After Scale AI, Meta is going after OpenAI co-founder Ilya Sutskever's startup Safe Superintelligence
Meta is on AI overdrive. After steadily losing talent to rivals and a faulty product launch during its developer event, Llamacon, in May, the social media giant is desperately trying to stay in the AI race. Its plan: if you can't build internally, then go shopping.
Last week (June 10), Meta agreed to pick 49% stake in data-labelling firm Scale AI — a deal that brought the company's founder Alexandr Wang under Meta's fold. Now, per media reports, the social network is looking to bolster Mr. Wangs' team.
After its failed attempt to buy OpenAI co-founder Ilya Sutskever's startup Safe Superintelligence Inc earlier this year, Meta has opened communication channels with SSI CEO Daniel Gross and ex-Github CEO Nat Friedman. It must be noted that both — Mr. Gross and Mr. Friedman — run a venture fund called NFDG.
Per media reports, the duo is expected to join Meta and report to Mr. Wang — Meta's superintelligence department chief. Meta will also own a portion of NFDG under the terms.
Neither Mr. Gross nor Mr. Friedman have confirmed or denied talks with Meta.
Apart from this, Meta has also attempted poaching OpenAI's AI researchers by offering some even upto $100 million as joining bonus
Mr. Sutskever's startup SSI, per an April funding round, was valued at $32 billion. The firm was established by the Canadian computer scientist after he quit OpenAI over alignment issues.
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Hans India
an hour ago
- Hans India
Apple Eyes Generative AI to Speed Up iPhone and Mac Chip Design, Confirms Hardware Chief
Apple is preparing to bring generative AI into the heart of its chip development process, a move that could transform how the company designs processors for its iPhones, Macs, and other devices. Johny Srouji, Apple's Senior Vice President of Hardware Technologies, recently confirmed the company's growing interest in using AI tools to make its silicon design process faster and more efficient. Speaking in Belgium while receiving an award from Imec, a renowned semiconductor research institute, Srouji revealed that Apple sees strong potential in applying AI-driven automation to speed up its chip development timeline. According to a report by Reuters, which reviewed a recording of the event, Srouji stated, 'Generative AI techniques have a high potential in getting more design work done in less time, and it can be a huge productivity boost.' Srouji emphasized that while AI offers immense benefits, it must be complemented by robust design infrastructure. He pointed out the crucial support Apple receives from electronic design automation (EDA) partners such as Cadence Design Systems and Synopsys. 'EDA companies are super critical in supporting our chip design complexities,' he said. Apple's exploration of generative AI for chipmaking reflects a broader industry trend, with tech rivals like Google and OpenAI investing heavily in artificial intelligence. Google, during its I/O 2025 developer event, showcased a range of AI innovations, while OpenAI continues to lead advancements in conversational AI through ChatGPT, further intensifying competition in the AI race. Though Apple has faced criticism for lagging in the consumer-facing AI space—especially after delays in rolling out promised Apple Intelligence features—its latest announcement indicates a shift toward strengthening its behind-the-scenes AI capabilities. Apple has been developing its in-house chips since 2010, beginning with the A4 processor for the iPhone. Since then, the tech giant has expanded its custom silicon portfolio to include the A-series for mobile devices and the M-series for its Mac lineup. The transition from Intel processors to Apple-designed M-series chips was a bold move, but one that has paid off significantly. These chips have helped Apple achieve industry-leading performance, battery efficiency, and tighter hardware-software integration across its devices. 'Moving the Mac to Apple Silicon was a huge bet for us,' Srouji reflected. 'There was no backup plan, no split-the-lineup plan, so we went all in, including a monumental software effort.' Now, with plans to incorporate AI into its chip design workflow, Apple is signaling a new chapter in its silicon strategy. While consumer-facing features may still be in development, the company is clearly investing in foundational technologies that could sharpen its competitive edge in the years ahead. As Apple continues to explore generative AI applications in hardware, the future of Apple Silicon may be even faster, more powerful, and more intelligently designed—bringing the company in closer competition with AI pioneers across the tech landscape.
