...
Andrei Agapi Andrei Agapi (Deutsche Telekom)
Ranny Haiby Ranny Haiby (Linux Foundation Networking)
...
At their core, telecoms are technology companies driven by the need to scale their networks to service millions of users, reliably, transparently and efficiently. To achieve these ambitious goals, they need to optimize their networks by incorporating the latest technologies to feed the connected world's insatiable appetite for ever more bandwidth. To do this efficiently, the networks themselves need to become more intelligent. At the end of 2021, a bit over 2.5 years ago, LFN published its first white paper on the state of intelligent networking in the telecom industry. Based on a survey of over 70 of its telecom community members, the findings pointed to a still nascent field made up of mostly research projects and lab experiments, with a few operational deployments related to automation and faster ticket resolutions. The survey did highlight the keen interest its respondents had in intelligent networking, machine learning and its promise for the future of the telecom industry in general.
Fast forward to 2023 when at the request of the LFN Governing Board and Strategic Planning Committee, the LFN AI Taskforce was created to coordinate and focus the efforts that were already starting to bubble up in both new project initiatives (Nephio, Network Super Blueprints, just to name a few) and within existing projects (Anuket, ONAP). The Taskforce was given the charter by the Governing Board to look at and make recommendations on what direction LFN should gotake with this exciting emerging field of research and technology. Some of the areas that the Taskforce looked into included:
...
Ironically, as generative AI and LLM adoption becomes more widespread in many industries, telecom has lagged somewhat due to a number of valid factors. As was covered in the previous white paper, the overall industry challenges remain the same, that is the constant pressure to increase the efficiency and capacity of operators’ infrastructures to delivery more services to customers for lower operational costs. The complexity and lack of a common standard understanding of network traffic data remains a barrier for the industry to speed the adoption of AI/ML to optimize network service delivery. Some of the challenges that are motivating continued research and adoption of intelligent networking in the industry include:
...
Deploying large models such as LLMs in production, especially at scale, also raises several other issues in terms of: 1) Performance, scalability and cost of inference, especially when using large context windows (most transformers scale poorly with context size); , 2) Deployment of models in the cloud, on premise, multi-cloud, or hybrid; , 3) Issues pertaining to privacy and security of the data for each particular application; , 4) Issues common to many other ML/AI applications, such as ML-Ops, continuous model validation and the need for continuous and costly re-training.
- Need for High Quality Structured Data: Telecommunications networks are very different from other AI human-computer data sets, in that a large number of interactions between systems use structured data. However, due to the complexity of network systems and differences in vendor implementations, the degree of standardization of these structured data is currently very low, causing "information islands" that cannot be uniformly "interpreted" across the systems. There is a lack of "standard bridge" to establish correlation between them, and it cannot be used as effective input to incubate modern data-driven AI/ML applications.
- AI Trustworthiness: In order to meet carrier-level reliability requirements, network operation management needs to be strict, precise, and prudent. Although operators have introduced various automation methods in the process of building autonomous networks, organizations with real people are still responsible for ensuring the quality of communication and network services. In other words, AI is still assisting people and is not advanced enough to replace people. Not only because the loss caused by the AI algorithm itself cannot be defined as the responsible party (the developer or the user of the algorithm), but also because the deep learning models based on the AI/ML algorithms themselves are based on mathematical statistical characteristics, resulting in behavior uncertainty leading to sometimes erratic behavior, meaning the credibility of the results being can be difficult to determine.
- Uneconomic Margin Costs: According to the analysis and research in the Intelligent Networking, AI and Machine Learning While Paper , there are a large number of potential network AI technology application scenarios, but to independently build a customized data-driven AI/ML model for each specific scenario, is uneconomical and hence unsustainable, both for research and operations. Determining how to build an effective business model between basic computing power providers, general AI/ML capability providers and algorithm application consumers is an essential prerequisite for its effective application in the field.
- Unsupportable Research Models: Compared with traditional data-driven dedicated AI/ML models in specific scenarios, the R&D research and operation of large language models have higher requirements for pre-training data scales, training computing power cluster scale, fine-tuning engineering and other resource requirements, energy consumption, management and maintenance, etc. Is it It is difficult to determine if it is possible to build a shared research and development application, operation and maintenance model for the telecommunications industry so that it can become a stepping stone for operators to realize high-end useful autonomous networks.
