Integrating artificial intelligence in Ericsson rApps
In the ever-evolving landscape of telecommunications, the integration of artificial intelligence (AI) into network management systems represents a pivotal shift toward automation and enhanced efficiency. However, as we inch closer to this new era, the importance of trust in AI systems becomes increasingly paramount.
Portfolio Director, AI/ML Innovation, Cognitive Network Solutions
Portfolio Director, AI/ML Innovation, Cognitive Network Solutions
Portfolio Director, AI/ML Innovation, Cognitive Network Solutions
Trustworthy AI, defined by its transparency, accountability, and reliability, is not just a futuristic concept but a present necessity that communication service providers (CSPs) are actively seeking.
In the next generation telecommunication networks, the convergence between trustworthy AI and service management and orchestration (SMO) in the open radio access network (O-RAN) architecture will play a pivotal role in the development of intelligent RAN automation networks. As defined by the O-RAN Alliance, non real time radio intelligent controller applications (rApps) will be the fundamental building block of the service management and orchestration platform.
This blog post explores the importance of trustworthy AI and highlights how Ericsson is spearheading advancements in this domain, shaping the industry by incorporating explainable AI methodologies in the rApps portfolio.
The imperative of trust in AI
Trust in AI is more than just a buzzword; it is a critical factor that significantly influences the adoption and success of AI-driven solutions, especially in the telecommunications sector. According to recent market research by OMDIA in Q2 2024, trust in AI, primarily driven by explainability, is vital for CSPs contemplating automation and closed-loop assurance. With the telecom industry on the cusp of a transformation, the demand for trustworthy AI is surging, and rightly so. Trustworthy AI ensures that the decisions made by AI systems are understandable, justifiable, and free from biases, thus fostering confidence among users.
In the case of rApps, AI explainability methodologies are even more important. The SMO ecosystem will bring a new level of scalability and openness, enabling multiple and heterogeneous non-rt RIC applications to coexist and interact with each other. In this scenario, building trust based on the transparency and understanding of the algorithms and insights provided will be a needle mover for the successful adoption of network automation technologies and the catalyst for adopting intelligent RAN automation technologies.
Ericsson's vision: From concept to real-world implementation.
Ericsson has long been at the forefront of AI innovation, committed to developing technologies that not only push the boundaries of what is possible but also ensure reliability and transparency. At the heart of our rApps portfolio transformation, there are two enabling technologies: Bayesian neural networks (BNNs) and explainability in graph neural networks (GNNs).
-
Bayesian Neural Networks: Transparency and confidence
BNNs offer a novel approach to understanding the uncertainty behind AI-driven decisions. Unlike traditional "black box" AI models, BNNs provide insights into the confidence levels of AI predictions. This shift from opacity to transparency allows CSPs to audit network optimization and diagnostic processes, thereby enhancing trust. With BNNs, customers gain higher visibility into how AI systems operate, resulting in increased confidence and adoption of AI recommendations. -
Explainability in graph neural networks: Bridging complexity and clarity
GNNs are renowned for their accuracy in simulating complex cellular networks. However, their intricate nature has often posed challenges in explainability. Ericsson's Cognitive Network Solutions team has pioneered a method to assess the effectiveness of explainability methodologies in GNNs, a technique currently under IPR review. This innovation drastically reduces the time required for deploying explainability processes — by up to 99 percent — while ensuring scalability in production environments. By elucidating the importance of cell-neighbor relations, this method boosts confidence in AI-driven recommendations.
Patents, publications, and open-source contributions: A testament to innovation
Ericsson's commitment to pioneering trustworthy AI is reflected in several significant milestones when it comes to our rApps portfolio realization:
- Patents: Among others, Ericsson’s Cognitive Network Solutions has recently filed one more patent related to trustworthy AI. Two additional patents are under review, with two more in the writing process. These patents symbolize the tangible advancements Ericsson is making in AI technologies.
- Research publications: In July 2024, Ericsson’s Cognitive Network Solutions published a paper on trustworthy AI at the prestigious 2nd World Conference on eXplainable AI), showcasing their groundbreaking work to the broader AI and telecom communities.
- Open-source contributions: Ericsson’s Cognitive Network Solutions team developed an open-source library nominated for the BCSS Tech leadership award in Q2 2024. This recognition underscores the significance of trustworthy AI in driving industry leadership.
Real-world impact: Where innovation meets application
The technologies developed as part of Ericsson's trustworthy AI initiatives are already making substantial impacts. Ericsson's Cognitive Software, incorporating these explainable AI technologies, is deployed across 11 million cells in 120 commercial contracts with leading service providers. These deployments showcase the tangible benefits of trustworthy AI:
- Boosting AI adoption: Transparent AI processes build trust with CSPs, making them more likely to adopt AI-driven automation.
- Increased confidence: CSPs can confidently execute closed-loop changes without human intervention, thanks to the transparency of AI processes.
- Efficiency and cost savings: Automated AI-driven processes reduce the complexity and cost of network management, leading to significant savings.
- Lower development costs: Enhanced visibility into AI algorithms reduces development costs and streamlines innovation.
Ericsson rApps include specific use cases, such as the root cause explainers and the optimizers, which will provide users with all the information needed to understand how the AI algorithm arrives at certain recommendations.
The path forward: Building trust through explainability
As CSPs continue their journey towards full automation, trustworthy AI remains a crucial enabler. The adoption of AI in network management hinges on the trust built through transparency and explainability. By integrating explainability and Bayesian methodologies into AI models, Ericsson is helping the telecom industry transition from reactive to proactive network management.
This is just the beginning. Ericsson's work on trustworthy AI will continue to evolve, playing a key role in shaping the future of intelligent networks. As the industry progresses, the focus will remain on making AI-driven solutions more transparent, trustworthy, and efficient. The ultimate goal is to build a telecommunications ecosystem where AI is not only a tool for automation but also a trusted partner in driving innovation and excellence in rApps towards the realization of the intelligent RAN automation vision.
In conclusion, the significance of trustworthy AI cannot be overstated, especially toward O-RAN realization for the Intelligent RAN Automation transformation journey. It is the foundation upon which the future of intelligent networks will be built. Through innovations in BNNs and explainability in GNNs, Ericsson is leading the way towards a more transparent and reliable AI-driven future for the rApps portfolio. The journey towards trustworthy AI is a testament to Ericsson's commitment to pushing the boundaries of what is possible while ensuring that trust remains at the core of every technological advancement.
RELATED CONTENT
Like what you’re reading? Please sign up for email updates on your favorite topics.
Subscribe nowAt the Ericsson Blog, we provide insight to make complex ideas on technology, innovation and business simple.