Skip to Main Content

Strategic Business Insights (SBI) logo

Mobile Communications August 2018 Viewpoints

Technology Analyst: Michael Gold

Open-Source AI for Telecommunications

Why is this topic significant?

Mobile-communications players are spearheading an effort to establish standards and practices for artificial-intelligence-software development, including open marketplaces for interoperable software components.

Description

AT&T Laboratories and outsourced-services specialist Tech Mahindra created Acumos AI—a platform for developing and distributing interoperable artificial-intelligence and machine-learning software. The companies recently donated Acumos AI's source code to the Linux Foundation. The donation does not contain advanced AI. Rather, Acumos AI consists of tools that aim to streamline implementations of machine-learning software in highly automated service environments. The architects of the tool set promise enhanced compatibility, anticipating that developed software will commute across diverse platforms, including competing cloud services and private clouds. An app store distributes business-to-business AI software, and other parties can build their own public or private app stores if they wish.

Although its creators intend Acumos AI for use by organizations in general, telecommunications companies are champions of the technology. Of 14 project sponsors, half are communications services or their suppliers, including Amdocs, AT&T, Ciena, Huawei, Orange, Nokia, and ZTE. Other sponsors, including Baidu and Tencent, run large networks internally.

Implications

Several application domains could benefit from having the ability to discover and mash up AI-based software components to form complete value-added services. The initial public app store has a few examples of applications, including a face-blurring privacy filter from AT&T and an algorithm from Tech Mahindra that analyzes transaction records and creates lists of "platinum customers" and other market segments. Other use cases for reusable AI software components could emerge, including monitoring cybersecurity and responding to attacks, detecting and mitigating wireless interference, chatting with text bots for customer service, automating and orchestrating connections among network resources in accord with changing patterns of demand, and intent-based networking (see the February 2018 Viewpoints).

A major motivation for Acumos AI was its developers' needs to systematize the use of diverse kinds of computers and cloud services. Achieving interoperability of AI software and functioning services cannot be an ongoing research project. Players need production-ready workflows. Developers of service software are often distinct from teams of data scientists, who are also distinct from machine-learning specialists. Each of these teams tends to use different kinds of computers. Availability of data for training and needs to control access to private data are additional factors that affect which systems are in use, further boosting needs for diverse computers to work together in a coherent service-development environment.

Impacts/Disruptions

AI and adherence to standards might be requirements for achieving the vision of highly automated orchestration of many different software components that play roles in 5G services. A system architect from Telus recently expressed interest in using Acumos AI as a development platform for machine-learning software that balances loads and performs closed-loop control over servers, switches, and other resources in software-defined networks—for example to optimize utilization of some nodes and put other nodes to sleep.

Network operators might need AI to coordinate all of the resources involved in delivering mobile-communications services, considering the remarkable complexities of 5G architectures (see the September 2016 Viewpoints). But the technology landscape for machine-learning tools is far from mature, and diverse pathways to automation exist for cloud services, core networks, and software-defined wireless networks. For Acumos AI to see common use, its champions will need to maintain momentum, and the tool set will need to evolve to reflect ongoing changes in automation practices.

Scale of Impact

  • Low
  • Medium
  • High
The scale of impact for this topic is: Medium

Time of Impact

  • Now
  • 5 Years
  • 10 Years
  • 15 Years
The time of impact for this topic is: 5 Years to 10 Years

Opportunities in the following industry areas:

Cellular services, cloud services, software development, communications equipment manufacturing, computer manufacturing, data centers, enterprise computing, information-technology management

Relevant to the following Explorer Technology Areas:

Challenges for the Galileo Satellite Constellation

Why is this topic significant?

Smartphones are increasingly able to make use of the Galileo geolocation system, whose construction is approaching completion. But recent developments raise doubts about the satellites' commercial potential.

Description

Four satellites for the Galileo constellation launched during July 2018 and could be in use within a few weeks to a few months. Galileo then promises to have 26 operating satellites—enough to provide service that is comparable to the US GPS and Russian Glonass services. Glonass has the advantage of superior performance at high latitudes, including the Siberian region. What are Galileo's advantages? The constellation's champions have cited the potential for enhanced accuracy, value-added commercial tiers of service, opportunities for open innovation that could arise with Galileo's civilian governance, and maintenance of strategic aerospace capabilities in the European Union.

In a somewhat surprising development, a March 2018 European Commission decision rescinded previous guidance that indicated a commercial version of Galileo service would provide accuracy to within 10 centimeters. Signals that purportedly allow an accuracy of 20 centimeters will be available free, but in a November 2017 interview, European GNSS Agency General Director Carlo des Dorides told Inside GNSS magazine that "...as a free service...we are talking about 20 centimeters with a convergence time on the order of 5 minutes." Idling for 5 minutes to establish a reference position is likely acceptable in some farming, construction, and other applications that can benefit from precision geolocation, but such use cases are distinct from real-time navigation.

Implications

Formerly, Galileo champions promised significant improvements to location accuracy. Research and development plans have tentatively relied on expectations that Galileo's accuracy would help deliver new benefits in transportation, pedestrian navigation, agriculture, and other application domains. Now, stakeholders might not know Galileo's accuracy in practice until they can experiment for themselves. Definitive answers might not be available until 2020, when the constellation is scheduled to enter its "full service supply" phase.

If users cannot rely on Galileo to detect, for example, the lane of travel a car is in or the side of a street where a car is parked, future drivers might instead rely on improved augmentation of satellite signals by ground stations. Significantly, organizations in Japan plan to make increased use of ground-based augmentation for that nation's Quasi-Zenith Satellite System.

Stakeholders cannot rule out an alternative business model for Galileo, whose satellites transmit digitally signed messages that can be distinguished from signals emanating from illegal devices that emulate satellites. Researchers have demonstrated the spoofing devices, and motivated fraud artists might be using them in secret. Employers or cargo shippers might value Galileo's authenticated geolocation messages. Trustworthy location capabilities might also help ensure security of bank-card transactions, providing the bank or payment processor with high-confidence data about a card user's true location.

Impacts/Disruptions

Additional recent news items concern potential disruption to Galileo's road map arising from the United Kingdom's planned withdrawal from the European Union. At least so far, nonmember states have reduced abilities to access capabilities of Galileo that are reserved for military use. Nations could be facing a complex process of separating lines of responsibility for Galileo. Plausibly, the United Kingdom might build and operate its own geolocation satellites.

Scale of Impact

  • Low
  • Medium
  • High
The scale of impact for this topic is: Low

Time of Impact

  • Now
  • 5 Years
  • 10 Years
  • 15 Years
The time of impact for this topic is: 10 Years

Opportunities in the following industry areas:

Semiconductor manufacturing, telecommunications equipment manufacturing, electronic products manufacturing, agriculture, construction

Relevant to the following Explorer Technology Areas: