 |
|
|
|||||
Mark Zuckerberg has decided he wants to claw back lost ground in the AI race, and hes willing to spend heavily to do it. The Meta CEO is reportedly hiring staff personally for a new superintelligence lab within the company, offering massive salaries to attract top talent. Meta has also announced the acquisition of Scale AI, with founder Alexandr Wang joining the superintelligence team. I’m not surprised with the choice of Alex Wanghe is friends with Zuckerberg, says a former Meta employee, granted anonymity due to restrictions from their current employer. Wangs political leanings may help deflect regulatory scrutiny, the former staffer adds, and he also has a track record of executing whereas the new Meta AI leads have messed up Llama big time. The real challenge may be finding people to join them. AI talent is in high demand, with leading labs offering millions to top engineers. Meta, with its vast resources, is reportedly offering up to nine-figure compensation packages. Meta has been trying to pull in top-tier talent from other labs for a while now, like the big open source AGI announcement on Zucks Instagram, says another industry source who requested anonymity. I guess they felt they needed to mix it up for whatever reason. But Metas reputation could be a hurdle. What you have to understand about Meta is that it’s very Game of Thrones-like, with people constantly fighting over power and influence, says the former employee. Bloat and excessive middle management remain problems. That’s a major reason why their products are crap and now it’s affecting their AI research too, the former employee adds. (Meta did not immediately respond to a request for comment.) According to the anonymous industry source, one potential draw for candidates is Metas vast supply of GPUs. Meta has an absolutely massive stockpile of GPUs it doesnt seem to be using particularly well, so possibly they think that could turn some heads, they say. Still, the companys AI output has often trailed competitors. The hope would probably be that the new org completely clears house. Ray Wang, a Washington-based semiconductor and technology analyst, agrees hiring could be tough, especially given Metas recent stumbles in large language model development. Still, he sees reasons some might sign on. Meta still possesses deep hardware and software resources, platforms with tens of millions of users, and forthcoming AI-enabled products, which together should offer researchers and engineers a compelling environment for developing future cutting-edge models and implementing them in the companys products and platforms at scale, Wang says. The money is likely to be the biggest motivator. The main pull is probably compensationthough its possible some people could feel like they are getting in on the ground floor of a new org despite Meta being a giant, says the industry source. But for potential hires, a lingering concern may be Zuckerbergs direct involvement in the new lab. Im also not surprised Zuckerberg is deeply involved, says the former Meta staffer. Hes always done that when things he cares about go badly. Or he feels he can make them better by getting involved. Sadly, he has a poor track record: Rooms, e-commerce, metaverseall flops.
Category:
E-Commerce
Elon Musk stepped back from his explosive feud with U.S. President Donald Trump, writing on X that he regrets some of his posts about his onetime ally and that they went “too far.”Early Wednesday morning, he posted “I regret some of my posts about President @realDonaldTrump last week. They went too far.”Musk’s break with a president whom he spent hundreds of millions of dollars to elect appeared to put an end to his influence in the White House and prompted concerns about effects on his companies. As a major government contractor, Musk’s businesses could be particularly vulnerable to retribution, and Trump has already threatened to cut Musk’s contracts.Musk earlier deleted a post in which he claimed without evidence that the government was concealing information about the president’s association with infamous pedophile Jeffrey Epstein. Meanwhile, other posts that irritated Trump, including ones in which Musk called the spending bill an “abomination” and claimed credit for Trump’s election victory, remained live.On Sunday, Trump told NBC’s Kristen Welker that he has no desire to repair their relationship and warned that Musk could face ” serious consequences ” if he tries to help Democrats in upcoming elections. Associated Press
Category:
E-Commerce
Databricks, known for secure data storage and AI, has launched a new AI business intelligence dashboard called Databricks One, designed for nontechnical business users across various departments. While Databricks initially focused on providing secure data storage for large organizations (it works with 60% of the Fortune 500), it has in recent years shifted toward helping customers harness the power of AI models situated close to their data.Until now, access to the Databricks platform was primarily limited to engineers and data scientists with SQL or Python skills. Databricks One introduces a completely redesigned interface, simplified specifically for users without a technical background.We want to make our experience for nontechnical users as amazing as our experience for technical users, said Databricks cofounder and CEO Ali Ghodsi. This is our first step of making this true so that everyone across the organization can unlock the full value of their data and drive innovation.The Databricks One tool resembles an AI chatbot interface. Users can describe the type of data or analysis they need, and a large language model translates those requests into actions. According to Ghodsi, this could involve deploying AI agents that tap into pipelines of various models and databases to retrieve data or perform specific analyses.Ordinarily, such requests would require coding, but the dashboard handles that complexity for the user. Behind the scenes its actually creating SQL and executing on a data warehouse, and all of that is abstracted away from you, Ghodsi says.[Animation: Databricks]To help users better understand the data, Databricks integrated visualizations directly into the One interface. Users can interact with an AI/BI Genie to refine their view, asking custom questionsnot just selecting from preset queries offered by the dashboard.A user might run advanced analytics on the effectiveness of a marketing campaign, conduct a legal review of overlapping (and potentially conflicting) business contracts, or quickly gather everything a salesperson needs before meeting a prospect.Ghodsi says the idea for Databricks One came from many conversations with large enterprise clients. These organizations wanted to put AI-powered business intelligence into the hands of more employees who could act on it. At the same time, they were wary of the security risks involved in broadening access to potentially sensitive data on the Databricks platform. Threading that needle and making sure that you have the right balance is very hard, and so were very excited we finally found the thing that works for all of our customers, Ghodsi says.[Image: Databricks]Specifically, customers were concerned that nontechnical users might misinterpret the data, access information they shouldnt, or use the platform in ways that couldnt be properly audited. Many of Databrickss clients operate in tightly regulated sectors like healthcare and financial services, where handling sensitive personal data is routine.In response, Databricks spent two years building guardrails to address these risks. This included systems that assign different levels of access based on user roles and permissions. According to Ghodsi, the One platform keeps customer data secure and sandboxed, preventing leaks. Some users have read-only access, and the company does not use one customers data to train models for another.Ghodsi believes the simplified interface could make the platform a strong fit for midsize and even small businesses in the future.Following a massive $10 billion Series J raise in December 2024, Databricks has brought in about $14.4 billion in venture funding, along with $5.25 billion in debt financing, reaching a $62 billion valuation. Its value has grown rapidly as large companies try to use AI to mine their own data for actionable insights. In 2023, the company was valued at $43 billionmarking a roughly 44% increase in a single year during the so-called AI boom. Many analysts expect a Databricks IPO in 2025 or 2026, depending on market conditions.Databricks One is currently in private preview, with a public beta expected later this summer. Ghodsi says the company wont charge seat licenses but will instead use a metered pricing model based on usage time and token consumption by AI models.Ghodsi says its ultimately up to customers to decide how widely to deploy the tools. The litmus test is easyare they like Okay, lets open it up to more people?
Category:
E-Commerce
All news |
||||||||||||||||||
|
||||||||||||||||||