November was an eventful month, with uproar at OpenAI and Nokia’s displeasure with Intel, overshadowing Nokia’s own AI breakthrough designed to enable verbal network configuration. Contributing Editor, Annie Turner reports.
Bell Labs, part of Nokia, claims its new AI and machine learning technology will enable telcos and their customers to reconfigure networks immediately, using verbal instructions.
Apparently the Natural-Language Networking technology can act on simple text or spoken requests to allocate and assemble network resources to suit end users’ specific requirements.
Csaba Vulkan, Network Systems Automation Research Leader at Nokia Bell Labs, said, “Operators won’t need to explore technical catalogues or complex API descriptions when they configure networks. Instead, a simple statement like ‘Optimise the network at X location for Y service’ will work.
“Those requests could be used to configure a wireless network in a factory for robot automation or optimise networks at a concert for a barrage of social media uploads.”
The Natural-Language Network will learn from the instructions it receives about how best to best optimise the network. The goal is that the technology will eventually anticipate users’ needs and automatically make adjustments.
The development is part of Nokia’s UNEXT programme designed to build a self-regulating, interactive operating system, “aiming to replicate the success of their seminal UNIX system,” which is celebrating its fiftieth birthday this year.
The death of Moore’s Law?
Lightreading reports that relations are not so rosy between chipmaker Intel and Nokia. The relationship first came under strain when Intel had problems with its 10-nanometer designs which caused delivery delays five years ago. Nokia was hardest hit and didn’t have another supplier lined up for its 5G products. After a change of top management, Nokia added Marvell and Broadcom to its suppliers.
Now Nokia has decided that Intel’s general-purpose processors are not suitable for running the code that processes signals in cloud RAN, known as Layer 1, although Ericsson is sticking with the silicon for that purpose. At an update in Oulu in early November, Nokia executives reportedly complained about vendor lock-in, claimed that Moore’s Law is finished and said the company is looking at cloud RAN products that don’t involve Intel.
This is quite a moment for Intel, cloud RAN, tech industries and every other sector that relies on them. Intel has dominated the general purpose chip market for decades. But Brian Cho, Nokia’s CTO for Europe and a former Intel engineer was quoted saying, “Nowadays, with every generation of pure general-purpose computing, there is almost no efficiency improvement…Moore’s Law ended, and this opens the door for a so-called special-purpose processor.”
General purpose has its place
In Cho’s view, general-purpose chips are useful when the workloads they process are unpredictable and flexibility is needed. However, many workloads are predictable and consequently there has been a steady shift to specialist silicon in the cloud for better performance and for AI. Nvidia has been the most notable beneficiary, joining the trillion-dollar market cap club in May, the first chipco to do so.
For its part, Intel points out that its silicon is not expected to run in cloud RAN deployments much before the end of next year and say that Nokia’s implementation with Marvell is not truly cloud native, requiring proprietary software and human oversight.
Marvell and Nokia began their partnership to jointly develop silicon for 5G in March 2020. They extended it in November 2022 to collaborate on the “industry’s most advanced Radio Access and Transport Processing Platform,” based on their jointly developed ReefShark chipset and Marvell’s 5nm OCTEON 10 DPU (data processing units).
Big implications for telecoms vendors
The latest survey by TelcoTitans, TelcoX: EMEA Leadership and Performance Report, asked respondents from Europe’s largest operator groups which vendors are on target in its new TelcoX Brand Leadership Bullseye. Microsoft came out smack in the middle, most immediately surrounded by Google, Nokia, Cisco, AWS, Huawei, VMware and Ericsson. The report takes a deep dive into the ramifications of how vendors are seen/positioned in the concentric-circles schematic for operators and vendors.
Who controls the future of AI?
Microsoft has had a tumultuous few days after non-profit OpenAI fired its iconic CEO and co-founder, Sam Altman, blindsiding the software giant completely. Despite the fact that Microsoft has invested some $13 billion in the firm, that Altman is widely seen as the face of AI globally and that just days before, Altman and Satya Nadella, Microsoft’s CEO appeared on stage together as “the best partnership in tech”.
Altman was back as CEO within a week after a huge backlash and one imagines considerable pressure from investors behind the scenes. There has been widespread speculation that it is inevitable that investors would be given seats on the board and much more say in the start-up whose organisational structure reflects its focus on safeguarding and altruism, eschewing profit. In other words, that commercialism and those with deep pockets would gain control of the future of AI.
However, The Information reported that OpenAI is to have a nine-person board. At the time of writing, the first three directors of the new board were awaiting confirmation: Chair Bret Taylor, former Treasury Secretary Larry Summers and Quora’s CEO Adam D’Angelo. However, there is no intention of giving seats on the board to any of the investors, which include Microsoft, Khosla Ventures and Thrive Capital.
And although it’s not fashionable to mention it right now, OpenAI is not the only AI game in town.