• Virgin Media O2 is reportedly prepping a £3bn bid for CityFibre
• BT’s Openreach is to delay the launch of its Equinox 2 pricing
• AT&T tests Open RAN controller with Nokia

In today’s industry news roundup: The latest UK fibre broadband M&A speculation involves VMO2 and CityFibre; BT’s Openreach delays the dawn of its Equinox 2 pricing offer; AT&T trials Nokia’s near real-time RIC: and much more! 

Virgin Media O2 (VMO2) is considering a £3bn takeover bid for UK wholesale fibre-to-the-premises (FTTP) network operator CityFibre, according to The Telegraph (subscription required). Early discussions have already been held, according to the report, between Greg Mesch, CEO of CityFibre, and Mike Fries, CEO of Liberty Global which, along with Telefónica, jointly owns VMO2. Such a move would secure VMO2’s position as the UK’s second-largest broadband network player in the UK behind BT’s quasi-autonomous division Openreach. VMO2 had previously explored the possibility of boosting its fixed broadband position by acquiring TalkTalk, but discussions related to that deal collapsed late last year. VMO2 currently operates an extensive cable broadband network based on DOCSIS technology, which it plans to upgrade to fibre by 2028: Acquiring CityFibre would fast-track its migration to a fibre access network infrastructure. The speculation comes at an interesting time for the UK’s fibre broadband sector, as there is evidence elsewhere of potential consolidation – Trooli is looking for a buyer and has been attracting interest from VMO2 and Zzoomm, according to industry gossip – as well as signs that regulator Ofcom is spending more time thinking about network-based competition – speaking of which…

Following last week’s decision by telecom watchdog Ofcom that the Equinox 2 tranche of discounted pricing offers recently proposed by Openreach, BT’s quasi-autonomous wholesale fixed access network operator, should not be introduced as planned on 1 April as it needs further regulatory scrutiny, Openreach has decided it will voluntarily delay its planned introduction of Equinox 2 pricing. “Whilst we continue to share Ofcom’s initial view that our offer isn’t anti-competitive, it’s important the regulator has time to consider all the feedback it has received fully and fairly, so our discounted Full Fibre prices won't come into effect on 1st April,” noted the UK wholesale operator in a statement posted on its website. “That said, we don't think customers should be forced to pay higher prices because of this delay so, if Ofcom's initial assessment is confirmed in May, we'll ensure our customers benefit from Equinox 2 pricing from 1 April.” Backdated discounts, eh? How generous!!

AT&T and Nokia are boasting of an interesting Open RAN trial involving the giant Finnish vendor’s near real-time RAN intelligent controller (RIC), the emerging network element that is regarded by many as the future nerve centre or brain of open, disaggregated radio access network deployments. According to the pair, Nokia deployed its near real-time RIC in the vendor’s AirScale base stations running in AT&T’s mobile network and managed xApps over an E2 interface – E2 is the O-RAN Alliance-developed interface that connects a near real-time RIC to Open RAN network elements, including the distributed unit (DU) and centralised unit (CU). In the trial, “near real-time xApps utilised E2SM [service model] policy services to dynamically perform targeted RAN optimisation,” noted Nokia, adding that “the benefit of near real-time RIC is that it can optimise services for specific user groups, different frequency layers, or based on quality of service (QoS) class identifiers in 5G networks.” Rob Soni, VP of RAN technology at AT&T, noted: “AT&T is excited about the potential benefits offered by the O-RAN E2 interface. We are pleased to see the maturing of the Nokia Service Enablement Platform in consecutive trials over the last few years. The results of the trial demonstrated that [the] E2 interface allows RAN policies to be updated several orders of magnitude faster than through legacy OAM interfaces, and thus unlocking new methods for RAN optimisation.” Read more. 

BT has unveiled a multimillion-pound investment to enable “high-speed, latency-sensitive and intensive 5G connectivity” for its UK business customers and the public sector looking to deploy high-bandwidth internet of things (IoT) technology. The UK operator has partnered with Amazon Web Services (AWS) for the move, which will see the hyperscaler shift closer to the network edge when its cloud services are combined with BT’s 5G and 4G infrastructure. EE’s, BT's mobile network division, national mobile network will be united with AWS Wavelength – an AWS infrastructure which is designed to run workloads requiring ultra-low latency over mobile networks. BT claimed the collaboration will open doors to “faster, secure and high-bandwidth connectivity on the move for use cases like policing, crowd management, healthcare and security.” The move is part of the telco’s investment in its existing mobile networks and aims to enable “5G-connected infrastructure-as-a-service” which will be trialled within a new AWS Wavelength Zone in Manchester before being deployed nationally. The end goal is for BT to roll out AWS Wavelength to UK businesses “more broadly” in the coming years, with potential IoT use cases including autonomous vehicles, cameras for policing, live media production and smart industrial robots. Find out more.

