Artificial Intelligence in the public and private sectors

You're not the only one nervous about AI -in light of rapid AI growth and adoption, the U.S. Government recently held three Subcommittee Meetings designed to understand the implications posed by the widespread adoption of AI technology in the public and private sectors. So why is the US Government concerned about AI in society, and what role should it be considering in the private sector? Sid Mair, senior vice president of Federal Systems at Penguin Computing, weighs in.

Beginning his technology career at NASA, Sid brings more than 30 years of expertise across all aspects of the federal market, including the Department of Defense, Homeland Security, civilian agencies in both classified and unclassified areas, as well the Executive and Congressional branches of government. Penguin Computing most recently built the world’s largest AI cluster in the private sector. The company also custom designs, builds and supports AI, HPC cloud and enterprise data center infrastructures, helping more than 2,500 customers in 40 countries – including government customers - to derive the greatest value from data.


ADM: What should legislators consider as they access the appropriate role of AI systems in the federal space?

Mair: On February 14, 2018, the first hearing was held and the Subcommittee heard from four experts from the (tech) private sector and academia. On March 7, 2018, the second hearing featured government witnesses to discuss how their agencies are engaging with AI through research, procurement and applications. The third and final hearing was held on in late April and examined the potential challenges for industries impacted by the widespread adoption of AI technology.

Artificial Intelligence (AI) has amazing potential to create efficiencies in how the government operates, from balancing the federal budget and bolstering GDP, to managing the complexities of the healthcare system and spotting potential national security threats. The government should evaluate its existing IT infrastructure to identify gaps in technology, audit their internal AI expertise and determine how heavily to regulate AI in the interest of public safety and well-being.

While artificial intelligence is increasingly cited as either a great fear or hope for future generations, few in the public or private sector have a clear understanding of it. What can you offer to help create more understanding around the use of AI?

For many, the idea of AI conjures the notion of replicating the human brain, but we’re far from that. AI technology has been under development for decades. It’s currently used to examine massive amounts of historical data and identify trends or patterns to solve complex computational problems. The explosion of data, improved algorithms and exponentially more powerful computing systems underpin the growing buzz surrounding AI, but this is not because we are inching closer to a fully automated reality where the machines will replace humans. Instead, AI technology and advanced algorithms have the ability to spot trends at a rate and scale that humans cannot. Using machines, which are capable of processing and understanding massive amounts of data with more speed and accuracy than humans, the government has the opportunity to streamline tasks that involve data-driven calculation or are repetitive in nature and identify patterns that are seemingly unrelated.

Artificial intelligence is not a “new” technology. It is a computing paradigm that has recently risen to prominence as algorithms have improved over time. As a new computing paradigm, AI can assist U.S. entities to better understand how to harness the power of data to create value and redefine operations.


ADM: What are the challenges and opportunities the government faces when exploring the use of AI?


Mair: As the government looks to adopt more AI-driven systems, the way government employees conduct their roles will likely change. Roles that follow clear processes and procedures are likely to shift to some level of machine automation. If public sector workers have the ability to defer computationally intensive or process-oriented elements of their jobs to machine learning systems, their time will be freed up to focus on different types of tasks.

When exploring the use of AI, transitioning and redefining certain roles and educating those who will be working in conjunction with AI systems could prove challenging at the outset. Building and maintaining the correct technology is also a potential roadblock to AI success. Security is also anticipated to be an ongoing trial as AI is powered by data (which is often personal, sensitive or confidential) and that data must only be made available to those intended to have access.

ADM: What should the government do to increase its understanding of artificial intelligence? How should it address uses and barriers to adoption when it comes to artificial intelligence?

Mair: Artificial intelligence (AI), and specifically machine learning (ML) and deep Learning (DL), is more than the next stage of data analytics. Combined with levels of computational power normally associated with high-performance computing (HPC), the adoption of ML/DL techniques is revolutionizing a wide range of application areas. By using these radically new approaches to solve difficult, existing problems, AI is leading to fundamental breakthroughs, however, entities are facing critical challenges in constructing an AI infrastructure that can actually drive the outcomes they seek. This is partly because of the speed at which research is progressing, but also because of the general lack of experience researchers have with building systems at scale.

ADM: Many AI systems are designed without sufficient thought in key areas, including the following themes:

1. Managing software ecosystem, focusing on orchestration and workload portability
    
2. Accurately sizing AI infrastructure to optimize utilization
    
3. Balancing overall data workflows and performance
    
4. Deploying and scaling AI infrastructure efficiently


Mair: The leading-edge nature of the work and the complexities involved in getting systems up and running mean that AI researchers often struggle to integrate the most robust technology currently available and find administrators with the technical expertise to use these technologies to their maximum potential. While there is no turnkey AI solution that fits all situations, the government should partner with technology leaders that understand which factors are critical to designing, building and deploying AI systems that are tailored to their specific needs.

ADM: Can you provide examples of how the government can implement AI systems and practices and the benefits that can and will result?


Mair: Expertise is essential. The correct people and technology must be in place as AI is deployed and continues to advance over time. The government should continue to partner with technology leaders who understand the intersection between the IT needs of the federal government and the ability to design and deliver infrastructure, solutions and expertise that allow federal agencies to successfully deploy AI- driven systems.

AI can reduce human error that stems from decision making. AI has the potential to improve data processing to identify cybersecurity and national security threats, grow the GDP by creating efficiencies and streamlining process redundancies, identify fraudulent activity in the U.S. Financial System, support military agencies in terms of intelligence, coordination and safety, improve the healthcare system, advance the Internet of Things (IoT), enhance weather modeling to predict events like hurricanes and more. 


ADM: What role should the government have in regulating AI use in the private sector? 


Mair: When you examine AI from a machine learning perspective, extreme regulation is not needed for machine learning applications as they are narrowly focused and designed to solve specific problems. When AI advances beyond machine learning applications to be self-educating or autonomous, the government will need to institute more stringent regulations in the areas that can be threatening to its citizens.

Regulatory policy should mirror many of the regulations we have in place today. For example, the SEC governs financial standards and we should consider similar standards for algorithmic trading. Similarly, just as we have driving laws to keep drivers and citizens safe, the same thinking should be applied to regulating autonomous vehicles.