GEOINT 2023 - NGA: The Way of AI
TLDRThe speaker discusses the National Geospatial Intelligence Agency's (NGA) vision and progress in artificial intelligence, emphasizing the need for continuous improvement in computer vision and machine learning. They highlight the challenges faced by small AI businesses in engaging with the government and the importance of adapting transformative technologies like large language models to the agency's needs. The speaker also addresses the future of the Maven program, distinguishing it from Project Sapphire and emphasizing the unique demands of different customer groups within the agency.
Takeaways
- 📚 The speaker discusses the 'Way of AI', a document that outlines the agency's approach to artificial intelligence, which was a year in the making and has already seen significant advancements in the field.
- 🚀 The agency is not pausing its AI efforts despite rapid advancements, as they believe they are not yet good enough in areas like computer vision and machine learning.
- 🔍 The National Geospatial Intelligence Agency (NGA) is focused on improving object detection, judicial accuracy, and the speed of detections to better support policymakers and combatant commanders.
- 🤖 The speaker emphasizes the need for better labels, algorithms, and models to enhance the capabilities of AI systems in the domain of computer vision.
- 📈 The NGA has set high standards for software delivery, expecting multiple times a day improvements and continuous integration and delivery.
- 🌐 The agency is working on an environment that allows for faster software accreditation and security, as well as housing and maintaining more data than ever before.
- 🔥 The speaker highlights the potential of applying large language model technology to images and computer vision, aiming to move from good to the best in the world.
- 📝 The NGA is open to new ideas and technologies, especially from small AI businesses, and encourages submissions of white papers and ideas.
- 🤝 The agency collaborates with industry and academia, as seen in their work with Moonshot Labs, focusing on synthetic generation of label data.
- 🎤 The speaker, Mark, has been selected as the Director of Data and Digital Innovation (DDI) at NGA, and plans to continue the course set by his predecessor, with a focus on communication and engagement.
Q & A
What is the main focus of the National Geospatial-Intelligence Agency (NGA) in terms of artificial intelligence?
-The main focus of NGA in AI is computer vision, machine learning, and the domain of object detection to derive information from objects with better judicial accuracy.
How does the speaker feel about the current state of AI technology in the industry?
-The speaker finds the current state of AI technology fascinating and awesome, but also acknowledges that there is still room for improvement, as they believe they are not yet good enough.
What is the significance of the 'AI Way' document mentioned in the transcript?
-The 'AI Way' document is a strategic document that outlines NGA's vision and approach to artificial intelligence, which has been approved and signed by the director.
What does the speaker think about the idea of pausing the development of AI technology?
-The speaker is against the idea of pausing AI development, as they believe it goes against the DNA of technologists and the mission of improving efficiency and outcomes.
What is the speaker's vision for the future of AI at NGA?
-The speaker's vision is to excel by bringing the best and brightest minds to support national security, focusing on applying large language model technology to images and computer vision to become the best in the world.
What is the role of the Data and Digital Innovation (DDI) directorate at NGA?
-The DDI directorate is responsible for shaping the future of data and digital innovation within NGA, organizing work, establishing offices, and ensuring budgets are set for data operations and AI.
How does the speaker plan to improve communication and engagement within the agency?
-The speaker plans to encourage more open communication and engagement, including discussing Maven more openly and fostering a culture of continuous improvement and innovation.
What is the difference between Maven and Project Sapphire in the context of AI and computer vision programs?
-Maven focuses on delivering AI capabilities to combat units like the 18th Airborne Corps, while Project Sapphire is centered on developing tools for analysts, highlighting the different needs and feedback loops for each group.
What are the expectations for AI-ML object detection performance at NGA?
-The expectations are for 100% accuracy in object identification and geolocation accuracy to be within two meters or less to support precise actions like delivering guided ammunition.
How does the speaker view the future of AI in terms of object detection capabilities?
-The speaker is confident that within the decade, AI object detection capabilities will reach 99.99% accuracy, aligning with the high expectations set by analysts and combatant commanders.