MTN has carried out a proof of concept (PoC) for what it claims is the first of its kind 5G standalone (SA) core deployment in Microsoft Azure. Within the demonstration, all 5G SA core elements, including control plane, user plane and management nodes, have been fully deployed in the Azure datacentre ‘region’ (cluster) in South Africa. According to MTN, this has enabled it to benefit from faster deployment time (a matter of days rather than months) and ease of scale. It plans to put its findings to good use in its network planning in terms of sustainability, commercial models, automation and disaster recovery. The PoC will also serve as “a stepping stone to drive further innovation in the cloud domain” as part of the African operator’s strategy until 2025, which will focus on driving “industry-leading connectivity operations through technology platforms that are second to none”. Read more.

It had to happen somewhere of course. Daktela, a cloud-based comms and call centre solutions company headquartered in Prague in the Czech Republic, has announced that it is the first provider of call centre-as-a-service (CCaaS) in the UK to integrate ChatGPT directly into its platform. Hitherto the company exploited some of the properties of AI in the functioning and communication of smart mailbots, voicebots and chatbots. Now though, Daktela’s goal is to create “the first real AI omnichannel, in which ChatGPT will become a part of the entire product portfolio” and provide contact centre agents with more efficient handling of incoming requests and faster communication with customers. In essence, the integrated ChatGPT bit of the solution will summarise the full communication thread with the customer so that agents don’t have to wade through reams of previous communications with the caller. Daktela also claims that the software will improve the speed and production of written communication. Thus, going forward, the human agent will only have to write a short bullet list with the main points of the conversation, which is then turned into ‘a complete prescribed e-mail with all formal and communication requirements via the “turbo version’ of the software, GPT-3.5, which allows access to other developers via its API interface. Simultaneously, an agent can mark any sentence and have it rewritten in a variety of styles, for example, “to sound more formal or more friendly.” Agents will also have to keep an eagle eye out for glitches. First off, although large language models might seem to ‘know’ a great deal, the problem with them is they can’t 'remember’ much for long. Secondly, on many, many recorded occasions (and many more that haven't been made public), ChatGPT has got things wrong because its responses are based on patterns rather than facts and data, and the software has only limited awareness of context. The result can be the generation of nonsensical, rude, inappropriate, incorrect and downright barking mad answers. When in use in call centres and it goes haywire, as it will, it’ll be a steep learning curve for agent and AI alike.

