Meta Platforms CEO Mark Zuckerberg and his deputies want other companies to freely use and profit from new artificial intelligence software Meta is developing, a decision that could have big implications for other AI developers and businesses that are increasingly adopting it. The Information: Meta is working on ways to make the next version of its open-source large-language model -- technology that can power chatbots like ChatGPT -- available for commercial use, said a person with direct knowledge of the situation and a person who was briefed about it. The move could prompt a feeding frenzy among AI developers eager for alternatives to proprietary software sold by rivals Google and OpenAI. It would also indirectly benefit Meta's own AI development.

[...] Meta stands to gain from releasing open-source AI models. As developers adopt and improve those models or patch their security holes, Meta will be able to incorporate those improvements in AI models for its own consumer and advertising products, Zuckerberg said in an April call with stock analysts. For instance, Zuckerberg has said he wants small businesses and content creators that use Facebook's apps to have access to "AI agents" who can act on their behalf by automatically communicating with fans or customers. "LLaMA or the language model underlying this is basically going to be the engine that powers that," he said in an interview last week with podcaster Lex Fridman.

TechCrunch's Kyle Wiggers writes: Not to be outdone by Google, Meta has released its own AI-powered music generator -- and, unlike Google, open-sourced it. Called MusicGen, Meta's music-generating tool, a demo of which can be found here, can turn a text description (e.g. "An '80s driving pop song with heavy drums and synth pads in the background") into about 12 seconds of audio, give or take. MusicGen can optionally be "steered" with reference audio, like an existing song, in which case it'll try to follow both the description and melody.

Meta says that MusicGen was trained on 20,000 hours of music, including 10,000 "high-quality" licensed music tracks and 390,000 instrument-only tracks from ShutterStock and Pond5, a large stock media library. The company hasn't provided the code it used to train the model, but it has made available pre-trained models that anyone with the right hardware -- chiefly a GPU with around 16GB of memory -- can run.

So how does MusicGen perform? Well, I'd say -- though certainly not well enough to put human musicians out of a job. Its songs are reasonably melodic, at least for basic prompts like "ambient chiptunes music," and -- to my ears -- on par (if not slightly better) with the results from Google's AI music generator, MusicLM. But they won't win any awards.

Meta said on Tuesday that it would provide researchers with access to components of a new "human-like" artificial intelligence model that it said can analyze and complete unfinished images more accurately than existing models. From a report: The model, I-JEPA, uses background knowledge about the world to fill in missing pieces of images, rather than looking only at nearby pixels like other generative AI models, the company said. That approach incorporates the kind of human-like reasoning advocated by Meta's top AI scientist Yann LeCun and helps the technology to avoid errors that are common to AI-generated images, like hands with extra fingers, it said.

Meta, which owns Facebook and Instagram, is a prolific publisher of open-sourced AI research via its in-house research lab. Chief Executive Mark Zuckerberg has said that sharing models developed by Meta's researchers can help the company by spurring innovation, spotting safety gaps and lowering costs. "For us, it's way better if the industry standardizes on the basic tools that we're using and therefore we can benefit from the improvements that others make," he told investors in April.

A French start-up founded four weeks ago by a trio of former Meta and Google artificial intelligence researchers has raised $113.3 million in Europe's largest-ever seed round. From a report: Mistral AI's first round of financing values the Paris-based concern at $259 million, including the funds raised, according to people close to the company. The record amount raised highlights the growing frenzy surrounding AI and Europe's desire to create a viable alternative to Silicon Valley companies such as Microsoft-backed OpenAI and Google's DeepMind.

"There is a rising awareness of the fact that this technology is transformative and Europe needs to do something about it, both as a regulator, as a customer and an investor," said Arthur Mensch, Mistral's chief executive. The former DeepMind researcher founded the start-up with Timothee Lacroix and Guillaume Lample, who both recently left Meta after working at Facebook's parent company for the past few years. [...] Mistral has yet to develop its first product, and its first few employees started work only days ago. It plans to launch early next year a new "large language model," similar to the "generative AI" system that powers OpenAI's breakout ChatGPT app.

