• This year’s Mobile World Congress (MWC) was all about AI across the hallways, presentation stages and meeting rooms
• We asked industry analysts to select the most notable AI innovations they saw at the Barcelona-based show 
• While AI is more than just hype, telcos will probably need to wait before realising its full potential

Mobile World Congress (MWC) 2024 was all about shiny innovations and new industry developments, and while AI undoubtedly dominated the show floor, some analysts suggest it will take a while before its full potential is unleashed.

At this year’s show in Barcelona, the AI theme was everywhere, from show floor demonstrations, keynote presentations and panel discussions to formal (and not so formal) meetings both at the show and in the city during the many evening events and dinners.  

“The story of the show floor was AI... and then [there was] everything else,” Appledore Research consulting analyst Robert Curran told TelecomTV. He added that just because there is so much hype around it doesn’t mean there’s no value, “just that you have to look harder to find it.”

So what caught Curran’s eye at the Fira? He highlighted a number of conversational interfaces with networks presented by the likes of Rakuten, Amdocs and Huawei.

The solutions demonstrated how AI can be used in a network operations centre, for example to extract useful information from a “much more complex data set” or change the configuration to improve the energy efficiency or the quality of service, rather than operating at the command line level or looking at multiple underlying analytic systems.

The analyst suggested that this type of technology is still only at the exploratory stage, but the technology is there and now it is up to operators to figure out how they will make use of it. “This isn’t slideware – there is real product here, and the rest of 2024 will see operators working to identify quick-win opportunities for delivering business value from the use of AI,” Curran explained.

He also suggested that telcos can transform their operations by using AI tools, but in order to get there, they need to look at what they can actually do today, where the problems are and work from the business problem down to the technology. 

And because AI, and especially generative AI (GenAI), requires quite a significant investment, Curran believes that telcos are aiming to discover “the easy places that we can see obvious benefit.” He outlined the customer service arena as one of the main areas that stands to gain from AI in the near term because of the expected efficiency but, aside from that, the analyst expects telcos to need more time to find out the most valuable ways for the technology to be used. However, he added, there is no doubt that “the potential is huge”.

Aside from conversational interfaces with networks, AI was powering a plethora of other showcases at MWC.

In a blog, PP Foresight founder and analyst Paolo Pescatore highlighted several demos that captured his attention at the event. A special mention went to a product called AI Pin, a wearable device developed by Humane and showcased at SK Telecom’s booth throughout the event, which interacts with large language models (LLMs) to offer answers to verbal requests and which can project visual responses onto the palm of a user’s hand (a video demonstration of the AI Pin can be seen in our Spotlight on 5G report from the show floor). “It represents a stepping stone towards how we will interact and engage with technology and wearable devices in the future”, wrote Pescatore of the showcase, which could be seen at SK Telecom’s booth throughout the event.

Another similar demo singled out by the PP Foresight analyst was Deutsche Telekom’s prototype T Phone, a device developed in collaboration with Qualcomm and Brain.ai that shows how voice can be used to interact with a mobile device without the need to use any on-device apps.

That particular showcase was also highlighted by several analysts at CCS Insight, who jointly wrote in a blog that this demonstration “gave an experience that felt as if the phone knew what users wanted to do before they completed a task.” They added that it uses a framework called a “large action model”, which differs from LLM as it provides “a more intuitive experience, seamlessly linking apps or other content on the device and presenting it in an accessible way. Crucially, this should result in a single interface for the user, rather than requiring them to switch between different apps and services.”

The team at CCS Insight believe the T Phone will help the company differentiate its offering from other operators.

Yet another AI-powered highlight, according to Pescatore, was a large language and vision assistant (LLaVA) demonstration from Qualcomm – a solution that “can accept multiple types of data inputs, including text and images, and generate multi-turn conversations with an AI assistant about an image.”

This demonstration was part of Qualcomm’s broader launch of its AI Hub, which provides more than 75 models designed to enable developers to quickly produce applications supporting on-device AI.

“This isn’t a surprising move from Qualcomm but is a critically important step to speed application development that can fully harness the features of its flagship chipset platforms. Qualcomm and others are promoting the potential for on-device AI to drive a significant replacement cycle in smartphones and PCs, but few mass-market applications available today offer the features promised by the introduction of dedicated neural processing units (NPUs). Addressing this shortfall is critical to delivering the touted benefits of on-device AI and ensuring that reality starts to catch up with industry hype,” the analysts at CCS Insight noted. 

They added that Qualcomm’s launch is a necessary step to a new range of “rich experiences enabled by AI running locally on devices.”

The CCS Insight team also mentioned the new AI-RAN Alliance, which was announced on the first day of the event – see AI-RAN Alliance launches at #MWC24. The goal of this new industry body is to use AI to improve radio access network (RAN) performance, address new revenue opportunities and help the shift towards shared edge platforms that can host and manage AI and RAN workloads.  

“Enhancing spectral efficiency is a natural extension of existing energy efficiency and 6G air interface work. Offering new AI-enabled applications represents an acceleration of edge content caching and edge applications, but now AI-focused. The most interesting focus is on asset usage, harnessing the fact that AI and RAN share the same infrastructure. Here, the alliance may seek to push AI inference out onto the edge rather than running in central cloud datacentres”, the CCS Insight analysts noted in their blog.

Although, they cautioned, using compute and storage resources capable of AI processing in the RAN will increase costs for telcos. “To persuade operators to buy AI hardware for the RAN, suppliers will need to show that the potential revenue from this will outweigh the additional electricity cost of running these loads, as well as the increased wear and tear on RAN hardware,” the analysts suggested.

They noted that the AI-RAN Alliance has “a good chance to succeed” if it can add to its membership more leading operators and a broader range of suppliers.

The main takeaway from this year’s MWC, according to various analysts, is that AI is not just a fad and that its impact will be transformative. However, as Appledore Research principal analyst Patrick Kelly told TelecomTV, “It will take many more years for its full potential to be realised.”

AI was also a central theme in many of the top-level telco executive interviews that TelecomTV conducted during MWC24 – check out those interviews, our daily video coverage (The Slice) and much more from Barcelona right here.  

- Yanitsa Boyadzhieva, Deputy Editor, TelecomTV