“Nvidia and Perplexity are partnering to offer what they call localized and sovereign artificial-intelligence models to their users in Europe, the companies said Wednesday.
The chip titan is working with local European partners, including the French AI firm H Co., to build open-source, sovereign AI models that will be offered through AI search startup Perplexity's platform.
The models will run on local AI infrastructure from European partners participating in DGX Cloud Lepton, a service designed to link AI developers with Nvidia's network of cloud providers. Businesses can also access and fine-tune or customize those AI models through an integration with Hugging Face, which runs a popular open-source model platform [1].
"We reached out to and have talked to about 13 AI model builders in the region, to be able to say that we're going to help them train models," said Kari Briski, Nvidia's vice president of generative AI software for enterprise. "To be able to both train on the languages that are unique to their culture, and then offer them in Perplexity."
With the localized, sovereign AI models, Perplexity users will be able to search and return answers in the European Union's 24 official languages, the company said.
Nvidia Chief Executive Jensen Huang has described the concept of countries investing directly in artificial intelligence as "sovereign AI" -- and it is an important element underpinning Nvidia's growth ambitions.
Entities including the Slovak Republic and the University of Ljubljana in Slovenia are teaming up with Nvidia to help optimize their models, some of which specialize in national language and culture, the company said.
The announcement comes amid a flurry of international dealmaking by Nvidia in recent months, which included a massive chip deal with the United Arab Emirates and designs for a supercomputer in Taiwan.
Expanding Nvidia's reach in Europe, including its recent partnerships in the U.K. around sovereign AI, is part of the plan for making AI more accessible to more people and enterprises around the world -- all of which ultimately benefits Nvidia's chip business, said Jim McGregor, founder of Tirias Research.
Perplexity, which has grown quickly following the public launch of OpenAI's ChatGPT in 2022, sells an AI-powered search tool that summarizes answers using information gathered from the web. It was recently in advanced talks to raise $500 million, valuing it at $14 billion -- a more than 50% increase from late last year.
The AI search company uses a variety of large language models to support its answer engine. That's partly why Perplexity co-founder and CEO Aravind Srinivas said he doesn't believe "the world should have one model to rule them all."
"I think everybody should be able to build tailor-made, custom AIs for their needs and cultural preferences and so on," Srinivas said in an interview. "We want to make Perplexity a pretty model-agnostic platform."
Last year, The Wall Street Journal's parent company, Dow Jones, sued Perplexity for copyright infringement.
It is likely that as regions like Europe expand their rules for sovereign technology -- especially for highly regulated sectors -- and have a greater need for localized AI, there will be more local versions of Perplexity and AI tools like it, Tirias Research's McGregor said.
For Perplexity, one benefit of working with Nvidia is that it raises the AI search firm's profile as it goes up against even deeper-pocketed competitors like OpenAI, McGregor said.
Nvidia and Perplexity have been working together since the start of 2024, Srinivas said, and the search company uses the chip giant's hardware as well as some of its AI models.” [2]
1. Hugging Face is involved in the development and sharing of open-source resources for building self-driving models.
Here's how:
LeRobot Platform: Hugging Face has expanded its LeRobot platform to include training data specifically for self-driving machines.
L2D Dataset: In collaboration with Yaak, they are developing L2D, which aims to be the largest open-source dataset for training end-to-end self-driving models.
Community Contribution: Hugging Face facilitates collaboration and knowledge sharing within the AI community, allowing users to build and share applications, including those for self-driving cars.
While they don't host a single, ready-to-deploy self-driving car model, they provide the necessary tools, data, and a platform for the community to build and share such models. They are actively supporting the development of open-source self-driving technology.
2. Perplexity, Nvidia Join For AI Push In Europe. Lin, Belle. Wall Street Journal, Eastern edition; New York, N.Y.. 12 June 2025: B1.
Komentarų nėra:
Rašyti komentarą