And sticking with the theme, it’s China versus the US again (though not everyone sees it that way)… The much-trumpeted debut of the Ernie AI bot developed by China’s Baidu, and the suggestion it could be a serious and superior rival to ChatGPT, developed by OpenAI from the US, was an underwhelming affair. After promising that Ernie, short for ‘Enhanced Representation through Knowledge Integration’, would perform its magic live at a Beijing inauguration, Ernie failed to do anything of the sort. Instead, reports Ars Technica, Robin Li Yanhong, Baidu’s co-founder and CEO, fronted an over-managed, pre-recorded demonstration of Ernie’s alleged and apparent capabilities. It is claimed that a live stream of the event was watched by 2 million people. That figure is unverifiable but investors who had been invited to watch in awe and admiration before delving deep into their pockets to stump up cash for the further development of the chatbot kept their wallets as tightly zipped as their pursed lips. Live-streaming a pre-recorded tech demonstration is not the same as presenting it as a live interaction, and attendees were not impressed by a few pre-prepared examples of Ernie solving a few algebraic equations and answering questions about a trilogy of Chinese sci-fi novels. Furthermore, it was evident from the heavily edited demo that Ernie cannot as yet hold a candle to its western competitors and will need some serious development if it’s ever to be regarded as a viable rival to the likes of ChatGPT. And for that to happen, Ernie’s users will need access to the global internet, not just the partial, partitioned, censored and over-policed and surveilled subset version that the Chinese authorities allow the country’s masses to see, and such unfettered freedom is just not going to happen. So while Wenxin Yiyan (as Ernie is called in China) may well be a big player in its home market, it will be an ‘also-ran’ irrelevancy in much of the rest of the world. Reaction to the demonstration was swift – Baidu’s share price fell by 10% after the curtain came down on the demo. Wenxin Yiyan/Ernie was due to be launched as a commercial service before the end of this month, initially as a standalone app that, later this year, would be consolidated into a service on the Baidu search engine itself. That timetable is now unlikely to be met. It’s hard to avoid the conclusion that Baidu, taken by surprise by the release of ChatGPT, painted itself into a corner by overpromising what Ernie is able to do and then, scared of losing face by postponing the promised live demonstration, went ahead and exhibited a turkey. To make matters worse it has been completely blindsided by the unexpected recent launch of the even more powerful GPT-4. Because of that, Baidu CEO Robin Li had to admit that Ernie’s technology is a work in progress, but then insisted, “Our expectations for Ernie Bot are close to ChatGPT, even GPT-4.” Great expectations are all very well, but as young Philip Pirrip (Pip) found in the Charles Dickens novel, they may not be fulfilled in the ways the protagonists thought they might be. However, the CEO did add, “Ernie Bot is not a tool of confrontation between China and the United States.” Good to know, given that everything else technological seems to be….

Fraudulent robocalls are expected to cost mobile subscribers a whopping $58bn globally in 2023, up from a $53bn loss in 2022, a new Juniper Research report has estimated. The increased losses are expected to be driven by various scam calls, including unauthorised call forwarding and caller ID spoofing. According to the research, innovative fraud methods will keep driving a rise in losses for end users, reaching $70bn globally by 2027. The picture is most gloomy in North America, which this year is expected to account for more than half of all losses incurred by deceiving robocalls. By 2025, however, such losses are predicted to decline for the first time in the region due to increased adoption of robocalling mitigation frameworks, such as Stir/Shaken. Beyond North America, Juniper Research is calling for stakeholders to develop frameworks that are designed to combat region-specific methods of fraud.

Another day, another startling and potentially massively important development in quantum computing. Scientists at the University of Sydney, Australia, working with a team of others at the University of Basel, Switzerland, have shown it is possible to manipulate ‘quantum light’ and manoeuvre and identify small numbers of interacting photons, the elementary particles and quanta of light that are the smallest possible ‘packets’ of electromagnetic energy and radiation. Photons do not have mass and therefore always travel at the speed of light in vacuum. Dr Sahand Mahmoodian, of the University of Sydney School of Physics and joint lead author of the research, said, “This fundamental science opens the pathway for advances in quantum-enhanced measurement techniques [for advanced medical imaging among other applications] and photonic quantum computing.” Dr Natasha Tomm of the University of Basel, the other joint author of the paper added, “The device we built induced such strong interactions between photons that we were able to observe the difference between one photon interacting with it compared to two. We observed that one photon was delayed by a longer time compared to two photons. With this really strong photon-photon interaction, the two photons become entangled in the form of what is called a two-photon bound state.” A new paper, “Photon bound state dynamics from a single artificial atom”, published today in the latest edition of Nature Physics, the respected monthly peer-reviewed scientific journal, explains the experiment in detail. It reports “the direct observations of a photon-number-dependent time delay in the scattering of a single artificial atom – a semiconductor quantum dot coupled to an optical cavity.” Normally photons do not easily interact with one another which is an attribute that makes possible almost distortion-free transfer of data at close to light speed and thus of immense value in telecommunications. However, while having photons interact is valuable, it is a tricky process. As Dr Mahmoodian stressed, “By demonstrating that we can identify and manipulate photon-bound states, we have taken a vital first step towards harnessing quantum light for practical use. The next steps… are to see how this approach can be used to generate states of light that are useful for fault-tolerant quantum computing. Dr Tomm agreed: “This experiment is beautiful, not only because it validates a fundamental effect – stimulated emission – at its ultimate limit but it also represents a huge technological step towards advanced applications. We can apply the same principles to develop more-efficient devices that give us photon-bound states. This is very promising for applications in a wide range of areas: From biology to advanced manufacturing and quantum information processing.”

- The staff, TelecomTV