Time of India
2 hours ago
- Time of India
Australia's teen social media ban faces a new wildcard: teenagers
HighlightsThirteen-year-old Jasmine Elkin expressed skepticism about the effectiveness of age-checking software designed to restrict social media access for users under 16, stating that individuals will find ways to bypass such measures. From December, social media platforms like Meta's Facebook and Instagram, Snapchat, and TikTok could face substantial fines for failing to implement reasonable steps to block younger users, amidst growing global scrutiny of child protection efforts online. The trial of photo-based age estimation products revealed that while many young testers had their ages guessed correctly, the reliability of the software remains uncertain, with the need for the government to establish acceptable effectiveness standards. When 13-year-old Jasmine Elkin tried out the age-checking software Australia might use to ban children and teenagers from social media, she was surprised some products could identify a person's age to the month - but she still doubts it will work. "People are always going to find a way to get past it," said the Perth schoolgirl who trialled five photo-based age estimation products with about 30 other students in May. "They can get their brother or sister to take a photo. There's nothing really that you can do about it." Elkin's view echoes one of the main concerns of child protection advocates, tech firms and even the trial organisers about the technology Australia hopes will enable the world's first national social media ban for under-16s: the software works, they say, but young people will find a way around it. From December, social media companies like Meta's Facebook and Instagram, Snapchat and TikTok will face a fine of as much as A$49.5 million ($32.17 million) if they fail to take what the law calls "reasonable steps" to block younger users in an effort to protect their mental and physical health. The platforms say users need to be at least 13 years old to sign up for an account. How well the ban works could reverberate across some of the world's largest companies and the governments seeking to contain them: already Britain, France and Singapore are making efforts to keep children and teens off social media, while U.S. states including Florida are challenging free speech laws by pushing for a ban. Even the law's opponents are likely to be watching closely: X owner Elon Musk, who has been advising U.S. President Donald Trump and is a vocal opponent of platform moderation and regulation, has criticised the measure and called the regulator overseeing it a "censorship commissar". "Everybody is looking at Australia," said Colm Gannon, CEO of the International Centre for Missing and Exploited Children Australia, a member of the trial's stakeholder adviser group. "Australia should be really focusing on robust technology, robust testing and making sure the scope of the actual project is in line with the needs that they're trying to address." The organisers of the trial, which ended this month, say it was designed to determine whether the software worked as promised, and that nearly 60 products were pitched. But it also underlined the teenagers' tech skills - testers were so fast completing their assignments, organisers doubled the number of products they tested and halved session times as the project progressed. "It hasn't been our intention to pull apart the software, rip the guts out and work out every different way that you could circumvent it," said Andrew Hammond, general manager at tech contractor KJR, which ran the trial. They will present an overview of the findings on June 20 and deliver a detailed report to the government by the end of next month. That will inform the eSafety Commissioner's advice to the government, which cited risks from cyberbullying, harmful depictions of body image and misogynist content in pushing forward with the legislation. "We know that social media age restrictions will not be the end-all be-all solution for harms experienced by young people online, but it's a step in the right direction to keep our kids safer," said a spokesperson for Communications Minister Anika Wells. What about the teens? For some of the young Australians who participated, the trial was a glimpse into a world six months in the future where, according to the law, they will no longer be allowed to use a platform they have come to rely on for daily communication. "I use it a lot, but I can still live without it," said Canberra school student Charlie Price, 14, who trialled four software options in a room with about 60 peers and had his age guessed accurately (someone in his testing group was wrongly assessed at over 20). "I know people that will get really shocked and upset," added Price, who uses Snapchat, Instagram and messaging platform Discord and plans to collect phone numbers of his online associates before December. Like Elkin, he said he thought some teenagers might try to get around the block. Emanuel Casa, 15, who was in the same group, said the test subjects tried to check the products for ease and accuracy, but "no one tried to challenge it necessarily, like no one tried to trick it." Hammond said software that revolved around a user submitting a selfie - sometimes with different facial expressions - proved the fastest and most accurate way to identify teenagers. Products involving credit card details proved impractical since few young teens had their own cards, while those that required a person to hold up their hand in various positions gave too broad an age estimate for people near the 16 cutoff, he added. No further trials have been scheduled, but Hammond said the government would need to decide on the level of software reliability it was prepared to accept. Most of the young testers had their ages guessed correctly most of the time, but a peer of Elkin, the 13-year-old, was placed at 42 by one product, she said. "There is no measure at the moment as to what 'good' is. Do they need to be 70% effective or 80% effective or 100% effective?" said Hammond. "The government so far hasn't indicated that they're going to mandate a particular solution." Nathanael Edwards, principal of Radiant Life College, a Queensland high school where 35 students along with a few parents and teachers participated, said his group tested a basic age-gating product where a person typed in their birthday. Some did as asked, while others faked a birthday to age themselves up - although not always successfully. "I think the mathematics caught a couple of kids out," he said.
