Team Credited For AI Networking at Graphcore Stolen By Meta

Growing increasingly reliant on AI technology, Meta has roped in an Oslo-based team of engineers that previously worked at Graphcore. The LinkedIn profiles of ten employees who joined the social networking giant in February or March this year indicate that they all worked at Graphcore until December 2022 or January 2023.

Meta spokesman Jon Carvill later released a statement confirming the recruitment of professionals to help develop supercomputing and AI technologies. However, he declined to share what exactly the team would be working on at Meta.

They bring deep expertise in the design and development of supercomputing systems to support AI and machine learning at scale in Meta's data centers.Jon CarvillnMeta's Race For AI Tech

The intense AI race is seeing Meta rush to catch up with Microsoft and Google to develop generative AI technologies. AI products like ChatGPT that are capable of generating human-like art, writing, and other forms of content are currently being eyed by investors as the next growth area in the tech industry. Hence, Meta's recruitment of top-tier AI engineers doesn't really come as a surprise.

The tech giant's in-house unit is currently working on developing several chips aimed at improving the speed and efficiency of AI operations.

These include a network chip that would serve an air traffic control function for servers. A complex computing chip meant to train AI models and help them generate judgments and responses is also in the works.

AI systems like Dall-E and ChatGPT are too large to fit in a single chip and need to be split across multiple chips. This calls for more efficient networking, which is exactly what Meta seems to be focused on.

Graphcore's Struggle to Capture The Chip Market

British chip unicorn Graphcore happens to be one of the most valuable tech startups in the country. Major investors like Microsoft and venture capital firm Sequoia once considered Graphcore to be a promising contender in the AI chip market, which is primarily dominated by NVidia. However, Graphcore fell behind in the competition and has been struggling to make any gains against NVidia and Advanced Micro Devices.

During a broad restructuring announced in October 2022, Graphcore also shut down its Oslo office.

In 2020, Microsoft scrapped an early deal to purchase Graphcore's chips for its cloud computing platform on Azure. Instead, Microsoft chose to use Nvidia's GPUs when building the vast AI networking infrastructure that powers OpenAI and ChatGPT.

Indeed a major setback for Graphcore - this was followed by Sequoia writing down its investment in the UK-based startup to zero.

These setbacks, along with the fact that Meta managed to scoop up the team of 10 engineers working on AI networking chips at Graphcore, clearly indicate a hard time for the startup. A Graphcore spokesperson acknowledged the difficulties but has assured that the company is ready to take advantage of the rapidly growing AI market.

So, What Comes Next?

Meta is definitely bent on developing advanced AI chip technologies, and the induction of the team from Graphcore gives the tech giant a huge boost. Meta's servers will likely be able to handle AI work much more efficiently in the near future. Teams across the company require more AI-oriented infrastructure, and Meta seems to be intent on fulfilling the same.

AI has already found extensive use in advertisement targeting, news feed post suggestions, and purging banned content on Facebook and Instagram. However, it remains to be seen if Meta can beat the competition from Google and Microsoft when it comes to generative AI.

The post Team Credited For AI Networking at Graphcore Stolen By Meta appeared first on The Tech Report.