- Contextual data sets: Another hurtle that is often overlooked is the need for the networking data sets to be understood in context. What that means is that networks need to work with all the layers of the IT stack, including but not limited to:
- Applications: Making sure that customer applications perform as expected withe underlying network
- Security: More important than ever as attack vector expand and customers expect the networks to be protected
- Interoperability: The data sets must support transparent interoperability with other operators, cloud providers and other systems in the telecom ecosystem
- OSS/BSS Systems: The operational and business applications that support network services
- Community Unity and Standards: Although equipment manufacturers can provide many domain AI solutions for professional networks/single-point equipment, these solutions are limited in "field of view" and cannot solve problems that require a "global view" such as end-to-end service quality assurance and rapid response to faults. . Operators can need to aggregate management and maintenance data in various network domains by building a unified data sharing platform, and based on this, further provide a unified computing resource pool, basic AI algorithms and inference platform (i.e. cross-domain AI platform) for various scenario-specific AI for , end-to-end scenarios, and intra-domain scenarios . Applied using an applied reasoning platform.
Emerging Opportunities
...
The Telecoms have been working on converged infrastructure for a while. Voice over IP has long been an industry standard, but there is far more than can be done to drive even more efficiencies in with network and infrastructure convergence.
Converged infrastructure is needed to support the growth and sustainability of AI. This is particularly important in the need for a single solution designed to integrate compute, storage, networking, and virtualization. Data volumes are trending to grow beyond hyperscale data, and with such massive data processing requirements the ability to execute is critical. The demands on existing infrastructure are already heavy., so bringing everything together to work in concert will be key, to maintain and grow the demand for resources. In order to do that the components will need to work together efficiently and the network will play an important role in linking specialized hardware accelerators like GPUs (Graphics Processing Units) or TPUs (Tensor Processing Units) to accelerate AI workloads beyond current capability. Converged infrastructure solutions will lead to the ability to deploy AI models faster, iterate them more efficiently, and extract insights with greater speedfaster. This can pave the way for the next generation of AI.
...
Data Privacy and Security: The ability to monetization the data comes with a big caveat, which is that the use of customer data must be handles with care to ensure data privacy, security and regulatory compliance. This requires clear policies and security procedures to ensure anonymity, safety and privacy at all times. The good news is that AI can be used to address the growing threat of network vulnerabilities, zero day exploits and other security related issues with predictive analytics and threat analysis.
Complex Data Model SimplificationSimplification of complex data models: Large language models can be used to understand large amounts of unstructured operation and maintenance data (for example, system logs, operation and maintenance work orders, operation guides, company documents, etc., which are traditionally used in human-computer interaction or human-to-human collaboration scenarios), from which effective knowledge is extracted to provide guidance for further automatic/intelligent operation and maintenance, thereby effectively expanding the scope of the application of autonomous mechanism.
...
This breakthrough in AI research is characterized by vast amounts of easily accessible data, extensive training models, and the ability to quickly generate human-like text. These models are trained on enormous datasets from sources particular to the a given area of research. LLMs have changed the way natural language processing tasks are interpreted. Some of the areas that have been particularly fruitful include: text generation, language translation, summarization, and automated question chatbots and routine query responses.
Generative AI (Gen AI)
Gen AI is a much broader category of artificial intelligence systems capable of generating new content, ideas, or solutions autonomously based on a human text, video, image or sound based input. This includes LLMs as resource data for content generation. As such, Gen AI seems to be able to produce human-like creative content in a fraction of the time. Content creation for web sites, images, videos, and music are a few of the capabilities of Gen AI. The rise of Gen AI provides has inspired numerous opportunities for business cases, from creating corporate logos, to corporate videos, to saleable products to end-consumers and businesses, to creating visual network maps on the basis of the datasets being accessed. Further, even being able to provide optimized maps for implementation to improve networking either autonomous or with human intervention are useful areas for further exploration.