20 miles west of Portland, engineers at an Intel lab are dunking expensive racks of servers "in a clear bath" made of motor oil-like petrochemicals, reports the Oregonian, where the servers "give off a greenish glow as they silently labor away on ordinary computing tasks." Intel's submerged computers operate just as they would on a dry server rack because they're not bathing in water, even though it looks just like it. They're soaking in a synthetic oil that doesn't conduct electricity. So the computers don't short out.

They thrive, in fact, because the fluid absorbs the heat from the hardworking computers much better than air does. It's the same reason a hot pan cools off a lot more quickly if you soak it in water than if you leave it on the stove.

As data centers grow increasingly powerful, the computers are generating so much heat that cooling them uses exorbitant amounts of energy. The cooling systems can use as much electricity as the computers themselves. So Intel and other big tech companies are designing liquid cooling systems that could use far less electricity, hoping to lower data centers' energy costs by as much as a third — and reducing the facilities' climate impact. It's a wholesale change in thinking for data centers, which already account for 2% of all the electricity consumption in the U.S... Skeptics caution that it may be difficult or prohibitively expensive to overhaul existing data centers to adapt to liquid cooling. Advocates of the shift, including Intel, say a transition is imperative to accommodate data centers' growing thirst for power. "It's really starting to come to a head as we're hitting the energy crisis and the need for climate action globally," said Jen Huffstetler, Intel's chief product sustainability officer...

Cooler computers can be packed more tightly together in data centers, since they don't need space for airflow. Computer manufacturers can pack chips together more tightly on the motherboard, enabling more computing power in the same space. And liquid cooling could significantly reduce data centers' environmental and economic costs. Conventional data centers' evaporative cooling systems require tremendous volumes of water and huge amounts of electricity...

Many other tech companies are backing immersion cooling, too. Google, Facebook and Microsoft are all helping fund immersion cooling research at Oregon State... [T]he timing may finally be right for data centers operators to make the shift away from air cooling to something far more efficient. Intel's Huffstetler said she expects to see liquid cooling become widespread in the next three to five years.
The article notes other challenges:

• liquid adds more weight than some buildings' upper floors can support
• Some metals degrade faster in liquid than they do in air.
• And the engineers had to modify the servers by removing their fans — "because they serve no purpose while immersed."

During a companywide meeting, Meta's chief product officer, Chris Cox, revealed a preview of the company's upcoming Twitter competitor, a standalone app based on Instagram that will integrate with the decentralized social media protocol, ActivityPub. "That will theoretically allow users of the new app to take their accounts and followers with them to other apps that support ActivityPub, including Mastodon," reports The Verge. From the report: The forthcoming app, which, in the meeting today, Meta chief product officer Chris Cox called "our response to Twitter," will use Instagram's account system to automatically populate a user's information. The internal codename for the app is "Project 92," and its public name could be Threads, based on internal documents also seen by The Verge.

"We've been hearing from creators and public figures who are interested in having a platform that is sanely run, that they believe that they can trust and rely upon for distribution," Cox said, throwing direct shade at Elon Musk's handling of Twitter, to cheers from the audience. He said the company's goal for the app was "safety, ease of use, reliability" and making sure that creators have a "stable place to build and grow their audiences."

Cox said the company already has celebrities committed to using the app, including DJ Slime, and was in discussions with other big names, including Oprah and the Dalai Lama. He said "coding began" for the app in January and that Meta will be making the app available "as soon as we can."

Meta is planning to introduce AI-generated stickers on Messenger, allowing users to create stickers based on text prompts. The Verge reports: During a companywide meeting today that The Verge listened to, Ahmad Al-Dahle, Meta's vice president of AI, told employees that the company will leverage its image generation model to let users create stickers based on text prompts. Employees will begin testing the feature internally before it's made available to the public. "With AI-generated stickers, our users can have infinitely more options for self-expression, cultural representations, and even trend relevance," Al-Dahle says. "Of course, stickers are just the tip of the iceberg."

Al-Dahle adds that the company is also "working on AI models that are going to transform any image you want in any way you want." That includes doing things like "changing the aspect ratio of your picture" or turning a picture of a corgi "into a painting."

