Decoding AI: Part 8, Data governance and generative AI
Welcome to Part 8 of our Decoding AI: A government perspective learning series! In our previous blog, we re-introduced a technique that can help us achieve better boundary conditions and quality outputs: Retrieval Augmented Generation (RAG) with GAI. In this module, we wrap up with a culmination of all our learnings till date, and how they can benefit the US government and the public. We will also share some of the best practices and tips for using generative AI, and how to avoid or mitigate some of the potential risks or challenges. — Kent Cunningham, Miri Rodriguez, Siddhartha Chaturvedi
Why generative AI matters for the US Government
Generative AI, particularly in its applications through platforms like Azure Open AI and Copilot, stands at the forefront of innovation in artificial intelligence, offering transformative possibilities for federal civilian agencies. Its capacity to enhance various government functions extends beyond the realms of proposal management and content creation. For instance, in the realm of public service, generative AI can revolutionize how agencies interact with citizens, offering more streamlined and efficient access to benefits and services. It can profoundly improve call center operations, not only in terms of efficiency but also by elevating customer satisfaction through more responsive and personalized support.
The application of generative AI in automating and refining the processes of drafting proposals, reports, and official documents is another area where its impact is deeply felt. By employing advanced natural language generation and summarization techniques, these AI tools can significantly reduce the time and resources spent on document creation, allowing agencies to focus more on strategic decision-making and mission-critical tasks. Additionally, generative AI can aid in data analysis and decision-making processes, providing insights and recommendations that are based on comprehensive data evaluation, thus enabling better-informed decision-making and program delivery.
However, the deployment of AI and Large Language Models in government sectors brings with it the imperative of adhering to stringent standards of data privacy, ethics, and quality. In the context of government operations, where data sensitivity and accuracy are paramount, it’s essential to ensure that the use of generative AI aligns with these principles. To this end, federal agencies must adopt a responsible and trustworthy approach towards the application of generative AI. This involves embracing robust data governance practices and frameworks, ensuring that the technology is not only effective and efficient in enhancing mission delivery but also compliant with the highest standards of data integrity and ethical considerations. In doing so, agencies can leverage the full potential of generative AI while maintaining public trust and ensuring the safeguarding of sensitive information.
Your data governance practices are crucial to set clear limits and safeguards, and use generative AI with your data in a responsible and efficient way.
How to use generative AI responsibly and effectively
In this learning series, we have covered some of the key concepts and techniques that can help government agencies to use generative AI responsibly and effectively, such as:
- Understanding the difference between generative AI and traditional AI or machine learning, and how they use data or content differently.
- Applying the framework and principles of responsible AI to generative AI, and ensuring that generative AI systems are fair, reliable, safe, private, secure, inclusive, transparent, and accountable.
- Creating boundary conditions in generative AI, which are the limits or specifications that are applied to generative AI systems, such as large language models (LLMs), to ensure that they produce outputs that are relevant, accurate, and appropriate for the intended domain, task, and audience.
- Using retrieval augmented generation (RAG) with GAI, which is a technique that combines retrieval and generation to create boundary conditions and quality outputs for generative AI, especially in high-value, high-volume critical domains such as healthcare and genomics.
Everything starts with Data
As we have discussed previously, generative AI is often at its most effective when it’s firmly rooted in reliable and trustworthy organizational data, finely tuned to deliver optimal results for specific outcomes. We can think of generative AI not as an entirely new technology, but more like a highly adaptable version of an existing tool that, when properly nurtured, becomes an integral part of the business fabric. As always was true in the past, this new adaptation and integration process involves a blend of critical data governance and grounding in relevant data to be most effective.
Strong data governance is the bedrock of this process. Just as a new puppy in your home needs the right care and environment to thrive, generative AI requires high-quality, well-governed data to function effectively. It’s not just about feeding it any data; it’s about feeding it the right data. This is where the traditional data governance practices which have been in place for many years really begin to shine; by ensuring that the data used is not only accurate and reliable but also ethically sourced and compliant with applicable regulations. Without this critical foundation which has always been in place for enterprise ready applications, the potential of generative AI (or any AI/ML) in mission-critical settings can be severely compromised.
Now, let’s talk about tuning and grounding. Generative AI, when grounded in an organization’s specific (and well governed) data, becomes more than a general-purpose tool; it transforms into a customized solution that knows the facts and terminology of your organization. By tuning it to the nuances of the organization’s data and desired outcomes, generative AI becomes more aligned with the business goals, yielding more relevant and impactful results. This tuning is about making the AI system responsive to the unique demands and context of the business.
Meta-prompting is another key aspect. It’s about guiding the generative AI with precise instructions, ensuring it generates content that’s not just creative but also aligned with specific business objectives. Much like how you would give specific commands to a pet to elicit certain behaviors, meta-prompting in generative AI steers the creative process in a direction that’s beneficial for the organization as it builds responses from the accessible data.
In a nutshell, generative AI in a mission-critical setting is a symphony of well-governed data, tuning, relevant grounding on that data, and strategic meta-prompting. It’s a technology that, with the right care and guidance – much like nurturing a new pet – can bring about transformative results. All of this starts with the well-managed data governance practices have long been the driving force behind successful mission applications. These time-tested practices form the cornerstone upon which the efficacy of AI, including generative AI, is built. Maintaining and enhancing these traditional data governance methodologies remains crucial in unlocking the full potential of AI technologies across the organization as well.
The future of generative AI for the US Government
Generative AI is not a futuristic or hypothetical technology, but rather a present and practical one, that can help the US government to achieve its mission and vision, and to serve the public better and faster. Generative AI is not a replacement or a threat, but rather a complement and a partner, that can augment the human capabilities and enhance the human outcomes. Generative AI is not a black box or a magic wand, but rather a complex and dynamic system, that requires careful design, implementation, monitoring, and evaluation.
We hope that this learning series has helped you to understand the basics and the benefits of generative AI, and to get started with your own generative AI projects and experiments. The possibilities and opportunities of generative AI are vast, but it is important to use it responsibly and effectively for the US government and the public.
Here below, we’ve provided links to the earlier modules in this learning series, however, we encourage you to sign up here to stay up to date on the latest news and announcements on AI for the US Government.
Catch up on past modules
- Part 1: Decoding generative AI
- Part 2: From anomalies to action
- Part 3: Making data speak human
- Part 4: Sensing a more complete reality
- Part 5: A crucial focus on trust
- Part 6: A vital step for responsible AI
- Part 7: A powerful technique for generative AI
Kent Cunningham Chief Technology Officer, Federal Civilian Government, Strategic Missions and Technologies, Microsoft Federal
Light
Dark
1 comment
For federal agencies, are there any legislative/ policy sources you would recommend tracking in particular?