Automation Versus AI In National Defense: Are We Overprescribing AI?

Innovation Automation Versus AI In National Defense: Are We Overprescribing AI? Chitra Sivanandam Forbes Councils Member Forbes Technology Council COUNCIL POST Expertise from Forbes Councils members, operated under license. Opinions expressed are those of the author. | Membership (fee-based) Jun 17, 2022, 09:15am EDT | Share to Facebook Share to Twitter Share to Linkedin Chitra Sivanandam is CTO for Reinventing Geospatial, Inc (RGi).

She oversees tech advancement for the defense and intelligence industries. getty The rise of artificial intelligence (AI) has influenced many sectors to embrace technology and apply some form of AI to their workflow. Because of this popularity, terms like “AI” and “automation” can start to lose their meaning—and their purpose.

Tech solutions should help organizations accomplish their goals more efficiently. Over the past several years, the federal government has increased its focus on utilizing more advanced technology for a wider variety of complex problems. As the ecosystem of solution providers grows to foster more innovation, the question becomes whether the government is asking the right questions upfront.

The Department of Defense (DoD) is one of the most secure federal departments that works with sensitive and intelligent data. The DoD has started to look at AI solutions and in 2020, the Congressional Research Office (CRS) asked the following questions for the department: What types of military AI applications are possible, and what limits, if any, should be imposed? What unique advantages and vulnerabilities come with employing AI for defense? How will AI change warfare, and what influence will it have on the military balance with U. S.

competitors? MORE FOR YOU Google Issues Warning For 2 Billion Chrome Users Forget The MacBook Pro, Apple Has Bigger Plans Google Discounts Pixel 6, Nest & Pixel Buds In Limited-Time Sale Event These questions are important to understand how AI can be developed and how it will be used in the defense industry. The nuance of military combat requires special attention to data in order to make informed and intelligent decisions. As an example, in our work with the Army, RGi has been working on experimental geospatial workflows for military missions including surveying, surveillance and reconnaissance.

We investigate how AI specifically can provide enriched and meaningful insights that will lead to more informed, accurate and timely decisions during combat operations. In the race to understand data and apply AI, have we over-prescribed it or intertwined the meaning with other techniques such as automation? Before we start applying AI to national security or defense, we should start asking targeted questions that will help us understand problems, identify baselines, iterate appropriately and measure along the way. As a CTO in the defense industry, I have found it beneficial to ask the following questions as part of my process to make decisions and prioritize tech advancement.

What problem are you trying to solve? We must understand which specific function we are looking to solve for before we start implementing solutions. Are we trying to analyze activity across reconnaissance photos, make more informed and accurate decisions in the field, optimize operational workflows—or is there some other problem to be solved? Almost all data problems can be aggregated as description problems, classification problems or prediction problems. Almost all problems have competing characteristics around cost, resources, timeliness, functionality, accuracy, precision and user experience (UX).

Focusing on problems and deeply understanding them allows us to think about design and constraints before advocating for a particular solution. Which tools can be used to solve the problem? Once you know which problem you are solving, you can figure out which tool is best suited for your organization. Do you know which workflows benefit from the technique? If you don’t have benchmarks and initial questions, you won’t know if you’ve just paid more for something that is fancier but not improving end results.

Learning algorithms are, by definition, hungry for more data. Sometimes the cost of implementing machine learning or deep learning may outweigh the benefit. Often, we think AI and automation are the same and if we just add “powered by AI” to the tagline, we’ll have a fully functional system.

That may not be the case. Often, we won’t know those tradeoffs without beginning to experiment. What are the differences between AI and automation? AI: One of the key components to AI is a learning-based feedback loop to better predict in the future.

A simplistic view of lifecycle management of AI looks like: Collect: Make data simple and accessible. Organize: Create a business-ready analytics foundation. Analyze: Build and scale AI with trust and transparency.

Infuse: Operationalize AI throughout a business. One must thoughtfully experiment with building the right measures. Data today will look different tomorrow because the data environment is constantly evolving.

It is imperative to design and iterate on a solution as well as the metrics and measures in order to keep up with the maintenance and management of AI. The more you invest in AI as a developing asset rather than a fixed good, the smarter and more human-like results will be. AI can only succeed when given constant data and learning behaviors.

Automation: Automation is typically rules-based and focused on alleviating specific pain points. Automation may be easier to set and forget because it’s not constantly learning and evolving. Strict rules and scenarios allow you to easily compute and accomplish tasks.

If you need to streamline workflows or collect information, automation may achieve this quickly and effectively. If we look at national security and the defense contractor point of view, automation tools should improve how an analyst performs their work. Automation can often improve inefficiencies when it comes to people working with large enterprises.

It can optimize based on workflow or structure; however, it cannot adapt without retooling. While automation may take less maintenance and learning than AI, organizations need to evaluate how and when they use automation versus AI and how they plan to nurture and improve these tools. Automation and AI live along a continuum, and most complex national security and defense problems require a thoughtful connection between the two.

In either case, if you set and forget your tools and your goals shift, your outcomes won’t align. How do you look at performance parameters of deep learning? Once you pick a technique to solve your problem, you will want to make sure you have clear parameters and measurements in place. If you don’t know what good looks like, you won’t know what better looks like.

Ask these questions to evaluate performance: Did this take more time? What did we improve? How did this benefit the outcome? Is this increasing what we want to measure? How do you improve or optimize using the tool? Once you evaluate and maintain performance, you will want to optimize. As you acquire new data and learn more about the UX impact, you can see where to iterate in order to build a function that mimics the behavior you need. You want to measure what is real, what isn’t and what has already been done to find the best solution.

You may or may not benefit from AI or automation, and that’s okay. The key is to know what is best for your organization. Once you ask the right questions, understand the problem and find out which technique is best for you, you can begin applying, evaluating and optimizing.

Have you been overprescribing AI? Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify? Follow me on Twitter or LinkedIn . Check out my website .

Chitra Sivanandam Editorial Standards Print Reprints & Permissions.