Meta's AI research and development

[firestar464]

This wouldn't be so dumb if it weren't for Facebook's AI spam problem, which Zuck seems to be ignoring.

[Powers]

This wouldn't be so dumb if it weren't for Facebook's AI spam problem, which Zuck seems to be ignoring.

Meta's curated images of Jesus.

[wjfox]

Meta proposes new scalable memory layers that improve knowledge, reduce hallucinations

January 7, 2025 1:45 PM

As enterprises continue to adopt large language models (LLMs) in various applications, one of the key challenges they face is improving the factual knowledge of models and reducing hallucinations. In a new paper, researchers at Meta AI propose “scalable memory layers,” which could be one of several possible solutions to this problem.

Scalable memory layers add more parameters to LLMs to increase their learning capacity without requiring additional compute resources. The architecture is useful for applications where you can spare extra memory for factual knowledge but also want the inference speed of nimbler models.

Traditional language models use “dense layers” to encode vast amounts of information in their parameters. In dense layers, all parameters are used at their full capacity and are mostly activated at the same time during inference. Dense layers can learn more complex functions as they grow larger, but increasing their size requires additional computational and energy resources.

In contrast, for simple factual knowledge, much simpler layers with associative memory architectures resembling lookup tables would be more efficient and interpretable. This is what memory layers do. They use simple sparse activations and key-value lookup mechanisms to encode and retrieve knowledge. Sparse layers take up more memory than dense layers but only use a small portion of the parameters at once, which makes them much more compute-efficient.

Memory layers have existed for several years but are rarely used in modern deep learning architectures. They are not optimized for current hardware accelerators.

https://venturebeat.com/ai/meta-propose ... cinations/

[wjfox]

[firestar464]

Meta has an AI for brain typing, but it’s stuck in the lab

https://www.technologyreview.com/2025/0 ... n-the-lab/

[wjfox]

Meta's still violating GDPR rules with latest plan to train AI on EU user data, says noyb

Wed 14 May 2025 // 15:34 UTC

There's a Max Schrems-shaped object standing in the way of Meta's plans to train its AI on the data of its European users, and he's come armed with several justifications for why Zuckercorp might be violating EU regulations with its stated plans.

The Austrian lawyer's noyb (None Of Your Business) privacy advocacy group sent a cease and desist letter to Meta Wednesday. Noyb's letter tells the Facebook giant that, if it opts not to discuss the matter or agree to make user data for AI data training an explicit opt-in, as opposed to the opt-out approach Meta is currently pursuing, Schrems and company are ready to file an injunction, or even take the matter to court in a class-action case.

This isn't the first time noyb has challenged Meta on plans to train its AI on the public posts and comments of EU users, as we noted in our coverage last month when Meta announced plans to resume training its AI in the EU.

https://www.theregister.com/2025/05/14/ ... dpr_rules/

[weatheriscool]

[spryfusion]

Blog post: "Introducing the V-JEPA 2 world model and new benchmarks for physical reasoning"

[wjfox]

Despite $2M salaries, Meta can't keep AI staff — talent reportedly flocks to rivals like OpenAI and Anthropic

By Jowi Morales
published 2 days ago

As companies pour billions of dollars into AI infrastructure, demand for AI talent to program and run these AI data centers is also greatly increasing. Deedy Das, a Venture Capitalist at Menlo Ventures and a former Google Search staff member, posted on X that Meta has an over $2 million annual pay package for AI talent, but still keeps losing its people to OpenAI and Anthropic. He said that he’s personally heard three such cases this week alone, which is major news given the size of Meta’s compensation.

https://www.tomshardware.com/tech-indus ... -anthropic

[weatheriscool]

Meta and Oakley May Launch AI Glasses This Friday, June 20
It's possible the new glasses will add in cycling, running, and sports videography features in addition to real-time voice translation.
By Jon Martindale June 17, 2025

Meta has teased a new partnership in its smart glasses initiative. Building on the success of its Rayban AI-powered smartglasses, it looks set to release a new line of sunglasses with Oakley, opening up new audiences and markets for the augmented reality specs.

Smart glasses have been tried numerous times over the years, with Google's discontinued Glass headset perhaps the most iconic. But in the past couple of years, Meta has had some modest success with its Ray-Ban-partnered smart glasses, which offer features like hands-free video and photo taking and AI-powered translations. It's now looking to bring that to a new style of glasses with its Oakley tie-up.

This is a no-brainer from Meta and Rayban's parent company, EssilorLuxottica, which also owns the Oakley brand. However, specific details on the partnership have yet to be released. Meta launched a new Instagram account this week called Oakley Meta, with the account following Meta, Oakley, Mark Zuckerberg, Rocco Basilico, chief wearables officer at Luxottica, and Caloa Mato, head of global marketing at Oakley, as per our colleagues at ZDNet.

https://www.extremetech.com/computing/m ... ay-june-20

[wjfox]

[weatheriscool]

[weatheriscool]

[weatheriscool]

[firestar464]

More on that by VentureBeat:

https://venturebeat.com/ai/mark-zuckerb ... ting-work/

[weatheriscool]

[firestar464]

Louisiana PSC approves Entergy's plan to power Meta's $10 billion data center

https://www.wwltv.com/article/news/loca ... 4796ea7702

[weatheriscool]

Meta partners with Midjourney on AI image and video models
Maxwell Zeff
2:41 PM PDT · August 22, 2025

Meta is partnering with Midjourney to license the startup’s AI image and video generation technology, Meta Chief AI Officer Alexandr Wang announced Friday in a post on Threads. Wang says Meta’s research teams will collaborate with Midjourney to bring its technology into future AI models and products.

“To ensure Meta is able to deliver the best possible products for people it will require taking an all-of-the-above approach,” Wang said. “This means world-class talent, ambitious compute roadmap, and working with the best players across the industry.”

The Midjourney partnership could help Meta develop products that compete with industry-leading AI image and video models, such as OpenAI’s Sora, Black Forest Lab’s Flux, and Google’s Veo. Last year, Meta rolled out its own AI image generation tool, Imagine, into several of its products, including Facebook, Instagram, and Messenger. Meta also has an AI video generation tool, Movie Gen, that allows users to create videos from prompts.

https://techcrunch.com/2025/08/22/meta- ... eo-models/

[firestar464]

Meta to launch California super PAC backing pro-AI candidates

https://www.reuters.com/world/us/meta-l ... 025-08-26/

[weatheriscool]