By John P. Desmond, AI Trends Editor
According to Ishaq Fabr, chief data scientist at the US Army’s integration center, the chief data scientist of the US military integration center, speaking in it, is a key point for the AI ​​stack to be taken by the US Army by the US Army for the AI ​​development platform efforts. AI World Government The event was personally and practically held last week from Alexandria, Wah,.
“If we want to transfer the military from the legacy system through digital modernization, I have found the biggest problem that is difficult to summarize the differences in the applications,” he said. “The most important part of digital change is the middle layer, a platform that makes it easier to stay on the cloud or on a local computer.” The desire is that you can move your software platform to another platform, with the same ease with which a new smartphone takes on user contacts and dates.
Ethics AI application reduces all layers of the stack, which is the top planning phase, followed by decision support, modeling, machine learning, large -scale data management and after the device layer or platform at the bottom.
“I am advocating that we think of the stack as the basic infrastructure and do not put it in a way to deploy requests and in our point of view,” he said. “We need to create a development environment for a global divided manpower.”
The military is working on a common operating environmental software (COES) platform, which was announced earlier in 2017, which is a design for DOD work that is expanded, Fertilly, Modular, Portable and open. “This is suitable for a wide range of AI projects. To put this effort into practice,” the devil is in detail, “he said.
The army is working on a prototype platform with CMU and private companies, including Vizimo Korapolis, Pa. , Which offers AI development services. Faber said they prefer cooperation and harmony with the private industry rather than buying products from the shelf. “The problem with this is that, you are stuck so much that you are being provided by this one vendor, which is not usually designed for the challenges of DOD networks,” he said.
Training a range of tech teams in Army AI
The army engages in AIwork Force Development efforts for several teams, including: leadership, graduate degree professionals; The technical staff, which is inserted through training to be certified. And AI users.
Tech teams in the army include various fields of focus: the general purpose software development, operational data science, deployment, which includes analytics, and a large team needed to create a computer vision system, such as machine learning operation team. “When people come by manpower, they need a place to cooperate, build and share,” said Faber.
Types of projects include diagnostic, which can connect the historical data series, according to predators and prescriptions, which recommend a prediction -based course. Faber said, “Very finally Ai. You don’t start with it.” The developer will have to solve three problems: Data Engineering, AI Development Platform, called “Green Bubble”, and the deployment platform, which he called “Red Bubble”.
“They are mutually special and all interconnected,” he said. These teams of different people need harmony with the program. Usually a good project team will have people from each of these bubbles. ” “If you haven’t done this yet, do not try to solve the green bubble problem. Unless you need operational, you have no meaning in pursuing AI.”
When asked by a partner, which group is the most difficult to reach and train, Faber said without hesitation, “The executives are the most difficult. They need to know what its price should be provided through the AI ​​environmental system. How to bring the biggest challenge,” he said.
Panel discusses AI use issues with the most capacity
In a panel on emerging AI -based panel, moderator Kurt Sioi, Program Director, World Smart City Strategy for IDC, Market Research Firm, asked that the emerging AI is the most capable of use.
“I will refer to the benefits of supporting pilots and operators for mission and resources planning, and the benefits of decisions at previous locations,” said Jane Charles Laide, the US Air Force’s autonomous tech adviser, the US Air Force’s Air Force Air Force Advisor.
“The natural language processing department is an opportunity to open the door of the AI ​​in the labor department,” said Crista Kannad, the chief of emerging technology. “Finally, we are dealing with the data of people, programs and organizations.”
Saoi asked, what are the major risks and risks to the panel people when implementing AI?
Anil Chowdhury, director of the Federal Services Administration (GSA), said in a general IT organization using traditional software development, that the effect of a developer’s decision is only far away. With AI, “You have to consider the effects of people, circles and stakeholders on the entire segment. With a simple change in the algorithm, you can delay the benefits or create a scale mistakes to millions of people. This is the most important threat.”
“Keep humans and humans on the loop in the loop,” he said.
“We have no intention of removing humans from the loop,” Kannad said, “This is really empowering people to make better decisions.”
It emphasized the importance of monitoring AI models after being deployed. “The models can flow as a data under changes,” he said. “So you need not only a level of critical thinking to work, but also to guess what the AI ​​model is doing is acceptable.”
He added, “We have set up use and contributions to the entire government to ensure that we are following the AI ​​responsible. We will never replace the algorithm people.”
The Air Force Lead said, “We often have issues where data does not exist. We cannot detect 50 years of combat data, so we use simulation. This risk is in teaching an algorithm that you have a real risk that you do not have a real risk.”
Chaudhry emphasized the importance of testing strategy for the AI ​​system. He warned the developers that “those who suffer from a device and forget the purpose of exercise.” He recommended development manager design in independent verification and verification strategy. “Your testing, at the same place, you have to focus on your energy as a leader. Before committing the resources, the leader needs an idea in mind how he will justify whether he is investing whether it is success.”
The Air Force’s lead talked about the importance of clarification. He said, “I am a technician. I do not make rules. The ability to explain the AI ​​function is important, the important is important. AI is a colleague with whom we have a conversation with us, rather than we come to the conclusion that we have no way to confirm.”
Get more information AI World Government.