Meta taps top researchers from Google and Sesame for new AI lab
Jack Rae, a principal researcher at Google DeepMind, is expected to join Meta's "superintelligence' team, according to multiple people familiar with the matter, who spoke on condition of anonymity as the information is not public. Meta has also recruited Johan Schalkwyk, a machine learning lead at AI voice startup Sesame, other people said.
Alexandr Wang, co-founder and chief executive officer of Scale, is also expected to be part of the team after Meta finalizes a multibillion-dollar investment in the data labeling startup that could be announced as soon as this week.
The new group is part of an ambitious, and costly, effort by Meta CEO Mark Zuckerberg to gain ground on rivals like Google and OpenAI after he was frustrated by the poor reception to the company's latest AI offering, Llama 4. Zuckerberg hopes the new hires can help improve Llama's models and build better AI tools for voice and personalization features.
Meta and Zuckerberg are offering compensation packages worth tens of millions of dollars over several years, including equity, according to people familiar with the matter. A spokesperson for Meta declined to comment. Rae confirmed he's leaving Google for Meta, but declined to comment further. Schalkwyk did not respond to a request for comment.
Zuckerberg has begun recruiting a brain trust of AI researchers and engineers, at times meeting with candidates at his homes in Lake Tahoe and Palo Alto, California. Often, the CEO has reached out to potential hires directly, according to people familiar with the process. Meta aims to hire around 50 people for the new team, including a chief scientist to help oversee the group, one person said.
The team is coming together as Meta is also nearing a deal to invest billions of dollars in Scale to bolster its AI efforts. Scale uses an army of contractors to label the data that tech firms such as Meta and OpenAI need to train and improve their AI models.
Wang, Scale's 28-year-old CEO, is an influential figure in the industry who has cultivated close ties with some in Washington. Other employees from Scale are likely to join Meta's superintelligence team after the investment is finalized, according to a person familiar with the deal.
The three pillars of AI development are chips, talent and data. Meta has lots of chips. The Scale partnership may help bolster its access to quality data. And with the power of its checkbook, Meta and Zuckerberg hope to make a bigger dent in the AI talent wars.
"There are very few people globally who can do those type of large AI trainings very, very efficiently,' said Vahan Petrosyan, co-founder of SuperAnnotate, an AI data platform. For that reason, he said, higher pay packages may make sense for companies like Meta.
Still, not everyone is jumping at Zuckerberg's recruitment push. At least one person at a leading AI research lab rejected a lucrative offer from Zuckerberg, according to a person familiar with the matter.
Meanwhile, Meta's competitors are racing to retain top talent. Google named Koray Kavukcuoglu, one of its top AI researchers, as its chief AI architect — a new senior vice president role that reports directly to Chief Executive Officer Sundar Pichai, according to an internal memo.
Kavukcuoglu will continue to serve as the GenAI unit lead and chief technology officer of Google DeepMind, Pichai said in the memo.
"Koray will help to accelerate how we bring our world-leading models into our products, with the goal of more seamless integration, faster iteration, and greater efficiency,' Pichai said.
Google didn't immediately respond to a request for comment.
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Nikkei Asia
an hour ago
- Nikkei Asia
Japan's Eneos to adopt AI for failure prevention at refineries
Eneos' Sendai refinery was one of the first to test out an AI system that will be used at all of its refineries in Japan. (Photo by Mao Kawano) MAO KAWANO August 15, 2025 07:15 JST TOKYO -- Japanese energy company Eneos Holdings will introduce AI-equipped failure prevention systems at all of its nine refineries in Japan by fiscal 2026.