Outlines
🗣️ Introduction to AI and NGA's Vision
The speaker introduces the topic of artificial intelligence (AI) and its significance in the context of the National Geospatial-Intelligence Agency (NGA). They mention a document called 'the way of AI' and its difficulty to distill the agency's AI efforts into a few pages. The speaker emphasizes the rapid advancements in AI since its inception, particularly in the last couple of years, and the need for continuous improvement and adaptation to stay ahead.
🚀 NGA's Progress and Challenges in AI
The speaker discusses NGA's focus on computer vision and machine learning, acknowledging that despite significant investment, there is still room for improvement. They highlight the need for better object detection, faster identification, and more accurate geolocation. The speaker also addresses the high standards set by NGA and the opportunity for the audience to contribute to these efforts, especially in the area of Transformer technology applied to computer vision.
📈 NGA's Data and Digital Innovation
The speaker, now as the Director of Data and Digital Innovation (DDI) at NGA, shares his vision for the directorate. He outlines the offices within DDI, including plans, programs, and strategy, as well as the QD (formerly known as the data core) and the Maven office. The speaker emphasizes the importance of open communication and the need to identify and incorporate unique and innovative technologies into NGA's operations.
🤖 AI Way and Software Way Integration
The speaker explains how the AI Way fits into NGA's Technology Focus areas, particularly in relation to the Software Way. They discuss the importance of measuring the speed and frequency of software delivery and the application of AI and machine learning to improve these processes. The speaker also highlights the potential for applying large language models to visual data and the expectation for breakthroughs in the coming years.
🔍 Future of Maven and Project Sapphire
The speaker addresses questions about the future of Maven, especially after its transition to NGA, and contrasts it with Project Sapphire. They clarify that while both programs are valuable, they serve different needs and audiences within the agency. Maven is designed for combat units, while Project Sapphire focuses on analyst tools. The speaker emphasizes the importance of both programs and the potential for synergies and shared core technologies.
🎯 Setting the Bar for AI/ML Performance
The speaker discusses the high expectations for AI/ML object detection performance at NGA, including 100% accuracy and geolocation accuracy within two meters. They acknowledge that while these goals are not yet achieved, they are confident that they will be reached in the future. The speaker also reflects on the potential impact of the technology being developed today on the future, hoping it will prevent self-destruction.
Mindmap
Keywords
💡AI
💡Computer Vision
💡Machine Learning
💡NGA
💡Maven
💡Transformer Technology
💡ML Ops
💡Software Way
💡AI Way
💡Project Sapphire
💡Synthetic Generation of Label Data
Highlights
The speaker discusses the challenges of distilling the agency's artificial intelligence efforts into a concise message.
The agency's vision statement for AI is difficult to convey due to the complexity of the field.
The speaker mentions the rapid advancements in AI since the first draft of their AI strategy a year ago.
The speaker highlights the importance of computer vision and machine learning in the agency's work.
The agency is not satisfied with the current state of AI capabilities and seeks improvements.
The speaker emphasizes the need for better algorithms and models for object detection and information derivation.
The agency is looking for innovative AI solutions and encourages participation in the conversation.
The speaker discusses the high bar set by the agency for AI capabilities and the expectation for continuous improvement.
The agency is interested in applying transformer technology from large language models to images for computer vision.
The speaker mentions the opportunity for the best and brightest to support national security through AI innovation.
The agency is working on synthetic generation of label data as part of its innovation efforts.
The speaker talks about the challenges small AI businesses face in engaging with the government.
The agency is focused on improving communication and engagement with industry partners.
The speaker discusses the AI Way and how it fits with the Software Way and NGA's technology focus areas.
The speaker anticipates breakthroughs in applying large language models to visual data in the next couple of years.
The future of Maven is discussed, including its differences from Project Sapphire and its focus on different customer needs.
The speaker expresses confidence in achieving 100% accuracy in AI object detection in the future.
The speaker concludes with a humorous remark about the future and the impact of current technology.