Artificial intelligence (AI) and machine learning (ML) have been transforming the business landscape for over a decade. However, the impact of AI is still being felt, and many organizations are just beginning to grasp its potential. In this article, we will explore how foundation models are being used to drive business success, and what lessons can be learned from the deployment of these models by Ordnance Survey (OS), a UK-based national mapping service.
Foundation models are a type of AI model that serves as a base for building more specialized applications. They are trained on a large dataset and can be used to extract features from that data. In the case of OS, the organization is using its foundation models to extract environmental features for analysis in a copyright-sensitive manner. By building these models from the ground up, OS is able to define the full training set with labeled data that it has internally.
One of the key benefits of foundation models is that they can be used to build subsequent output models. This allows organizations to connect to the problem they are trying to solve with source data, rather than having to train multiple models. For example, OS can use its foundation models to learn about roof materials, green spaces, or biodiversity, and then fine-tune the models to get more accurate results.
Another key aspect of foundation models is that they can be used in conjunction with commercially available tools to exploit and distribute geospatial data. This allows organizations to tap into the vast amounts of data that are available, and to use that data to drive business decisions. For example, OS is using its foundation models to analyze environmental features and to identify areas of interest for analysis.
So, what can business leaders learn from the deployment of foundation models by OS? Here are five key lessons:
• Develop a strong use case
• Establish purposeful methods
• Use other LLMs for fine-tuning
• Think about commercialization
• Keep one eye on the future
In this article, we will explore each of these lessons in more detail, and provide examples of how they can be applied in a business context.
**Lesson 1: Develop a strong use case**
Developing a strong use case is critical to the success of foundation models. This involves defining a clear problem or opportunity that the models can address, and identifying the specific features or data that are required to solve that problem. In the case of OS, the organization is developing foundation models to extract environmental features for analysis in a copyright-sensitive manner. By defining a clear use case, OS is able to build models that are tailored to its specific needs, and that can deliver accurate results.
**Lesson 2: Establish purposeful methods**
Establishing purposeful methods is also critical to the success of foundation models. This involves using focused training data to constrain costs and ensure that the models are building on a solid foundation. In the case of OS, the organization is using a combination of internal and external data sources to build its foundation models. By using purposeful methods, OS is able to build models that are accurate and reliable, and that can deliver results in a timely and cost-effective manner.
**Lesson 3: Use other LLMs for fine-tuning**
Another key aspect of foundation models is the use of other LLMs for fine-tuning. This involves using commercially available tools to refine and improve the models, and to adapt them to specific use cases. In the case of OS, the organization is using a combination of Microsoft’s Azure machine learning models and Python-based tools to fine-tune its foundation models. By using other LLMs, OS is able to build models that are highly accurate and reliable, and that can deliver results in a timely and cost-effective manner.
**Lesson 4: Think about commercialization**
Thinking about commercialization is also critical to the success of foundation models. This involves considering the potential benefits and challenges of sharing or selling the models, and identifying opportunities for revenue generation. In the case of OS, the organization is exploring the potential for commercializing its foundation models, and is considering partnerships with external organizations to generate collaborative solutions to data-led challenges.
**Lesson 5: Keep one eye on the future**
Finally, it is essential to keep one eye on the future when deploying foundation models. This involves staying up-to-date with the latest developments in AI and ML, and identifying opportunities for innovation and growth. In the case of OS, the organization is exploring the potential for using generative AI to open up new access to in-depth insight, and to create definitive answers to prompts using trusted sources.
In conclusion, foundation models are a powerful tool for business success. By developing a strong use case, establishing purposeful methods, using other LLMs for fine-tuning, thinking about commercialization, and keeping one eye on the future, organizations can unlock the full potential of these models and achieve their goals.