An anonymous reader quotes a report from The Verge: Mark Zuckerberg doesn't seem fazed by Apple's introduction of the Vision Pro. In a companywide meeting with Meta employees today that The Verge watched, the CEO said Apple's device didn't present any major breakthroughs in technology that Meta hadn't "already explored" and that its vision for how people will use the device is "not the one that I want." He also pointed to the fact that Meta's upcoming Quest 3 headset will be much cheaper, at $499 compared to the Vision Pro's $3,499 price tag, giving Meta the opening to reach a wider user base.

"I think that their announcement really showcases the difference in the values and the vision that our companies bring to this in a way that I think is really important," Zuckerberg told employees, who were gathered at the company's Menlo Park, California, headquarters for its first all-hands meeting since 2020. Zuckerberg said that the Quest is about "people interacting in new ways and feeling closer" while also "about being active and doing things." "By contrast, every demo that they showed was a person sitting on a couch by themself," he said of Apple's WWDC keynote earlier this week. "I mean, that could be the vision of the future of computing, but like, it's not the one that I want."

Meta CEO Mark Zuckerberg announced a plan to employees on Thursday that will see it put generative AI text, image and video generators into its flagship products, such as Facebook and Instagram. From a report: At an all-hands meeting with workers on Thursday, Zuckerberg announced a range of technologies at various stages of development, with some for internal use but many designed directly for consumers. One, for example, will allow customers to use a text prompt to modify their own photos and share them in Instagram Stories.

Another will bring AI agents with different personalities and capabilities to help or entertain. That's focused initially for use in Messenger and WhatsApp. The company is also hosting an internal hackathon in July focused on generative AI.

According to a joint investigation from The Wall Street Journal and researchers at Stanford University and the University of Massachusetts Amherst, Instagram's algorithms are actively promoting networks of pedophiles who commission and sell child sexual abuse content on the app. The Verge reports: Accounts found by the researchers are advertised using blatant and explicit hashtags like #pedowhore, #preteensex, and #pedobait. They offer "menus" of content for users to buy or commission, including videos and imagery of self-harm and bestiality. When researchers set up a test account and viewed content shared by these networks, they were immediately recommended more accounts to follow. As the WSJ reports: "Following just a handful of these recommendations was enough to flood a test account with content that sexualizes children."

In addition to problems with Instagram's recommendation algorithms, the investigation also found that the site's moderation practices frequently ignored or rejected reports of child abuse material. The WSJ recounts incidents where users reported posts and accounts containing suspect content (including one account that advertised underage abuse material with the caption "this teen is ready for you pervs") only for the content to be cleared by Instagram's review team or told in an automated message [...]. The report also looked at other platforms but found them less amenable to growing such networks. According to the WSJ, the Stanford investigators found "128 accounts offering to sell child-sex-abuse material on Twitter, less than a third the number they found on Instagram" despite Twitter having far fewer users, and that such content "does not appear to proliferate" on TikTok. The report noted that Snapchat did not actively promote such networks as it's mainly used for direct messaging.

In response to the report, Meta said it was setting up an internal task force to address the issues raised by the investigation. "Child exploitation is a horrific crime," the company said. "We're continuously investigating ways to actively defend against this behavior." Meta noted that in January alone it took down 490,000 accounts that violated its child safety policies and over the last two years has removed 27 pedophile networks. The company, which also owns Facebook and WhatsApp, said it's also blocked thousands of hashtags associated with the sexualization of children and restricted these terms from user searches.

schwit1 shares a report from CBS News: When whistleblower Frances Haugen pulled back the curtain on Facebook in the fall of 2021, thousands of pages of internal documents showed troubling signs that the social media giant knew its platforms could be negatively impacting youth, and were doing little to effectively change it. With around 21 million American adolescents on social media, parents took note. Now, families are suing social media. Since we first reported this story last December, the number of families pursuing lawsuits has grown to over 2,000. More than 350 lawsuits are expected to move forward this year against TikTok, Snapchat, YouTube, Roblox and Meta -- the parent company to Instagram and Facebook.