The two combined open the question as to what should Gen AI be used for, and more importantly how is it made distinguishable from human work. There are many regulatory bodies looking at solutions around identification of decisions and what content has been generated to solve a particular problem and solution. The foundation of this combination is to ensure security, safety, mitigate biases, and identify what changes which behaviors were illustrated and acted upon by Gen AI, and what changes which were not. Gen AI requires an organizational framework for each organization to know and ensure these factors.
The advent of transformer models and attention mechanisms [1][2] and the sudden popularity of ChatGPT and other LLMs, transfer learning and foundation models in the NLP domain have all sparked vivid discussions and efforts to apply generative models in many other domains. Interestingly, let us not forget that word embeddings [3][4][5], sequence models such as LSTMs [6] and GRUs [7], attention mechanisms [8], transformer models [1] and pretrained LLMs [2] have been around long before the launch of the ChatGPT tool in late 2022. Pretrained transformers like BERT [2] in particular (especially transformer-encoder models) were widely used in NLP for tasks like sentiment analysis, text classification [9], extractive question answering [10] etc. long before ChatGPT made chatbots and decoder-based generative models go viral. That said, there has clearly been a spectacular explosion of academic research, commercial activity and ecosystems that have emerged since ChatGPT came outwas launched to the public, in the area of both open [11][12] and closed source [13][14][15] LLM foundation models, related software, services and training datasets.
...
A natural question arises on how the the power of LLMs can be harnessed for problems and applications related to Intelligent Networking, network automation and for operating and optimizing telecommunication networks in general, at any level of the network stack. Datasets Datasets in telco-related applications have a few particularities unique to the industry. For one, the data one might encounter ranges from fully structured (e.g. code, scripts, configuration, or time series KPIs), to semi-structured (syslogs, design templates etc.), to unstructured data (design documents and specifications, Wikis, Github issues, emails, chatbot conversations).
...
For finetuning LLMs, unlike for regular neural network models, several specialized techniques exist in the general area of PEFT (Parameter Efficient Fine Tuning), allowing one to only finetune fine tune a very small percentage of the many billions of parameters of a typical LLM. In general, the best techniques to achieve domain adaptation for an LLM will heavily depend depends on:
- The type of data and how much domain data is available,
- The specific downstream task
- The initial foundation model
...
In conclusion, while foundation models and transfer learning have been shown to work very well on general human language when pretraining is done on large corpuses of human text (such as Wikipedia, or the Pile [29]), it remains an open question to be answered whether domain adaptation and downstream task adaptation work equally well on the kinds of domain-specific, semi-structured, mixed modality datasets we can find in the telecom networksindustry. To enable this, telecoms should focus on standardization and data governance efforts, such as standardized and unified data collection policies and developing high quality structured data across a common definition and understanding.
Projects and Research
1.5 page
3GPP Intelligent Radio Access Network
The intelligent evolution of wireless access networks is currently in a phase of rapid evolution and continuous innovation. In June 2022, 3GPP announced the freezing of R17 and described the process diagram of an intelligent RAN in TR37.817, including data collection, model training, model inference, and execution modules, which together form the infrastructure of an intelligent RAN. This promotes the rapid implementation and deployment of 5G RAN intelligence and provides support for intelligent scenarios such as energy saving, load balancing, and mobility optimization.
...
However, despite the remarkable progress made in the 5G wireless access network intelligence industry, some challenges and issues remain to be addressed. For example, network security and data privacy protection are pressing issues that require effective measures to be implemented. The energy consumption issue of 5G networks also needs attention, necessitating technological innovations and energy-saving measures to reduce energy consumption. In the future, continuous efforts should be made in technological innovation, market application, and other aspects to promote the sustainable and healthy development of the 5G wireless access network intelligence industry.
Core Network Transformation
Mobile core network is kind of central or brain networks can be thought of as the brains of mobile communication, it is the only and largest part which experienced the . In recent years, these networks have experienced a huge transformation from legacy proprietary hardware into telco to telecom cloud native systems. Today, almost 100% of today mobile core network has been networks are deployed based on telco cloud architecture architectures supported by NFV technologies. It mostly consisted of packet core network such as 5GC and UPF which is responsible for packet forwarding, IMS which help operators multimedia communications such as voice, message and video communication of telecommunication operators, and the management of core network including telco cloud infrastructure and 5G network functionalities. F.or those 3 parts of AI evolvement,
...