Amidst increasing discussions about the use of AI and ChatGPT in the Federal Government, we turned to SoKat, a rising 8(a), WOSB and 8(a) STARS III Prime that has been building and deploying award-winning Artificial Intelligence, Machine Learning and Blockchain technology solutions in the Federal government since 2017. Speaking with SoKat President, Dr. Jim Liew, who is also an Associate Professor at Johns Hopkins University teaching AI and Machine Learning, we looked to learn more about the evolution of ChatGPT, the opportunity and next steps for government, and SoKat’s plans to continue being a trusted partner in ethical, transparent and transformative AI.

The Evolution of ChatGPT

SoKat’s team of researchers have been closely following the evolution of ChatGPT since it was merely an academic thought experiment that birthed GPT-1. The team quickly realized GPT’s revolutionary capabilities for text analytics and advanced language generation, and therefore built their Veteran-facing healthcare application using GPT-3, which won SoKat 2nd place honors (1st place in Patient Experience category) in the VA National Artificial Intelligence Institute’s AI Tech Sprint in 2021. SoKat was also brought into the Veterans Experience Office to build VA’s first large language model using patient-generated data for detection of suicide ideation.

Since then, OpenAI’s release of ChatGPT, which initially launched with GPT-3 as the base model, and has currently evolved to GPT-4, unleashed  a torrent of excitement over the long-awaited arrival of AI on the stage of national public discourse. “The rapid pace of adoption of ChatGPT is due to the fact that this latest evolution draws from a vast pool of collective knowledge available on the internet that can be customized and generated from a simple user-provided prompt – power and simplicity.”

The success of any AI-generated response still requires the user knowing the right prompts and the right paths to delve further into the information. Liew suggests that “the future of coding will involve crafting elegant prompts that enable people to interact with technology and bring ideas that solve real world problems to life.”

As would be expected, each iteration of generative AI models, be it ChatGPT, Bard or the next emergent model, will continue to improve, possibly far more quickly than we can anticipate.

The Call Center Opportunity and Beyond

AI and machine learning have the opportunity to solve a variety of Federal agency challenges, and currently, the low hanging fruit are call centers. GPT can help improve customer experience by supporting interactions with Government staff and at the same time, reduce the workload away from simpler queries, freeing staff to respond to more complex requests.

“I envision help desk support being drastically improved and providing a better customer experience. We’ve proven in our own work that large language model (LLM) chatbots can vastly improve user experience.  For example, we designed a machine learning application that empowered Veterans in their healthcare journey: the application digested the Veteran’s electronic health record, using algorithms to “learn” from the unique EHR, then provided a chatbot with which the Veteran can interact and get customized answers to all of their healthcare questions.”

Another opportunity SoKat is working  on to leverage LLM to improve quality of care for Veterans is in building an AI-based suicide ideation engine in the VA, which is a quantum-leap forward from the incumbent rules-based engine. SoKat was also one of 30 companies to win a Phase 2 award in the VA Mission Daybreak Challenge, beating out more than 1300 companies who entered this national challenge to find innovative technology and approaches to dramatically reduce Veteran suicide. SoKat’s winning AI solution, in this challenge, also leveraged LLM to analyze text data from multiple sources and to predict suicide ideation.

Government’s Next Steps

Presently, how can AI bring transformative change to the way that the federal government operates and serves the American public? The list is long, but at the very top in terms of impact are 1) preventing security vulnerabilities, 2) fraud, waste and abuse detection, 3) risk management 4) improving customer experience and 5) government talent recruitment, training and retention. “The more government personnel understand AI’s limitations and potential, the better equipped we are to harness its power and manage its risks, therefore pushing the U.S. closer to being the global leader in AI.”

At the same time, SoKat seeks a measured and balanced approach when leveraging the most cutting-edge AI technologies with ethical and safety concerns. For this reason, all of SoKat’s AI solutions require human judgment as the final step in the process of employing AI, otherwise known as ‘human in the loop’.  “AI offers immense potential, and I use it daily; however, it’s crucial to treat AI as a guide and a tool, not the primary decisionmaker.”

One way that SoKat is currently leveraging AI and GPT technology to transform the federal acquisitions process is through the launch of its AI-powered compliance tool for the FAR, called FARSight (https://the-far-sight.com).  “When we spoke to federal acquisitions staff about their pain points in their workflow, of which there were many, we decided to go after the problem that we could solve the most quickly and effectively, given our expertise in LLM generally, and GPT specifically.” FARSight is a tool for acquisition staff that leverages AI technologies to provide immediate and accurate responses about the FAR for all stages of the acquisitions process, thereby reducing time and stress and increasing FAR-compliance.

Another increasingly popular use case for AI, in both government and industry, is in supporting decision making. Whether it’s a DHS executive who is being overwhelmed with data that may or may not be relevant for the decision at hand to accurately calculate the risk of a threat, or an agency’s CIO being pitched by multiple vendors with conflicting views, an unbiased fact-based tool saves critical time and provides crucial accuracy. SoKat’s Risk Management as a Service software leverages AI to address this very real business problem of decision making support. In a world that is being inundated with increasingly more data, decision-makers need a tool that will aggregate the data, digest it, and maximize its value for optimal decision making, which may include providing risk scores and detecting anomalies.

Democratize AI – through Office Hours

As part of SoKat’s overarching and long-term mission to democratize AI and bring everyone along on this AI journey, which is the only way to ensure that AI serves all people, SoKat plans to roll out “office hours” for government staff so they stay apace with the startlingly rapid development and deployment of AI technology. Furthermore, SoKat’s team of academics, researchers, and industry experts will aim to help the federal government translate the vast frontier of AI into immediately useful and concrete tools by simply asking: “what is the urgent business problem you need to solve?”

“We will start by identifying business problems, collecting relevant sample data, and then engage in a robust discussion with Feds to brainstorm on potential solutions, risks, frictions, and barriers. It’ll be fun, creative, public-service driven and best of all, it’s the attendees who will be designing and driving the innovative solution to further their agency’s mission.”

About SoKat

SoKat creates award winning, Artificial Intelligence, Machine Learning and Blockchain technology solutions and is a trusted partner of Federal Government agencies in leveraging AI to support mission success. SoKat unlocks the hidden value of data through thoughtful, creative and human-centric design solutions, comprising of actionable intelligence, transparent data analytics, and robust predictive AI models.



Is your company an OS AI Premium Member? Learn about all the benefits here. Packages start at $500.

How useful was this post?

Click on a star to rate it!

We are sorry that this post was not useful for you!

Let us improve this post!

Tell us how we can improve this post?

Leave a Reply