to jobs being lost to AI for productivity gains and roles that are seemingly becoming redundant, it's believed that a large portion of these cuts could be driven by the expectation that AI will improve productivity, rather than by actual data reflecting this.

"I don't know if they are directly related to actual productivity gains," said Babak Hodjat, chief AI officer at Cognizant, of the job cuts. "Sometimes, you know, AI becomes the scapegoat from a financial perspective, like when a company hired too many, or they want to resize, and it gets blamed on AI." And with that, it would still take up to a year before companies start seeing notable productivity gains from AI.

Two new features aimed at helping users reach a wider audience and edit images are making their way onto Elon Musk's social media platform X. Both features are powered by xAI's Grok models and aim to improve the in-app experience while keeping users inside the app rather than relying on third-party tools.

Late Tuesday, X's head of product, Nikita Bier, announced that automatic translation is rolling out worldwide, allowing users to instantly read content in their preferred language without manually translating it. The feature is designed to give posts in any language a global reach and reduce reliance on Google Translate. Users who prefer the original text can tap the gear icon on a translated post and toggle off automatic translation for that particular language.

Alongside translation, X has also introduced a new in-app photo editor on iOS. The tool gives users access to drawing and text tools, plus a blur feature that lets users hide sensitive information, such as credit card or Social Security numbers, from images.

Anthropic's unreleased Claude Mythos is described as a "general-purpose, unreleased frontier model," but its coding capabilities are so powerful that it poses a real and unprecedented cybersecurity threat. According to the AI company, the Mythos Preview has already "found thousands of high-severity vulnerabilities" in every major operating system and web browser.

Its coding capabilities are so strong that, apparently, only the "most skilled humans" could beat it at finding and exploiting these vulnerabilities. "Given the rate of AI progress, it will not be long before such capabilities proliferate, potentially beyond actors who are committed to deploying them safely," Anthropic explains. "The fallout - for economies, public safety, and national security - could be severe."

And with that, Claude Mythos won't be getting a public release anytime soon, but it's currently being deployed as part of the new Project Glasswing initiative designed to "secure the world's most critical software."

Tesla's TeraFab project is a joint venture between Tesla, SpaceX, and xAI, dubbed by Elon Musk as "the most epic chip-building exercise in history." Tesla claims its $25 billion Austin chip factory would roughly match 70% of TSMC's current global output. This would allow Tesla to reduce its reliance on foundries like TSMC and Samsung by building domestic semiconductor production lines capable of producing chips at volumes significantly larger than those of any other foundry.

However, given that Tesla has no node IP or semiconductor industry experience, many were skeptical about how a car company, a rocket company, and an AI startup could develop 2nm-class process technology from scratch and scale it to 1 million. The most likely path forward was always a partnership with an existing foundry, and it appears Intel has now been chosen for that role.

Intel confirmed its involvement in the TeraFab project on its official X page, stating that it aims to help TeraFab reach 1 trillion watts of compute per year for future AI and robotics workloads.

Recently, several people noticed that Microsoft's Copilot Terms of Use page for its AI tool, which has become increasingly embedded in the company's Windows 11 operating system and Microsoft 365 Office suite, included several notable disclaimers.

"Copilot is for entertainment purposes only," one reads. Adding that "it can make mistakes," "may not work as intended," and that users should not "rely on Copilot for important advice." And to top it all off, there's a final "Use Copilot at your own risk." Naturally, this statement makes it sound like Copilot should not be used in any formal or business-facing capacity, which runs counter to the company's AI push to get its vast customer base to use Copilot for everything from search to summarizing documents, and even when firing up Paint or Notepad.

To make matters worse, the disclosure also states that Microsoft makes "no guarantees" that Copilot will operate as intended, which kind of makes it seem more experimental than practical. Again, this contradicts Microsoft's big Copilot marketing push across all its software offerings, as well as the arrival of Copilot+ PCs, which include dedicated NPUs to run local AI Copilot tools.

AMD has announced that its annual AMD Advancing AI 2026 event is scheduled for July 22 to 23 in San Francisco, California. The annual AI everything event is designed to bring together developers, researchers, partners, and other customers to "experience the latest in AI infrastructure, architecture, and development."

As the company's flagship AI event, similar in scope to NVIDIA GTC, we can expect a few announcements during the opening keynote with AMD CEO Dr. Lisa Su. During last year's event, AMD lifted the lid on its Instinct MI350 accelerators, teased its Helios rack, and even presented a look at EPYC CPUs powered by next-gen Zen 6 architecture.

