By John P. Desmond, AI Trends Editor
Pushing reliable AI and machine learning to reduce the risk of the agency is a priority for the US Department of Energy (DOE), and identifying the best methods for implementation of the AI scale is the priority of the US General Services Administration (GSA).
Participants learned the same in two sessions AI World Government The direct and virtual event was held last week in Alexandria, Wah.
Pamela Asom, director of the AI and Technology Office in the DOE, who talks about pushing reliable AI and ML techniques to reduce the agency’s risks, has been involved in enhancing the use of AI in the agency for many years. With the emphasis on applicable AI and data science, it has been involved in monitoring risk -reduction policies and standards and has been involved in saving AI to save lives, fight fraud and stabilize cyberscript infrastructure.
It emphasized the need for an AI project effort to become part of the strategic portfolio. “My office is there to reduce the threat by gathered us to run a comprehensive view of AI and tackle the challenges,” he said. This effort helps with the DOE’s AII and Technology Office, which focuses on converting the DOE into AI enterprise globally by accelerating research, development, delivery and adoption of AI.
“I am telling my organization to keep in mind the fact that you may have tons and tons of data, but it may not be a representative,” he said. Its team looks at examples of international partners, industry, academia and other agencies that “we can trust” with Systems that add AI.
“We know that the AI is disrupting, trying to do a human work and do it better,” he said. “This is beyond human ability,” he said. It is beyond data in the spreadsheet. It can tell me what I will do before I think it myself. It is so powerful. “
As a result, data sources should be paid close attention. Iceom said, “AI is very important for the economy and our national security. We need health. We need the algorithm which we can trust. We need accuracy. We do not need prejudice,” and don’t forget that you need to monitor the production of models after they are deployed. “
Executive Orders Guide GSAA Work
Executive Order 14028, which is a detailed step to tackle the cyberciction of government agencies released in May this year, and the executive order 13960, which promotes the use of reliable AI in the federal government issued in December 2020, provides valuable leaders.
To help handle the risk of development and deployment of AI, ISOM has developed AI Risk Management Playbook, which provides guidance around the system’s characteristics and reduction techniques. It also contains a filter for ethical and reliable principles that are considered in the AI life cycle stages and types of risk. In addition, playboxes are related to relevant executive orders.
And it provides examples, such as your results 80 % accuracy, but you wanted 90 %. Iceom added, “There is something wrong,” Playbok helps you see these types of problems and what you can do to reduce the risk, and what factors you should weigh your project when you design and prepare your project. “
Currently, the internal, agency is considering next steps for the external version. “We will soon share it with other federal agencies,” he said.
GSA AI projects have been outlined for the best ways to scale
Anil Chowdhury, the director of the Federal AI’s implementation for the GSA’s AI Center of Excellence (COE), who spoke on the best ways for the implementation of AI on a scale, has more than 20 years of technology, operations and program management in defense, intelligence and national security sectors.
COE’s mission is to accelerate technology modernization throughout the government, improve public experience and enhance operational performance. “Our business model is to contribute with industry articles experts to solve problems,” said Chaudhry, “Chaudhry added,” We are not in the business of regenerating industry solutions and creating them. ”
The COE is providing recommendations to the partner agencies and is working with them to enforce the AI system as the federal government is very busy in the development of AI. “For AI, public landscape is wide, every federal agency is still underway,” he said, “he said, and the maturity of the AI experience varies widely in the agencies.
Common use cases they are seeing include AI’s attention on high speed and efficiency, cost saving and cost, increasing time and quality and compliance with better response. As an excellent process, he recommended agencies See their commercial experience With major datases they will face government.
Chaudhry said, “We are talking about structural and unintentional figures here about Bits and Axis. (ED. Note: A pata byte is 1,000 tarbbitis.) “Also ask industry partners about their strategies and processes about how they do macro and micro trend analysis, and their experience has been in the deployment of boats such as robotic process automation, and how they show stability as a result of the growth of data.”
He also asks the potential partners of the industry Explain the AI talent in their team Or what skills can they get? If the company is weak on AI talent, Chaudhry asks, “If you buy something, how do you know what will you find when you have no way to evaluate it?”
He added, “One of the best practices in the implementation of the AI explain how you train your workforce to benefit from AI tools, techniques and methods, and explains how to enhance and strengthen your manpower. Access to talent is either successful or achievement, as a result of success. It is a matter of doing. “
In another excellent exercise, Chaudhry recommended to check the industry partner Access to financial capital. “AI is a field where the capital flow is extremely stable,” he said. You can’t predict or project that you will spend X -dollars to arrive this year where you want, “he said, because an AI development team may need to detect another speculation, or need to clear some data that is not transparent or potentially biased.” If you do not have access to the fund. “
Another is the best process Access to logistics capitalSuch as data that sensors submit for the AI IOT system. “AI requires a lot of statistics for AI, which is authentic and timely. Direct access to this figure is essential,” Chaudhry said. He recommended that data sharing agreements with AI system -related organizations exist. “You may not need it right now, but to access the data, so you can use it immediately and think in cases of confidentiality before needing data, this is a good process for measuring AI programs,” he said.
One last best practice plans Physical infrastructure, Such as a data center. “When you are in a pilot, you need to know how much capacity you need in your data center, and when you need how many closing points you need to measure the application,” Chaudhry said.
Get more information AI World Government.