meta platform hired an Oslo-based team that was building out late last year artificial intelligence networking technology and the British chip unicorn Graphcore.
A Meta spokesperson confirmed the hiring in response to a request for comment Reuters identified 10 people whose LinkedIn profiles said they worked at Graphcore until December 2022 or January 2023 and later in February or March this year Joined Meta.
“We recently welcomed a number of highly-specialised engineers to our infrastructure team at META in Oslo. They are involved in the design and development of supercomputing systems to support AI and machine learning in META’s data centres,” said John Carville. Brings deep expertise.” Meta Spokesperson.
The move brings additional strength to the social media giant’s bid to improve how its data centers handle AI work, as it faces demand for AI-oriented infrastructure from the company’s teams building new features. runs to.
meta, owned by Facebook And Instagramhas become increasingly reliant on AI technology to target ads, select posts for its app’s feed, and remove restricted content from its platforms.
On top of that, it is now racing to join competitors like Microsoft And Alphabet‘S Google In releasing generative AI products capable of creating human-like writing, art and other content, which investors see as the next big growth area for tech companies.
Job descriptions of the 10 employees on LinkedIn indicate the team worked on AI-specific networking technology at Graphcore, which develops computer chips and systems optimized for AI work.
Carville declined to say what work he would be doing at Meta.
Graphcore closed its Oslo office as part of a wider restructuring announced last October, as it struggled to maneuver against US-based firms, a spokesman for the startup said. NVIDIA And Advanced Micro Devices Which dominates the market for AI chips.
Two sources told Reuters that Meta already has an in-house unit designing a range of chips, including a network chip, with the aim of speeding up and maximizing efficiency for its AI work. Which acts as a sort of air traffic control for the server.
Efficient networking is particularly useful for modern AI systems such as those behind chatbots. chatgpt or image creation tool dul-ewhich are too large to fit on a single computing chip and must instead be spliced together across multiple chips.
A new class of network chip has emerged to help keep data running smoothly within those computing clusters. Nvidia, AMD and intel Everyone makes such network chips.
In addition to its network chip, Meta is also designing a complex computing chip to both train AI models and make inferences, a process in which trained models make decisions and generate responses to signals, although it does not expect That’s around 2025 until the chip is ready.
Graphcore, once one of Britain’s most valuable tech startups, was seen by investors such as Microsoft and venture capital firm Sequoia as a promising potential challenger to Nvidia’s commanding lead in the market for AI chip systems.
However, it suffered a setback in 2020 when Microsoft canceled an initial deal to buy Graphcore’s chips for its Azure cloud computing platform, according to a report in UK newspaper The Times. Microsoft instead used Nvidia’s GPUs to build the large-scale infrastructure that powers ChatGPT developer OpenAI, which Microsoft also supports.
Sequoia has since reduced its investment in Graphcore to zero, according to a source familiar with the relationship, although it remains on the company’s board. The write-down was first reported by Insider in October.
A Graphcore spokesperson confirmed the setback, but said the company was “fully positioned” to take advantage of accelerating commercial adoption of AI.
© Thomson Reuters 2023