UAE Releases Falcon Arabic AI Model To Outcompete Mideast Rivals

A research arm of the Abu Dhabi government has released a powerful new Arabic-language artificial intelligence model in a bid to preserve its technological lead over rivals in the Middle East.
The new system, called Falcon Arabic, was trained on a dataset spanning Modern Standard Arabic and regional dialects. The Technology Innovation Institute (TII), the Abu Dhabi group behind Falcon, claims the new offering matches the performance of models up to 10 times its size. TII also launched Falcon H1, a small model that it said outperforms similarly sized options from Meta Platforms Inc. and Alibaba Group Holding Ltd.

Try Our AI Features

Explore what Daily8 AI can do for you:

Learn with AIFact CheckingText to SpeechAI Summary

Related Articles

Yahoo

35 minutes ago

• Yahoo

Meta (META) Ramps Up AI Talent War With New Hires From Safe Superintelligence

Meta Platforms (META, Financials) is set to bring on Daniel Gross, CEO of Safe Superintelligence, and former GitHub CEO Nat Friedman after a failed attempt to acquire the $32 billion AI startup founded by OpenAI co-founder Ilya Sutskever, according to sources cited by CNBC. Warning! GuruFocus has detected 5 Warning Sign with META. Meta will integrate Gross and Friedman into its AI unit under Alexandr Wang, whom Zuckerberg hired last week alongside a 49% stake in Wang's company, Scale AI. Meta will also take a stake in NFDG, the venture firm co-run by Gross and Friedman. Gross previously led machine learning efforts at Apple (AAPL, Financials) and co-founded Safe Superintelligence with Sutskever. Friedman served as GitHub's CEO following its acquisition by Microsoft (MSFT, Financials). Their firm NFDG has invested in companies such as Coinbase, CoreWeave and Perplexity, though CNBC said it's unclear what becomes of those holdings under Meta. Meta's efforts come as major players, including OpenAI, Google and Microsoft, vie to secure top AI engineers. OpenAI CEO Sam Altman recently claimed Meta offered signing bonuses up to $100 million to poach his staff offers that were declined. This article first appeared on GuruFocus.

Yahoo

40 minutes ago

• Yahoo

Meta Fails to Buy $32B AI Firm, Hires CEO Instead in Aggressive Talent Grab

June 20 - Meta (NASDAQ:META) is set to bring on Daniel Gross,the CEO of AI startup Safe Superintelligence, following a failed effort to acquire the company earlier this year. The startup, co-founded by OpenAI veteran Ilya Sutskever, had reportedly turned down a takeover bid from Meta despite being valued near $32 billion in a recent funding round. Sources familiar with the matter said Sutskever also declined an individual offer to join Meta. Warning! GuruFocus has detected 5 Warning Sign with META. Instead, Meta is now hiring Gross along with former GitHub CEO Nat Friedman. The pair co-run NFDG, a venture firm that Meta will take a stake in as part of the hiring arrangement. Both will report to Scale AI founder Alexandr Wang, whom Meta onboarded last week through a separate $14.3 billion investment. The moves come as Meta accelerates efforts to build advanced AI systems and compete more directly with rivals like OpenAI, Google, and Microsoft (NASDAQ:MSFT). A Meta spokesperson said the company will provide further details on its superintelligence plans in the coming weeks. Gross previously led Apple's (NASDAQ:AAPL) machine learning efforts after selling his search engine startup Cue. Friedman served as GitHub's CEO following its acquisition by Microsoft. This article first appeared on GuruFocus. Error in retrieving data Sign in to access your portfolio Error in retrieving data Error in retrieving data Error in retrieving data Error in retrieving data

