Tag:

scale

Meta Invests $14 Billion in Scale AI to ‘Deepen’ its Work on Superintelligence

by admin June 13, 2025

In brief

Meta has acquired a 49% stake in Scale AI for $14.3B, its second-largest deal ever.
Scale CEO Alexandr Wang has joined Meta to lead a new superintelligence lab.
The investment illustrates Meta’s aggressive strategy to compete with the likes of OpenAI and Google.

American multinational tech company Meta has made a significant investment in Scale AI, the data-labeling startup essential for training artificial intelligence systems, recruiting its young founder Alexandr Wang to build out a “superintelligence” lab focused on achieving artificial general intelligence (AGI).

The $14.3 billion investment acquired a 49% stake in Scale AI, raising its valuation to over $29 billion. This is Meta’s second-largest investment, following its $19 billion WhatsApp acquisition in 2014.

Scale AI announced that Wang would continue to serve as a director on the company’s Board of Directors, with the firm’s Chief Strategy Officer, Jason Droege, stepping into his shoes as interim CEO.

In a tweet, Wang shared a note to employees, describing Meta’s investment as a “major milestone and a powerful validation of the hard work you’ve all put into Scale’s mission.”

For Meta, which has been working on its own AI strategies and recently released a new model and a standalone AI app in the past couple of months, the deal would “deepen the work” it has been engaged in for “producing data for A.I. models,” according to company statements shared with media.

Meta’s AI race

This speaks to the urgency of Meta’s move to catch up in the AI race, as it faces mounting pressure from other players working on “frontier models” which are advanced, large-scale systems pushing the boundaries of AI.

Yet the move also highlights a key tension between large tech firms such as Meta and others building on decentralized platforms, according to Renz Chong, CEO of a16z-backed modular on-chain platform Sovrun.

“Earlier this year, we saw the rise of open-source frontier models that can go toe-to-toe with closed models from Big Tech,” Chong told Decrypt. “That’s a clear signal: ‘state-of-the-art’ no longer has to mean centralized or proprietary.”

Because most other AI-on-chain projects still lean on “centralized inference endpoints or off-chain APIs,” the predicament places pressure on decentralized players.

“Early infrastructure players are laying critical groundwork, offering decentralized compute and incentivized training layers,” Chong noted.

What is Scale AI?

Scale AI specializes in data labeling services that are essential for training AI systems, working with clients including Google and OpenAI. The startup employs human annotators to classify data that fuels AI models, with much of the work conducted outside the U.S. through labor-intensive processes.

Meta’s investment would reportedly give it a minority stake in Scale AI, allowing the startup to maintain operational independence—a structure that could help Meta avoid additional regulatory scrutiny amid ongoing antitrust battles. Earlier in April, Meta was accused of assisting China’s AI ambitions, prompting a Senate inquiry.

Still, the consolidation of AI resources between tech giants has prompted alternative visions from the decentralized sector. Sovrun, for instance, recently entered a joint venture with Virtuals Protocol to build ReadyGamer, a platform that integrates AI-driven NPCs into popular game worlds.

While those projects experienced a decline in revenue earlier this year, a resurgence is underway, with daily numbers slowly returning to previous levels, according to data maintained by Virtuals Protocol.

Chong argues the real shift may be found not just in making better systems, but in “changing who gets to shape them” and building “outcomes that matter to the communities they serve.”

Generally Intelligent Newsletter

A weekly AI journey narrated by Gen, a generative AI model.

Source link

June 13, 2025 0 comments

Gaming Gear

Meta invests $14 billion in Scale to catch up in the AI race

by admin June 13, 2025

Meta is paying $14.3 billion to acquire 49 percent of Scale AI and hire its CEO, Alexandr Wang, to revamp its troubled AI efforts.

As part of the deal, Wang will report directly to Meta CEO Mark Zuckerberg and remain on Scale’s board of directors, both companies announced on Thursday. Sources say Wang will lead a new AI lab at Meta tasked with building “superintelligence.” Meta spokesperson Ashley Zandy says the company “will share more about this effort and the great people joining this team in the coming weeks.”

Zuckerberg has been actively recruiting a new team of researchers from rival firms to join Wang’s team, according to people familiar with the matter and other press reports. The Meta founder has reached out to potential recruits directly — usually via a cold email or WhatsApp message — and lured some of them away from companies like Google with seven- and eight-figure compensation packages.