India Today
3 hours ago
- India Today
Midjourney launches V1 AI video generation model right after Disney accuses it of plagiarism
Midjourney, the AI startup famous for its surreal image generation tools, is making a bold leap into video. Recently, the company unveiled V1, its long-awaited video-generation model that promises to breathe life into your static images. It's a big move for Midjourney as it throws the company into direct competition with other big-hitters like OpenAI, Runway, Adobe and Google.V1 is designed as an image-to-video model, allowing users to transform either their own uploaded pictures or Midjourney's AI-generated images into short five-second video clips. Like its sibling image models, V1 is only accessible via Discord for now and is web-only at launch. advertisementAnd it's not just videos Midjourney has in its sights. In a blog post, CEO David Holz set out some pretty ambitious goals for the company's AI, saying V1 is just the next stepping stone toward real-time 'open-world simulations.' The company also revealed its plans to branch into 3D renderings and real-time generative models down the line. While Midjourney's image tools have long appealed to artists and designers, the company has taken a slightly different tack with video. Many of its rivals — such as Sora by OpenAI, Runway's Gen-4, Firefly by Adobe and Veo 3 by Google — are going after commercial filmmakers and studios with highly controllable AI tools. Midjourney, however, is positioning itself as more of a creative playground for those looking for something a little more V1 AI video generation model: Pricing and availabilityadvertisementDespite this, Midjourney is pushing ahead. Video generation doesn't come cheap, though. V1 consumes eight times more credits per clip than Midjourney's still-image tools, so subscribers will burn through their monthly allowances far faster. At launch, Basic subscribers — who pay $10 (around Rs 866) per month — can access V1, but unlimited video generation is limited to the $60 (around Rs 5,200) Pro and $120 (approximately Rs 10,400) Mega plans, and only on the 'Relax' mode, which produces videos more slowly. However, the company says it will review this pricing structure in the coming weeks as it gathers feedback from for the tools themselves, V1 offers a surprising level of control. You can opt for an 'auto' mode that lets the AI generate motion for you or a 'manual' mode that accepts text prompts to dictate exactly how you want your animation to move. Plus, there are settings for adjusting movement intensity — 'low motion' if you want subtle shifts, or 'high motion' for more energetic effects. Clips last five seconds by default but can be extended up to 21 seconds in four-second accuses Midjourney of plagiarismThat said, Midjourney is entering the video arena under a legal cloud. Only a week ago, Disney and Universal sued the startup over its image-generation models, claiming they can produce unauthorised versions of famous characters like Darth Vader and Homer Simpson. It's part of a growing backlash across Hollywood as studios grow nervous about AI tools replacing human creatives — and AI companies face questions about training data and copyright examples of V1's output suggest Midjourney is sticking to its trademark surreal aesthetic rather than aiming for hyper-realism, the sort of style that fans of the platform have come to love. The initial reaction from users has been mostly positive so far, though it's still too early to tell how V1 will stack up against more established players like Runway and Sora.