With the AMD Instinct MI450 and Zen6-powered EPYC Venice CPUs on track for release this year, this will probably be the venue and time we get a deep dive and more details on these new technologies. Plus, introduce the next-gen Instinct MI500 accelerators, reportedly on track for 2027, and additional software updates.

Salesforce is bringing a massive overhaul of its popular enterprise communication platform, Slack. The headline is a serious makeover for Slackbot, the platform's AI-powered personal agent, transforming it from a simple chatbot into a full-fledged enterprise agent capable of drafting emails, scheduling meetings, and going through your inbox for specific information. This release puts Slack closer to what the company describes as an agentic operating system.

The most useful addition is "reusable AI skills", which let users define specific tasks for Slackbot from start to finish. Once created, these tasks can be deployed on demand across a variety of scenarios. There is also a strong focus on sharing and co-developing tools within an organization, meaning if you find a prompt useful, you can share it with your team. Slack comes with a built-in library of AI skills, but users can also create their own.

Slackbot also gets a Deep Research mode, and the chatbot will now conduct multi-step investigations rather than instantly responding. The new version also brings MCP (Model Context Protocol) client integration, allowing Slackbot to connect to and coordinate with external services and tools. This includes creating Google Slides, drafting Google Docs, and interacting with more than 2,600 apps in the Slack Marketplace.

Anthropic is one of the biggest AI companies on the planet, and a leak was detected on Tuesday morning that exposed the source code of Claude Code, a developer-focused capability that integrates Anthropic's AI assistant, Claude, into programming workflows.

According to the company, the leak was detected shortly after version 2.1.88 of Claude Code was made public, as the newly released version mistakenly included a source map file that exposed more than 500,000 lines of code and nearly 2,000 files. As you can probably imagine, internet sleuths were able to extract the files before Anthropic could deploy a fix, as a link to an archive containing them was posted to X by security researcher Chaofan Shou. The post caught the attention of more than 27 million users.

An Anthropic spokesperson confirmed the leak, saying the company is now reviewing steps to prevent a similar egregious human error from occurring again. The spokesperson provided a statement, saying, "Earlier today, a Claude Code release included some internal source code. No sensitive customer data or credentials were involved or exposed."

Meta has launched its newest smart glasses, developed in collaboration with Ray-Ban, aimed specifically at prescription wearers. Although people have already used Meta's current glasses with prescription lenses, the company has not actively targeted this market until now.

Meta claims its new glasses support nearly all prescriptions and are the most comfortable it has ever designed. CEO Mark Zuckerberg highlighted that billions of people wear glasses or contacts for vision correction, a market that Meta is now entering more deliberately.

The new glasses start at $499 and come in two styles: the Blayzer and the Scriber. The Blayzer adopts a rectangular design similar to the existing Wayfarer frames, while the Scriber offers a more rounded approach, close to the Headliner style. Both are part of the Gen 2 Ray-Ban Meta glasses lineup, but are significantly more customizable.

Meta is currently building a new $27 billion data center in Richland Parish, Louisiana, the company's biggest to date. Naturally, it's all about AI, with Meta noting that this 4 million-square-foot Richland Parish Data Center will play a "vital role in accelerating our AI progress, delivering over two gigawatts of compute capacity to train future open-source large language models."

With the energy requirements of these new data centers straining local grids and energy supplies, Meta has reportedly struck a deal with Entergy Louisiana to fund and help build seven new natural gas power plants in the region. With a combined power output of 5,200 megawatts, this is enough to cover "two gigawatts of compute" of the data center and then some.

As reported by the Wall Street Journal, the energy deal also includes 240 miles of 500-kilovolt transmission or power lines and battery storage. Now, although "natural gas" might sound good on paper, it's not exactly clean - it's still a fossil fuel. Albeit one with fewer emissions than coal-fired plants.

Google has developed three AI compression algorithms designed to reduce the memory footprint of large language models without sacrificing performance and quality. Published on Google Research, the tech is described as a way to shrink AI's working memory, known as the "KV cache", by using a form of vector quantization.

The company plans to present its findings at the ICLR 2026 conference next month, along with the three algorithms making this possible, namely TurboQuant, PolarQuant, and Quantized Johnson-Lindenstrauss.

TurboQuant would allow AI to remember more information while taking up less space and maintaining accuracy. There is a lot more detail in the Google Research article on how the compression technology works, but the results are what's exciting.