Can technology make Congress artificially intelligent?
The House of Representatives is exploring its own generative AI model for internal use, as Congress looks for ways to use new tech tools to overhaul its work.
House Deputy Chief Administrative Officer John Clocker told lawmakers Tuesday that his team is looking at getting a customized and secure generative AI “large language” model into the hands of members.
Large language models, or complex algorithms, grew in popularity in 2023 with the rise of conversational AI tools called chatbots such as OpenAI’s ChatGPT. There is talk AI programs could help senators and House members with tasks such as scheduling, debate prep and even drafting bill language.
“At this time, the CAO does not envision building a GenAI LLM from scratch,” Mr. Clocker said in written testimony to the Committee on House Administration. “As the entity responsible for House technology deployment, the CAO can attest to the cost and innovation benefits of leveraging third-party technology versus trying to build and maintain applications in-house, especially considering the House’s broad scope of business needs.”
The Capitol is playing catch-up: Legislatures around the world already have incorporated AI into their work.
Singapore built its own AI model into legislators’ government laptops, Brazil is experimenting with facial recognition for legislators to work remotely, and India’s parliament uses AI to control access to its building with biometric scanning, according to a POPVOX Foundation study published this month.
Rep. Morgan Griffith, Virginia Republican, said Tuesday he hoped members could one day use AI to listen to draft legislation, reports and amendments while traveling, such as when flying away from Washington or driving around their districts.
AI could also help lawmakers craft amendments “almost instantaneously” for use in committees and on the House floor, he added.
“When you talk about drafting amendments, I think that’s a very different story involving legal counsel and I can tell you they’re concerned about the appropriate use of drafting legislative language,” Mr. Clocker said to Mr. Griffith at the hearing.
“Well I think that’s up for the member to decide,” Mr. Griffith interjected. “But it’s a tool where if you hear something and you want to make a quick amendment.”
Mr. Clocker told the committee that his office has conducted legal and cybersecurity reviews of ChatGPT, Google’s Bard, and Microsoft’s Bing Chat, now known as Copilot. Upon changes made by the developers, Mr. Clocker said the House allowed limited use of ChatGPT Plus in June.
Widespread adoption of new AI tools across the federal government is coming. The Government Accountability Office told House lawmakers that it already has incorporated an AI model into its oversight work.
Taka Ariga, GAO chief data scientist, said the agency’s innovation lab he oversees is experimenting with several AI tools.
“One of our AI tools is a large language model we recently began deploying to explore generative AI capabilities like those found in industry, augmented with GAO-specific information and appropriate security controls,” Mr. Ariga said in written testimony. “Among other capabilities, we believe this prototype will help us efficiently synthesize our past work to meet congressional and taxpayer needs.”
Mr. Ariga said his team is also looking at using AI for editing, such as making texts consistent with the office’s style guide.
While federal officials from the Pentagon to the Library of Congress are examining how to adapt and deploy AI models safely and effectively, Mr. Clocker said his team wants to ensure the AI tools will not replace human legislators.
“These tools are nowhere [near] ready to replace human judgment and the CAO will not use these tools to replace human judgment anytime soon,” he said. “And I can’t imagine us doing that anytime soon.”
• Ryan Lovelace can be reached at rlovelace@washingtontimes.com.
Please read our comment policy before commenting.