Kathleen Spence: They're holding our children hostage and they're seeking and preying on them. Sharyn Alfonsi: Preying on them? Kathleen Spence: Yes. The Spence family is suing social media giant Meta. Kathleen and Jeff Spence say Instagram led their daughter Alexis into depression and to an eating disorder at the age of 12. [...] Attorney Matt Bergman represents the Spence family. He started the Social Media Victims Law Center after reading the Facebook papers and is now working with more than 1,800 families who are pursuing lawsuits against social media companies like Meta. Matt Bergman: Time and time again, when they have an opportunity to choose between safety of our kids and profits, they always choose profits.

This summer, Bergman and his team plan on starting the discovery process for the federal case against Meta and other social media companies, a multi-million dollar suit that he says is more about changing policy than financial compensation. This summer, Bergman and his team plan on starting the discovery process for the federal case against Meta and other social media companies, a multi-million dollar suit that he says is more about changing policy than financial compensation. Matt Bergman: They have intentionally designed a product that is addictive. They understand that if children stay online, they make more money. It doesn't matter how harmful the material is.

Security guru Bruce Schneier and CS professor Jim Waldo think big tech has underestimated the impact of open source principles on AI research: In February, Meta released its large language model: LLaMA. Unlike OpenAI and its ChatGPT, Meta didn't just give the world a chat window to play with. Instead, it released the code into the open-source community, and shortly thereafter the model itself was leaked. Researchers and programmers immediately started modifying it, improving it, and getting it to do things no one else anticipated. And their results have been immediate, innovative, and an indication of how the future of this technology is going to play out. Training speeds have hugely increased, and the size of the models themselves has shrunk to the point that you can create and run them on a laptop. The world of A.I. research has dramatically changed.

This development hasn't made the same splash as other corporate announcements, but its effects will be much greater. It will wrest power from the large tech corporations, resulting in both much more innovation and a much more challenging regulatory landscape. The large corporations that had controlled these models warn that this free-for-all will lead to potentially dangerous developments, and problematic uses of the open technology have already been documented. But those who are working on the open models counter that a more democratic research environment is better than having this powerful technology controlled by a small number of corporations...

[B]uilding on public models like Meta's LLaMa, the open-source community has innovated in ways that allow results nearly as good as the huge models — but run on home machines with common data sets. What was once the reserve of the resource-rich has become a playground for anyone with curiosity, coding skills, and a good laptop.

Bigger may be better, but the open-source community is showing that smaller is often good enough. This opens the door to more efficient, accessible, and resource-friendly LLMs.
Low-cost customization will foster rapid innovation, the article argues, and "takes control away from large companies like Google and OpenAI." Although this may have one unforeseen consequence...

"Now that the open-source community is remixing LLMs, it's no longer possible to regulate the technology by dictating what research and development can be done; there are simply too many researchers doing too many different things in too many different countries."

Thanks to long-time Slashdot reader mrflash818 for submitting the article

New submitter Peppercopia writes: CTV News is reporting that Meta will begin testing the blocking of news sites in Canada. If the argument is that the social media giants are unfairly benefitting from content from Canadian news organizations, this move should be moot as the 'stealing' would now be stopping. Unfortunately the opposite is likely the case, and the news organizations will find out how important the free traffic and promotion they are getting from social media giants really is. It feels a bit like killing the golden goose to get the eggs. The move is designed to "work out the kinks" before permanently blocking news on its platforms when the Canadian government passes the Online News Act. According to CTV News, the test "will affect up to five percent of its 24 million Canadian users."

"The company says the randomly selected users won't be able to see some content including news links as well as reels, which are short-form videos, and stories, which are photos and videos that disappear after 24 hours." Media organizations will be chosen at random.

Meta announced that it would remove news content from its platform in California if the state government passes legislation requiring tech companies to pay publishers. Reuters reports: The proposed California Journalism Preservation Act would require "online platforms" to pay a "journalism usage fee" to news providers whose work appears on their services, aimed at reversing a decline in the local news sector. In a tweeted statement, Meta spokesman Andy Stone called the payment structure a "slush fund" and said the bill would primarily benefit "big, out-of-state media companies under the guise of aiding California publishers."

The statement was Meta's first on the California bill specifically, although the company has been waging similar battles over compensation for news publishers at the federal level and in countries outside the United States.