Scale works with Google, OpenAI, Anthropic, and others to help train their models by having humans annotate and label the data that feeds them. Most of that work is done through low-cost labor outside of the US, and it has become a critical component of AI development. With Wang joining Meta, Scale is naming Jason Droege, its former chief strategy officer, as its new interim CEO.

“We’ve grown to over 1,500 people and become the trusted partner for model builders, enterprises, and governments building and deploying the smartest Al tools and applications,” Wang, who at age 21 became known as the world’s youngest self-made billionaire because of his stake in the company, said in a memo to Scale employees.

“Scale is now one of the most impactful companies in the world, accelerating the development of what may be the most important technology in human history. Today, we are announcing a massive new investment from Meta. This is a major milestone and a powerful validation of the hard work you’ve all put into Scale’s mission.”

Wang added that the “proceeds from Meta’s investment will be distributed to those of you who are shareholders and vested equity holders, while maintaining the opportunity to continue participating in our future growth as ongoing equity holders.”

By not fully acquiring Scale, Meta is following Big Tech’s established playbook of hiring key AI talent while also investing in their startups. This process is designed to reduce the risk of antitrust scrutiny; however, the deal with Scale is likely to still be scrutinized by governments, even though Meta will receive only non-voting shares in the company.

This year, Scale signed a deal with the Department of Defense for a first-of-its-kind AI agent program for US military planning. Since then, it has expanded its business with other governments, recently inking a five-year deal to provide AI tools to Qatar. The company’s work with governments in Asia and Europe could result in a sizable portion of its overall sales over the next two quarters.

Meta, meanwhile, just went to trial to defend itself from being broken up by the US government. The outcome of that case is still uncertain.

Since the disappointing debut of Llama 4 earlier this year, Meta CEO Mark Zuckerberg has been eager to catch up with competitors like Google, OpenAI, Anthropic, and DeepSeek. Llama 4’s release was delayed multiple times, and then Meta was caught gaming a public leaderboard to make the model appear better than it actually was. The company has still not released Llama 4 Behemoth, its largest and most expensive version that was teased in April.

Last month, Zuckerberg said that two of Meta’s top priorities for 2025 are making its ChatGPT rival, Meta AI, “the leading personal AI” and “building full general intelligence.” He recently stated that Meta AI has reached one billion monthly users, although the prominence of the assistant across Instagram, WhatsApp, and Facebook heavily influences this number.

In April, Meta released a standalone app for Meta AI, featuring a social feed that showcases how people are using it. The app briefly reached the top of the App Store but hasn’t sustained that level of popularity.

Source link

June 13, 2025 0 comments

Crypto Trends

Meta is assembling a ‘superintelligence’ AI team with $15b investment in Scale AI

by admin June 11, 2025

Meta hopes to bridge its AI gap with other major tech firms, with CEO Mark Zuckerberg personally assembling an AI “superintelligence group.”

Meta is ramping up efforts to control the future of artificial intelligence. According to a June 10 Bloomberg report, CEO Mark Zuckerberg is personally assembling a team of experts for a new division focused on AI research. At the same time, the company has made a $15 billion investment in Scale AI and recruited its CEO, Alexandr Wang.

Meta’s “superintelligence group” will focus on outpacing companies like OpenAI, Google, and Microsoft in the race toward artificial general intelligence. The goal is to develop machines capable of performing as well as humans on valuable tasks.

Once Meta develops its models, the company plans to integrate them across its product ecosystem, including Facebook, Instagram, and WhatsApp. This also includes its Meta chatbot and its Ray-Ban smart glasses. Zuckerberg reportedly plans to hire 50 people for the new team.

Meta invests nearly $15n in AI startup

A part of its efforts to assemble the superintelligence group is a major investment in an AI startup. Meta made a $14.8 billion investment in Scale AI, according to a June 10 report by The Information. The investment will secure it a 49% stake in the company, as well as recruit its CEO CEO, Alexandr Wang for the new team.

According to Bloomberg, insiders at Meta expressed frustration that the company appears to be trailing its rivals in the AI race. In particular, they cited underwhelming performance from Meta’s Llama 4 large language model, which they believe falls short of competitors from Google and OpenAI.

The AI competition among tech giants is intensifying. Google recently launched a new model that is being integrated into its search product. Meanwhile, Microsoft and Apple have signed partnerships with OpenAI to incorporate its ChatGPT model into their services.