Japan Today
3 hours ago
- Japan Today
Japan consortium developing misinformation-fighting AI model
A Japanese consortium involving Fujitsu Ltd is developing a fact-checking platform using artificial intelligence, seeking to counter the type of online misinformation often seen during disasters and around elections. The group of nine organizations -- including the National Institute of Informatics, NEC Corp and academic institutions -- aims to complete the system by the end of fiscal 2025. The AI will analyze, collect supporting data and assess the authenticity of information online. "It involves lots of checking when you want to ascertain truth on your own, but the system could help us make quick judgments," said Dai Yamamoto, a senior project director at Fujitsu. In late May, Yamamoto tried the system during its development, entering the statement, "A group of foreign thieves went to quake-hit areas immediately after the Noto earthquake," and asking it to verify. A few seconds later, a message judging the statement as "False" popped up. The system backed its assessment with a newspaper article, saying it found information that contradicted the claim. It also concluded the reliability of the news organization that published the report was "high." The system, which uses a large language model specializing in fighting false information, is being developed with 6 billion yen in funding provided by the New Energy and Industrial Technology Development Organization, a government agency. Using the platform, even "deepfakes" -- images and videos featuring real people but created by AI -- can be detected as it spots characteristics that tend to be overlooked, according to the developers. The move comes as concerns grow about the dissemination of online disinformation and misinformation, such as fake videos of disasters, and their negative impact at times of crisis. © KYODO
Japan Times
18 hours ago
- Japan Times
Apple plots expansion into AI robots, home security and smart displays
Apple is plotting its artificial intelligence comeback with an ambitious slate of new devices, including robots, a lifelike version of Siri, a smart speaker with a display and home-security cameras. A tabletop robot that serves as a virtual companion, targeted for 2027, is the centerpiece of the AI strategy, according to people with knowledge of the matter. The smart speaker with a display, meanwhile, is slated to arrive next year, part of a push into entry-level smart-home products. Home security is seen as another big growth opportunity. New cameras will anchor an Apple security system that can automate household functions. The approach should help make Apple's product ecosystem stickier with consumers, said the people, who asked not to be identified because the initiatives haven't been announced. It's all part of an effort to restore Apple's mojo. Its most recent moonshot project, the Vision Pro headset, remains a sales flop, and the design of its bestselling devices has remained largely unchanged for years. At the same time, Apple has come under fire for missing the generative AI revolution. And OpenAI may even threaten the company's home turf by developing new AI-driven devices with the help of former Apple design chief Jony Ive. Though Apple is still in the early stages of turning around its AI software, executives see the pipeline of hardware as a key piece of its resurgence — helping it challenge Samsung, Meta and others in new categories. A spokesperson for Cupertino, California-based Apple declined to comment. Because the products haven't been announced, the company's plans could still change or be scrapped. Many of the initiatives and their timelines rely on Apple's continued progress in AI-powered software. Chief Executive Officer Tim Cook told employees in an all-hands meeting this month that Apple must win in AI and hinted at the upcoming devices. "The product pipeline — which I can't talk about — it's amazing, guys. It's amazing,' Cook said. "Some of it you'll see soon. Some of it will come later. But there's a lot to see.' Beyond the home devices, Apple is preparing thinner and redesigned iPhones for release this year. And further out, it aims to introduce smart glasses, a foldable phone, a 20-year anniversary iPhone and a revamped headset dubbed N100. It's also planning a large foldable device that melds a MacBook and an iPad. Apple CEO Tim Cook is banking on an ambitious product road map to help get the company's AI effort on track. | Bloomberg Apple is looking to boost sales after years of slowing growth for its flagship products. It also nixed some expansions into new areas, like self-driving cars, adding pressure to find other sources of revenue. Moreover, the new initiatives will help rebut the idea that the company is no longer innovating like it used to. It was first reported last year that Apple was moving forward with a tabletop robotics project, codenamed J595, and developing a new smart-home strategy. But now a clearer picture is forming of its push into that market — and what it means for its AI ambitions. The tabletop robot resembles an iPad mounted on a movable limb that can swivel and reposition itself to follow users in a room. Like a human head, it can turn toward a person who is speaking or summoning it, and even seek to draw the attention of someone not facing it. The hope is to bring AI to life in ways that other hardware makers have yet to do. Apple imagines customers placing it on a desk or kitchen counter and using it to get work done, consume media and manage their day. FaceTime calls will also be a key function of the device. During videoconferencing, the display will be able to shift to lock on to people around a room. Apple is testing a feature that turns an iPhone screen into a joystick, letting users move around the robot to show different people or items in a room during video calls. Attendees view new products during an event at Apple Park campus in Cupertino, California, on Sept. 9, 2024. | Bloomberg But the hallmark of the device is an entirely new version of the Siri voice assistant that can inject itself into conversations between multiple people. It will be able to engage with users throughout the day and more easily recall information. The idea is for the device to act like a person in a room. It could interrupt a conversation between friends about dinner plans, say, and suggest nearby restaurants or relevant recipes. It's also being designed to engage in back-and-forth discussions for things like planning a trip or getting tasks done — similar to OpenAI's voice mode. Apple is planning to put Siri at the center of the device operating system and give it a visual personality to make it feel lifelike. The approach, dubbed Bubbles, is vaguely reminiscent of Clippy, an animated paper clip from the 1990s that served as a virtual assistant in Microsoft Office. Apple has tested making Siri look like an animated version of the Finder logo, the iconic smiley face representing the Mac's file management system. A