Britain's competition watchdog on Friday said social media giant Meta had offered to limit its use of other businesses' advertising data for its Facebook Marketplace service to address the regulator's competition concerns. From a report: The Competition and Markets Authority (CMA) said it was minded to accept the commitments, which include advertisers being able to opt out of allowing their data to be used to improve the Facebook Marketplace classified ads platform. CMA executive director of enforcement Michael Grenfell said: "Reducing the risk of Meta unfairly exploiting the data of businesses who advertise on its platform for its own competitive advantage could help many UK businesses who advertise there. We are now consulting on these commitments which we believe, at this stage, will address our concerns."

Meta employees received news Wednesday of the final round of previously announced job cuts, which affected thousands of workers in the company's business departments. Now, remaining staff are hoping an uncomfortable limbo at the company can end. From a report: The layoffs complete the bulk of the restructuring Chief Executive Officer Mark Zuckerberg announced in March to eliminate 10,000 positions. Initial reductions affected the company's recruiting and human resources departments, and in late April jobs in Meta's tech groups were slashed. Zuckerberg has said further cuts will come in only a "small number of cases" for the rest of the year, giving those people left a cold sense of relief.

The company, which owns Facebook, Instagram and WhatsApp, said the layoffs were necessary to improve efficiency, after over-hiring during the pandemic. Meta promised faster product development and decision-making that sent its shares up more than 100% so far this year. But employees said some important work and planning has been at a standstill. Notably, Meta is still deciding on its product roadmap for the rest of the year, while it sorts out resources following cuts in the tech group, a person familiar with the matter said. During the limbo, employees have been unsure who to collaborate with, how to shift responsibilities on their teams or who would be cut next, according to current and recently let-go employees, who asked not to be named discussing internal issues.

Facebook's parent company Meta is reportedly in discussions with augmented reality start-up Magic Leap to establish a multiyear agreement for intellectual property licensing and contract manufacturing in North America. While the partnership is not expected to result in a joint headset, Magic Leap's technology could play a crucial role in Meta's ambitious metaverse project as it seeks to compete with Apple's upcoming mixed reality device. The Financial Times reports: Magic Leap produces custom components, including high-tech lenses and associated software, which are key technologies that may be required to build a metaverse. Two former employees said Magic Leap's "biggest asset" is the sophistication of its "waveguides" -- technology that allows thin glass in front of the user's eyes to conjure up realistic images at different depths.

Meta sells nearly 80 per cent of all VR/AR headsets, thanks to its VR Quest models. But the market itself is small -- fewer than 9mn units sold last year, according to IDC -- a tenuous lead given Apple's expected entry into the market during its developer conference next month. The company told the Financial Times: "Given the complexities of developing true augmented reality technologies and the intricacies involved with manufacturing these optics, as well as the issues many companies experience with overseas supply chain dependencies, we have entered into several non-exclusive IP licensing and manufacturing partnerships with companies looking to enter the AR market or expand their current position."

Facebook owner Meta was hit by a record $1.3 billion European Union privacy fine and given a deadline to stop shipping users' data to the US after regulators said it failed to protect personal information from the prying eyes of American security services. Bloomberg News: The social network giant's continued data transfers to the US didn't address "the risks to the fundamental rights and freedoms" of people whose data was being transfered across the Atlantic, according to a decision by the Irish Data Protection Commission announced on Monday. On top of the fine, which eclipses a $806 million EU privacy penalty previously doled out to Amazon, Meta was given five months to "suspend any future transfer of personal data to the US" and six months to stop "the unlawful processing, including storage, in the US" of transferred personal EU data. A data-transfers ban for Meta was widely expected and once prompted the US firm to threaten a total withdrawal from the EU. But its impact has now been muted by the transition phase given in the decision and the prospect of a new EU-US data flows agreement that could already be operational by the middle of this year.

An anonymous reader shared this report from the Verge: Meta is building its first custom chip specifically for running AI models, the company announced on Thursday. As Meta increases its AI efforts — CEO Mark Zuckerberg recently said the company sees "an opportunity to introduce AI agents to billions of people in ways that will be useful and meaningful" — the chip and other infrastructure plans revealed Thursday could be critical tools for Meta to compete with other tech giants also investing significant resources into AI.

