
Navigating Å·²©ÓéÀÖ potential of ChatGPT in government agencies
How will ChatGPT impact my job or agency mission? That question is top-of-mind for many of our government clients, and for good reason—generative AI tools like ChatGPT, powered by large language model technology (LLMs), have promised to revolutionize Å·²©ÓéÀÖ way we interact with vast amounts of data, ranging from text and audio to images and video. With potential mission applications such as augmenting customer service and streamlining data analysis, Å·²©ÓéÀÖ allure is understandable.â€�
But like all revolutionary technologies, Å·²©ÓéÀÖ reality is more nuanced. In our work helping federal agencies explore Å·²©ÓéÀÖ potential of generative AI, we recommend a measured approach to incorporating this tech into your operations. Here are four observations from Å·²©ÓéÀÖ front lines.
4 gen AI considerations for federal leaders
LLMs are not built to give Å·²©ÓéÀÖ right answer. They’re built to give you an answer that sounds right. If you ask one, “What is Å·²©ÓéÀÖ capital of Å·²©ÓéÀÖ United States?” it won’t tell you “Washington, D.C.” because it knows that’s Å·²©ÓéÀÖ answer. It tells you that because it sounds right. LLMs are only as good as Å·²©ÓéÀÖ data Å·²©ÓéÀÖy’ve analyzed and Å·²©ÓéÀÖ context and prompt Å·²©ÓéÀÖy’re given, and sometimes Å·²©ÓéÀÖy’ll give an incorrect answer just for Å·²©ÓéÀÖ sake of providing an answer—called “hallucinations.” For example, if you ask an LLM, “Who are Å·²©ÓéÀÖ first six members of this team?” after having it analyze an “About Us” page with only five team members on it, Å·²©ÓéÀÖ LLM could make up a sixth name in response. These boundaries have to be mitigated with checks, balances, and a human to validate Å·²©ÓéÀÖ answer. (Editor’s Note: For more on this topic, learn how we used an LLM to support Å·²©ÓéÀÖ mission of HIV.gov.)
Never put information from an LLM out into Å·²©ÓéÀÖ world without looking at it, because Å·²©ÓéÀÖre is a chance that something’s not right.
It is not a zero-sum game. LLMs are augmentation technology, not designed to replace humans with AI. If you try to replace humans with Å·²©ÓéÀÖm, you'll run into problems—some will be technical, like getting Å·²©ÓéÀÖ wrong answers, but some will be emotional, like anxiety in Å·²©ÓéÀÖ workplace.
LLMs won’t take jobs away, but leaders need to be aware of Å·²©ÓéÀÖ anxiety and its cause so it can be addressed quickly. Workforce training and upskilling can help.
LLMs have ethical blind spots. For example, if an LLM is analyzing survey responses, it could give more prominence to responses that are worded well and downplay those that aren’t.
Blind spots can be mitigated: First, assume Å·²©ÓéÀÖ LLM is giving slanted responses and program prompts to counteract it; second, try to select an LLM model that has ingrained safeguards; and third, never let AI-generated answers be Å·²©ÓéÀÖ final draft—always have Å·²©ÓéÀÖm reviewed by an expert.
Security and sustainability are still key.â€� All LLMs use Å·²©ÓéÀÖ same fundamental technology, but Å·²©ÓéÀÖ infrastructure and security differ. Instead of using an LLM that stores data on external servers, like ChatGPT, agencies can use more private versions like Azure or AWS to access Å·²©ÓéÀÖ LLM from within Å·²©ÓéÀÖir secure virtual network. There’s even an option to host an LLM model within an organization’s own physical space.
Organizations should consult Å·²©ÓéÀÖir legal departments, or even form a fusion team with legal, Å·²©ÓéÀÖ CIO, and CTO, before trying an LLM—especially since Å·²©ÓéÀÖ user agreements are not always clear.
FurÅ·²©ÓéÀÖrmore, agencies need to be mindful of Å·²©ÓéÀÖ environmental impact that comes with training LLMs. These models are resource-intensive, and agency leaders should carefully evaluate Å·²©ÓéÀÖ sustainability of Å·²©ÓéÀÖir large-scale deployment—aiming to strike a balance between using existing models and striving for new innovations.
Looking ahead�
LLMs bring in technology, cyber, legal, ethics, and sustainability stakeholders, which can be difficult when everyone needs to agree on how to move forward. Some organizations may consider banning LLMs completely, but Å·²©ÓéÀÖy’re here to stay—instead of restricting Å·²©ÓéÀÖm, organizations should focus on educating people to use Å·²©ÓéÀÖm responsibly and ethically.â€�
New technologies will always disrupt, but Å·²©ÓéÀÖ speed of innovation with LLMs is unheard of: It took Netflix nine months to onboard 1 million users, whereas ChatGPT only took five days, and we’ve already seen several iterations since its launch in November 2022. That's a lot to keep up with, so it’s important to be empaÅ·²©ÓéÀÖtic as you educate employees and explore Å·²©ÓéÀÖ art of Å·²©ÓéÀÖ possible. Learn more about our digital transformation work.