Source link

June 11, 2025 0 comments

Esports

Non-profit Games for Change appoints 5 new board members to “scale impact in meaningful new ways”

by admin June 11, 2025

Games for Change, a non-profit organization “empowering game creators and social innovators to drive impact through digital games and immersive tech,” has expanded its board and appointed five more members: Samir El Agili, (Tilting Point and Brandible), Benjamin Golant (Tencent Americas), (Xbox, Microsoft), Evan Segal (Segal Ventures), and Mark Stanley (Yelnats Collective).

They join existing board members to guide the expansion and growth of G4C’s mission and its latest initiatives, including its XR for Change, G4C Learn, and Raising Good Gamers programs. Previous projects have included the Games for Change Diverse Voices, New Stories Challenge, which ran in 2023.

“As communities and organizations worldwide experience change at an unprecedented rate, Games for Change is needed more than ever,” said president Susanna Pollack. “We stand at the intersection of creativity and purpose, clear-eyed about the unique co-creation and storytelling power of games to foster empathy and inspire the next generation of changemakers.

“Individually, each of our Board members has worked to empower others through the power of play, not just as entertainment, but for education, entrepreneurship, and global progress. Collectively, the experience of our Board of Directors across marketing, policy, and philanthropy strengthens our ability to forge innovative partnerships and scale our impact in meaningful new ways.”

Source link

June 11, 2025 0 comments

Crypto Trends

Ethereum to Scale Layer 1 by 10X in a Year, Says Buterin

by admin June 2, 2025

Vitalik Buterin, one of Ethereum’s co-founders, has announced that Ethereum’s layer 1 (L1) will be scaled by 10X in the next year. Buterin stated at ETHGlobal Prague that the network wants to increase its capacity while maintaining decentralization.

Buterin said it will take about a year and a half to complete this scaling roadmap, so the team can make careful changes to the network. He underlined the need to avoid making decisions too quickly that could hurt Ethereum’s main qualities, like being able to resist censorship and maintain user privacy.

The plan covers boosting stateless nodes and making updates to the Ethereum Virtual Machine (EVM). This is different from what Ethereum researcher Dankrad Feist suggests, who believes Ethereum should scale its L1 by 100 times within the next five years.

Buterin wants to grow the network slowly to keep it safe, secure and controlled by many people. Although ETH prices have dropped and on-chain activity was low in Q1 2025, more people are interested in scaling Ethereum.

Organizations such as SharpLink Gaming are buying billions of Ethereum by using BTSC and ETFs which indicates that companies are feeling more confident. With the major upgrade approaching, ETH is expected to rise to $3,000 in order to recover some of its earlier losses.

Buterin’s balanced approach to scaling may help Ethereum achieve its next phase of development by blending speed, stability and decentralization.

Also Read: Ethereum Price Prediction June 2025: Will ETH Reclaim $3000?

Source link

June 2, 2025 0 comments

Game Updates

People Can Fly cancels two projects and confirms it will once again have to “scale down” its teams

by admin June 1, 2025

People Can Fly has cancelled two projects and “scaled down” its teams.

In a statement posted to LinkedIn this afternoon, CEO Sebastian Wojciechowski said it had been a “difficult decision” but attributed the suspension of Project Gemini and Bifrost to the publisher “not present[ing] us with a draft of the subsequent content rider to the Publishing Agreement”, and the Group’s cash flow, “which showed a lack of prospects”, respectively.

“Today we made a very difficult decision to suspend the development of project Gemini and project Bifrost – the relevant current reports have been released to the market,” Wojciechowski wrote.

“The suspension of the Gemini project is a consequence of the fact that the Publisher has not presented us with a draft of the subsequent content rider to the Publishing Agreement covering the terms and conditions of further milestone on project Gemini and the lack of communication from the Publisher as to its willingness to continue or terminate the Gemini project.

“Project Bifrost was suspended due to the above and the analysis of the Group’s cash flow, which showed a lack of prospects for securing organisational resources and funds necessary to continue the production and release of this project.

“As a result, we have to significantly regroup as a studio and scale down our teams, which hurts the most,” Wojciechowski added. “We wish to express our deepest regret and sadness over how these events have unfolded and our sincere gratitude for everyone’s contribution up to this point.”

Wojciechowski did not specify which publisher he thought had failed to “communicate”, but the studio has worked with several publishing partners, most recently Square Enix, Krafton, and Sony. It’s also working on Gears of War: E-Day.

In December, People Can Fly the Outriders developer announced its second round of layoffs in less than a year, this time affecting “more than 120” people. The move accompanies the cancellation or downsizing of several projects, with the studio blaming “external market pressures”.