Meta's new MTIA chip, which stands for Meta Training and Inference Accelerator, is its "in-house, custom accelerator chip family targeting inference workloads," Meta VP and head of infrastructure Santosh Janardhan wrote in a blog post... But the MTIA chip is seemingly a long ways away: it's not set to come out until 2025, TechCrunch reports.
Meta has been working on "a massive project to upgrade its AI infrastructure in the past year," Reuters reports, "after executives realized it lacked the hardware and software to support demand from product teams building AI-powered features."

As a result, the company scrapped plans for a large-scale rollout of an in-house inference chip and started work on a more ambitious chip capable of performing training and inference, Reuters reported...

Meta said it has an AI-powered system to help its engineers create computer code, similar to tools offered by Microsoft, Amazon and Alphabet.
TechCrunch calls these announcements "an attempt at a projection of strength from Meta, which historically has been slow to adopt AI-friendly hardware systems — hobbling its ability to keep pace with rivals such as Google and Microsoft."

Meta's VP of Infrastructure told TechCrunch "This level of vertical integration is needed to push the boundaries of AI research at scale." Over the past decade or so, Meta has spent billions of dollars recruiting top data scientists and building new kinds of AI, including AI that now powers the discovery engines, moderation filters and ad recommenders found throughout its apps and services. But the company has struggled to turn many of its more ambitious AI research innovations into products, particularly on the generative AI front. Until 2022, Meta largely ran its AI workloads using a combination of CPUs — which tend to be less efficient for those sorts of tasks than GPUs — and a custom chip designed for accelerating AI algorithms...

The MTIA is an ASIC, a kind of chip that combines different circuits on one board, allowing it to be programmed to carry out one or many tasks in parallel... Custom AI chips are increasingly the name of the game among the Big Tech players. Google created a processor, the TPU (short for "tensor processing unit"), to train large generative AI systems like PaLM-2 and Imagen. Amazon offers proprietary chips to AWS customers both for training (Trainium) and inferencing (Inferentia). And Microsoft, reportedly, is working with AMD to develop an in-house AI chip called Athena.

Meta says that it created the first generation of the MTIA — MTIA v1 — in 2020, built on a 7-nanometer process. It can scale beyond its internal 128 MB of memory to up to 128 GB, and in a Meta-designed benchmark test — which, of course, has to be taken with a grain of salt — Meta claims that the MTIA handled "low-complexity" and "medium-complexity" AI models more efficiently than a GPU. Work remains to be done in the memory and networking areas of the chip, Meta says, which present bottlenecks as the size of AI models grow, requiring workloads to be split up across several chips. (Not coincidentally, Meta recently acquired an Oslo-based team building AI networking tech at British chip unicorn Graphcore.) And for now, the MTIA's focus is strictly on inference — not training — for "recommendation workloads" across Meta's app family...

If there's a common thread in today's hardware announcements, it's that Meta's attempting desperately to pick up the pace where it concerns AI, specifically generative AI... In part, Meta's feeling increasing pressure from investors concerned that the company's not moving fast enough to capture the (potentially large) market for generative AI. It has no answer — yet — to chatbots like Bard, Bing Chat or ChatGPT. Nor has it made much progress on image generation, another key segment that's seen explosive growth.

If the predictions are right, the total addressable market for generative AI software could be $150 billion. Goldman Sachs predicts that it'll raise GDP by 7%. Even a small slice of that could erase the billions Meta's lost in investments in "metaverse" technologies like augmented reality headsets, meetings software and VR playgrounds like Horizon Worlds.

The Supreme Court on Thursday sidestepped a case against Google that might have allowed more lawsuits against social media companies. From a report: The justices' decision returns to a lower court the case of a family of an American college student who was killed in an Islamic State terrorist attack in Paris. The family wants to sue Google for YouTube videos they said helped attract IS recruits and radicalize them. Google claims immunity from the lawsuit under a 1996 law that generally shields social media company for content posted by others. Lower courts agreed with Google. The justices had agreed to consider whether the legal shield is too broad. But in arguments in February, several sounded reluctant to weigh in now. In an unsigned opinion Thursday, the court wrote that it was declining to address the law at issue.