Forbes

2 hours ago

• Forbes

Uber Is Making A Push In Data Labeling After Scale AI's Deal With Meta

Uber is making a push in data labeling. Last week, Scale AI's bombshell deal for Meta to take a 49% stake in the company sent shockwaves throughout the industry: In its wake, prominent clients like OpenAI have pulled back on working with Scale, which the ChatGPT maker had already been doing for months. Google is also planning its split. And a host of data-labeling rivals has been emboldened to step in to fill the void. Among them is a little-known unit from a familiar giant: Uber. Since last November, the ride-hailing behemoth has operated Uber AI Solutions, a data-labeling platform focused on training AI models for enterprise clients. Now as Scale's deal has carved a new opening in the market, Uber is making its pitch to new customers. 'For Uber, our core has always been being the platform of choice for flexible on-demand work,' Megha Yethadka, general manager of the unit and a 10-year veteran of the company, told Forbes. Uber drivers, of course, are contractors that shuttle around passengers and deliveries all across the globe. 'That extends itself really well to this business of digital tasks now.' On Friday, Uber told Forbes it's making a push to expand the service. Among the updates: a new service that provides ready-to-use datasets, including audio, video, images and text, to customers training their own models. The company will also license out the platforms it uses internally for managing data labeling projects and accessing its network of contracted clickworkers, making them available for clients to use. Beyond just training models, Uber is now also offering clients tools to develop AI agents, which can take specific actions for users, like helping with customer support. Another change: Launched as Uber Scaled Solutions, the company recently swapped out the 'scaled' in its name for 'AI.' Yethadka said the rebrand had nothing to do with avoiding confusion with its similarly-named rival, but wanting to more simply convey the AI of the unit. Going forward, Uber wants to separate itself from its data-labeling rivals by automating more of the process to set up clickwork projects. The company is developing a software interface that allows clients 'to simply describe their data needs in plain language,' while the platform automatically handles assigning tasks, setting up workflows and maintaining quality control. The idea is to hand over the project to human clickworkers more quickly, instead of doing manual work to onboard workers. 'We do see an opportunity to build this into a meaningful business line for Uber.' The company said Uber AI Solutions is now available in more than 30 countries, an expansion from its five initial launch markets last November, which included the U.S., Canada and India. Since the start of this year, Yethadka said Uber has doubled the number of clickworkers on its platform. She declined to disclose how many taskers are in the company's network overall, but said there are 'tens of thousands' of people working on each topic area of tasks, including STEM, coding and law. The most engaged clickworkers spend about 3 to 4 hours a day performing tasks, which can range from $20 to $200 per hour, depending on the complexity of the work, Yethadka said. The unit has more than 50 corporate customers, including the autonomous vehicle company Aurora and Niantic, the creator of Pokemon Go that recently ditched the games business to pivot to enterprise AI. The expansion comes as Scale's tie-up with Meta has thrown the data labeling industry into a frenzy. As part of the deal, Scale CEO and founder Alex Wang is heading to Meta to lead the tech giant's newly-formed Superintelligence Lab, an effort to compete with other deep-pocketed frontier labs including those from OpenAI, Anthropic and Google. Now, 'a number of companies are, of course, looking to revisit their partner strategy for data,' said Yethadka, aiming to find vendors that are 'neutral and impartial." In the aftermath, smaller rivals, including unicorns Mercor and Turing and startup Invisible Technologies, are clamoring to pounce. But Uber stands out among the competition because of its sheer size and resources, Yethadka argues. 'A lot of companies in this space are a lot smaller, VC-funding dependent,' she said. Meanwhile Uber, which is worth $175 billion and tallied $43.9 billion in revenue last year, is a more reliable long-term bet, Yethadka said. (She declined to break out the data-labeling unit's revenue.) While other companies in the space more resemble service providers, Uber has a long history of shipping products, which brings a different perspective when the company collaborates with customers, she said. 'We have been a product company and an operations company, and have done this for a living ourselves,' said Yethadka. 'We do see an opportunity to build this into a meaningful business line for Uber.' Even Scale had taken notice of Uber before the Meta deal. 'This space is full of opportunities. I think more people are seeing the value of the work we're doing here, which is why even a business like Uber will want to try their hand at the same space,' Xiaote Zhu, GM of Scale's Outlier platform for generative AI clickwork, told Forbes earlier this year. Still, it's not a foregone conclusion that Uber will come out on top, competitors say. The spoils will go to the company that compiles the best pool of clickworkers. 'Data annotation is transitioning towards higher and higher-skilled work,' Brendan Foody, CEO of $2 billion-valued Mercor, told Forbes. 'Uber's success will depend on how effectively they build this high-skilled talent network.' What's more, Uber has its own baggage. For years, the company limped through controversies around regulation and the treatment of its contract drivers. Yethadka said customers have not minded, and that Uber has committed to 'do the right thing' when it comes to data confidentiality and security controls. 'And that continues to be applied to this new line of business as well,' she said.