Source link

June 1, 2025 0 comments

Product Reviews

Staples Union & Scale Electric Standing Desk Review: Micro Movements

by admin May 31, 2025

The desk height ranges from 25.5 to 51.1 inches, and the dual motors glide effortlessly up and down, and they’re not too loud. You’ll want to be careful when lowering it—there’s collision detection, and it successfully stopped the desk from crushing a stool after I put it underneath, but it seems to depend on where the object under the desk is. I rolled my seat in so that the arms of my chair were directly under the desktop, and when I pressed the lower height preset, the desk kept going and would have probably broken my chair arms (or toppled forward) had I not forced it to stop. Maybe had I not stopped it, the sensor would have paused the operation automatically, but it was a little too close to my liking.

I haven’t noticed much wobbling when the desk goes up and down, though if you put pressure on the front of the desk (maybe you’re leaning forward with your elbows on the surface), there’s a slight forward shift. This didn’t cause any problems.

Staples offers a seven-year warranty for the motor and legs, along with a few other parts, but the control box, cables, control panel, and desktop are only covered for five years. That’s about the same as what you’ll find on a Flexispot, though short of desks like the Branch Duo that don’t cost that much more. The nice thing about buying from Staples is that returns are always free, and you can drop it off at a nearby retail store (though you’ll still have to disassemble it).

Photograph: Julian Chokkattu

Small Shift

The middle button on the control panel turns on Micro Movement mode, and at first, I thought it didn’t work, but that was because I didn’t realize how micro the movements are. Once turned on, the desk will move up 2.36 inches within five minutes, stop for two minutes, then move down the same amount in another five-minute increment. This will repeat after 48 minutes and cycle four times throughout four hours.

Staples says this will help reduce fatigue and encourage movement—subtle shifts in posture and muscle engagement—but the only reason I started noticing it was because my arms felt pinched. I set the desk up perfectly to my height and seat—my elbows sat level with the desktop without touching the surface. But for a few minutes with this feature turned on, my elbows suddenly were resting on the edge of the table, and it wasn’t pleasant. It works a little better when you’re standing, though you’ll still feel your arms resting on the table as the desk creeps upward.

I didn’t find the Micro Movements feature helpful. I didn’t feel like it encouraged movement, and I’m still fatigued from, well, everything. The best part of a standing desk is that it enables you to stand up, which will hopefully then induce some actual movement throughout your workday. The Micro Movements feel akin to the active seating hype, which has never felt particularly effective.

Thankfully, it’s an optional feature, and you can ignore it. And you probably should if budget is on your mind, because Staples has other dual-motor 55-inch desks without this feature, but they’re not cheaper. You can size down to the 48-inch standing desk, but it only has a single motor.

This Union & Scale is a solid desk for the money, with easy assembly, simple configuration, and a dual-motor system that operates smoothly. That said, Flexispot’s desks have more desktop customization options, sizes, colors, and accessories you can add on, and they cost roughly the same, if not less. I’ve had more trouble with Flexispot’s assembly process in the past, though, if that’s worth something.

Source link

May 31, 2025 0 comments

Product Reviews

Nvidia to boost AI server racks to megawatt scale, increasing power delivery by five times or more

by admin May 28, 2025

Nvidia is developing a new power infrastructure called the 800V HVDC architecture to deliver the power requirements of 1 MW server racks and more, with plans to deploy it by 2027. According to Nvidia, the current 54V DC power distribution system is already reaching its limit as racks begin to exceed 200 kilowatts. As AI chips become more powerful and demand more electricity, these existing systems would no longer be able to practically keep up, requiring data centers to build new solutions so that their electrical circuits do not get overwhelmed.

For example, Nvidia says that its GB200 NVL72 or GB300 NVL72 needs around eight power shelves. If it used 54V DC power distribution, the power shelves would consume 64 U of rack space, which is more than what the average server rack can accommodate. Aside from this, it also said that delivering 1 MW using 54V DC requires a 200 kg copper busbar — that means a gigawatt AI data center, which many companies are now racing to build, would need 500,000 tons of copper. This is nearly half of the U.S.’s total copper output in 2024, and that’s just for one site.

So, instead of using the 54V DC system, which is installed directly at the server cabinet, Nvidia is proposing to use the 800V HVDC, which will connect near the site’s 13.8kV AC power source. Aside from freeing up space in the server racks, this will also streamline the approach and make power transmission within the data center more efficient. It will also remove the multiple AC to DC and DC to DC transformations used in the current system, which added complexity.