Skip navigation
Like what you’re reading?

Applied Generative AI for Enterprise

Available in English Français

The advancement in Generative AI during late 2022 has opened up a myriad of new possibilities for harnessing the power of AI. The hype surrounding this breakthrough has revealed an abundance of novel use cases and an overwhelming demand to embrace Generative AI.

Innovation Manager

Generative AI lead for software technology

Head of AI Business Enablement, Enterprise Automation & AI

Global Head of AI, Quantum and Blockchain Execution

VP and Head of Automation & AI

Intelligent automation

Innovation Manager

Generative AI lead for software technology

Head of AI Business Enablement, Enterprise Automation & AI

Global Head of AI, Quantum and Blockchain Execution

VP and Head of Automation & AI

Innovation Manager

Contributor (+4)

Generative AI lead for software technology

Head of AI Business Enablement, Enterprise Automation & AI

Global Head of AI, Quantum and Blockchain Execution

VP and Head of Automation & AI

With the advent of Generative AI (GenAI), a remarkable array of capabilities has emerged, encompassing the generation of diverse digital content, including text, code, images, simulations, 3D objects, music, video, and potentially even more yet-to-be-unveiled creative outputs.

The foundation of GenAI lies in the transformer model, a deep learning architecture that excels in comprehending context and meaning by tracking relationships within sequential data, such as the words in this very sentence. Unlike conventional neural networks, the transformer employs a self-attention mechanism, allowing it to focus on different parts of the input sequence at each step. This remarkable feature enables the model to capture intricate relationships between elements within the sequence and process lengthy sequences efficiently.

Initially designed for language translation, the Transformer's standout attribute lies in its unparalleled adaptability to diverse tasks. Pretrained Transformer models can swiftly and effortlessly adjust to novel tasks for which they haven't been explicitly trained. Consequently, as a machine learning practitioner, the burden of training large models on extensive datasets diminishes. Instead, one can conveniently repurpose the pretrained model for their specific task, perhaps with only minor adjustments and a significantly smaller dataset.

Generative AI functional architecture

Generative AI functional architecture

 

The present generation of Large Language Models (LLMs) has exhibited human-level proficiency in tasks encompassing text generation, question answering, and language translation. Notably, among the distinguished examples of LLMs are GPT-4 developed by OpenAI, which proficiently generates human-like text responses across a wide array of prompts. Additionally, the BERT model by Google is tailored to comprehend the contextual nuances of words and phrases within sentences.

The domain of GenAI is experiencing rapid expansion, marked by substantial contributions from key industry players, such as Google, Microsoft, OpenAI, and Anthropic.

LLMs have found widespread utility in diverse applications, including chatbots, virtual assistants, and other systems that necessitate computer-generated natural language responses. Their widespread adoption in these contexts attests to their capability in comprehending and generating human-like language outputs, thereby enhancing user experiences and interactions with computer systems.

 

Implementing GenAI in Ericsson

As one of the biggest technology breakthroughs, GenAI emerges as a remarkably versatile technology with the potential to revolutionize various aspects of any organization, provided it is employed responsibly. The rapidly evolving landscape of GenAI is continually unveiling new possibilities for its application, promising significant advancements in the future.

Within Ericsson, our focus in the Enterprise Automation and AI team centers on harnessing the potential of GenAI within our internal operations. To promote its widespread adoption, three overarching use cases have been identified as initial drivers for integrating GenAI internally. These use cases are the first step to successful incorporate of GenAI throughout the organization.

Intelligent assistant

The first use case successfully implemented within our organization is an intelligent assistant, a highly adaptable tool featuring a chat-based conversational interface coupled with a generative language model backend. Initially leveraging the GPT model developed by OpenAI, we continually explore the integration of other pre-trained models or fine-tuned models using our proprietary data to cater to specific use cases. This intelligent assistant serves as a valuable asset across diverse departments, supporting employees in their day-to-day tasks.

We closely monitor the implementation of the Intelligent assistant, anticipating valuable insights on its potential impact. The solution shows promise, though its specific applications are yet to be fully understood. This is also what is expected from first generation GenAI implementations and we are very curious in how it’s going to be used.

Coding buddy

Generative AI holds significant promise in the realm of coding, with the digital coding buddy emerging as a transformative tool for software developers. It seamlessly aids in various everyday tasks, including code explanation, generation, bug identification, correction, test generation, and vulnerability mitigation. Serving as a digital pair programmer, it collaborates with human developers to expedite code writing and reduce workload. A Microsoft research study revealed that developers equipped with GitHub Copilot completed a coding exercise 56% faster than the control group (https://arxiv.org/abs/2302.06590). With such compelling evidence, we anticipate that this technology will significantly boost productivity in coding tasks.

Intelligent search

Generative AI has significant potential in intelligent search. Today, internal search tools  typically rely on keyword matching and predefined algorithms to retrieve relevant information. GenAI can enhance the search experience in three different ways;

  • Enhanced intent interpretation - the ability to understand context enables it to interpret ambiguous queries accurately. It can generate related search terms based on the user's input, expanding the search query to capture a broader range of relevant information.
  • Enhanced indexation - index vast amounts of information, including textual, visual, and multimedia content. It can employ semantic indexing, which considers the meaning and context of words and phrases, leading to more accurate and contextually relevant search results.
  • Generative response – help formulate an answer responding to your specific question.

By training a model on our proprietary data, it can be utilized both internally and externally, offering the potential to enhance accuracy, speed, and leave a lasting positive impression.

As a final remark, we must be aware to the fact that every technology comes with its own set of opportunities and risks. While Generative AI can unlock new levels of efficiency, productivity, and innovation, it also raises ethical questions around privacy, security, and the future of work. Thus, we must ensure that we use Generative AI in a mindful and responsible manner.

 

Other interesting blog posts:

https://www.ericsson.com/en/blog/2021/8/ai-powered-organization-decision-making

https://www.ericsson.com/en/blog/2021/4/blockchain-technology

https://www.ericsson.com/en/blog/2020/10/democratizing-ai

https://www.ericsson.com/en/blog/2021/10/process-mining-ai-task-mining

The Ericsson Blog

Like what you’re reading? Please sign up for email updates on your favorite topics.

Subscribe now

At the Ericsson Blog, we provide insight to make complex ideas on technology, innovation and business simple.