final decision on its appearance hasn't been made, with designers considering ideas that veer closer to Memoji, the playful characters that represent Apple user accounts. Device prototypes use a roughly 7-inch horizontal display, approaching the size of an iPad mini. The motorized arm can extend the display away from the base roughly half a foot in any direction. Some people familiar with the product call it the "Pixar Lamp,' referring to the animated film company's famous logo. Apple has previously disclosed some research in this area: It published a paper in January detailing a light fixture that uses robotics to move around. Apple has multiple teams across its AI, hardware, software and interface design groups tackling on the project. The work is being led in part by Kevin Lynch, who previously oversaw a push into smart watch software and cars. The technology giant is developing several other robots. It has teams exploring a mobile bot with wheels — something akin to Amazon's Astro — and has loosely discussed humanoid models. Apple has a group actively developing a large mechanical arm for use in manufacturing facilities or handling tasks in the back of retail stores, a move that could potentially replace some staff. Such a robot, codenamed T1333, remains several years away. Google has long offered smart display products. | Bloomberg The smart-home push includes a standalone display poised to launch by the middle of next year. That device, codenamed J490, is a stripped-down variant of the robot, lacking the arm and conversational Siri — at least to start. It will still have home control, music playback, note taking, web browsing and videoconferencing. It may also include the new Siri visual interface. Both the smart display and tabletop robot will run a new operating system dubbed Charismatic, which is designed to be used by multiple people. The interface largely centers on clock faces and widgets — small software features that are typically dedicated to specific tasks. Charismatic, which was previously known as Pebble and Rock earlier in development, blends the approach of the Apple TV and Apple Watch operating systems. It offers features like multiuser modes and clock-face themes, such as one based on Snoopy, the beagle from the Peanuts comic strip. The devices are meant to be easily shared: They include a front-facing camera that can scan users' faces as they walk toward it and then automatically change the layout, features and content to the preferences of that person. Some versions of the software use circular app icons and feature a hexagonal grid of apps. Apple is planning to include many of its core apps, including the calendar, camera, music, reminders and notes software. But the interface will be heavily reliant on voice interaction and widgets, rather than jumping in and out of apps. Though the device will have a touch screen, the primary input method will be Siri and an upcoming upgrade to a feature dubbed App Intents. That software lets users precisely control the interface and applications via voice. The hardware itself looks similar to a Google Nest Hub but is shaped like a square, with thin black or white bezels and rounded corners. The non-robotic 7-inch display sits on a half-dome-shaped base, which includes some of the electronics and is perforated around the bottom edges for speakers and microphones. It can also be mounted on a wall. The launch will mark the first time Apple is making a serious push into the smart home and comes nearly a decade after Amazon and Google started shipping smart speakers with screens. The home is a critical space for Apple to target, especially as more users consume content from the living room and automate household functions. Apple has long had a strong foothold in mobile devices and quickly became a player in the automotive industry via CarPlay — but that success hasn't followed into the smart home. Though the company launched HomeKit for controlling third-party devices in 2014, it has had limited success with its own HomePod speakers. The Apple logo during the opening of the new Apple Malaysia store in Kuala Lumpur, Malaysia, in 2024. | Bloomberg Core to the new home devices — and current products like iPhones and iPads — is an overhaul to the underpinnings of Siri. Engineers are working on a version codenamed Linwood with an entirely new brain built around large language models — the foundation of generative AI. The goal is to tap into personal data to fulfill queries, an ability that was delayed due to hiccups with the current version. That new software, known internally as LLM Siri, is planned for release as early as next spring. But work is going even further: Apple is preparing a visually redesigned assistant for iPhones and iPads that will also debut as early as next year. Craig Federighi, senior vice president of software engineering, hinted at a bigger-than-anticipated overhaul in an internal meeting with employees this month. "The work we've done on this end-to-end revamp of Siri has given us the results we needed,' he said, adding that "this has put us in a position to not just deliver what we announced, but to deliver a much bigger upgrade than we envisioned.' He said that "there is no project people are taking more seriously.' Linwood is based on technology developed by the Apple Foundation Models team, but the company has a competing project dubbed Glenwood as well that would power Siri with outside technology. A final decision hasn't been made on which models will be used, but Apple has been testing Anthropic's Claude for this purpose. Mike Rockwell, the former Vision Pro chief who was put in charge of Siri earlier this year, is overseeing both the Linwood and Glenwood efforts. During development of the tabletop robot, Apple engineers have made heavy use of ChatGPT and Google Gemini to build and test features. Within Apple's AI and Siri teams as a whole, software developers are increasingly using third-party systems as part of their development process. Apple is working on a camera, codenamed J450, designed for home security, detecting people and automating tasks. The device will be battery-powered and could last from several months to a year on a single charge, on par with rival products. The device has facial recognition and infrared sensors to determine who is in a room. Apple believes users will place cameras throughout their home to help with automation. That could mean turning lights off when someone leaves a room or automatically playing music liked by a particular family member. Apple is planning to develop multiple types of cameras and home-security products as part of an entirely new hardware and software lineup. The goal is to compete with Amazon Ring, Google Nest and Roku, capitalizing on its customer loyalty to sell more products. It has also tested a doorbell that uses facial recognition technology to unlock a door. Apple already sells iCloud+ subscriptions with online storage for security footage, but they're aimed at third-party cameras.
