When does an organization need to consider building a solution using AI as opposed to conventional Software Engineering?

The primary question an organization should examine while innovating in this area, is whether the customer problem they are trying to solve really requires an algorithmically and operationally complex approach like AI, instead of a relatively more wieldy technology. For starters, a solution built using AI doesn’t usually give a single concrete answer to a given question or query, instead it gives probabilities representing which set of answers are likely to be correct. Furthermore, this process is not bullet proof as it lacks immunity to producing false positives and even missing the correct answers altogether.

Sometimes the newest technological advancement isn’t the most effective or suitable solution. AI, while becoming an increasingly hot topic in tech and the media, isn’t always the best choice. Instead, innovators need to evaluate the most effective technology options without overcomplicating the development process. Sometimes, the smartest choice is based on tried and tested legacy technology.

Imagine a law enforcement agency needs a dashboard containing crime statistics in a given geographical area. We can provide these reports by utilizing conventional software, including databases and the appropriate query language, to calculate those crime analytics. This is a deterministic approach by definition and as such it would be counterproductive to use AI for the same exact purpose. However, an algorithmic approach would be called for in a more complicated case, such as when a law enforcement agency wants a tool to optimize the allocation of patrol officers over a given geographical area based on myriad set of factors like time of day, crime rate and personnel availability. In this case AI might be a potential candidate for a solution that is worthy of investigating.

To summarize: while conventional software, analytics and AI algorithms maybe overlapping, they all have different purposes, risks and associated development costs. Making it is essential to educate an organization on what AI is and what it isn’t.

Furthermore, AI is quite unsustainable in the long-term unless you can apply it in a responsible manner. Irresponsible AI which doesn’t take into account the impact on society, is easy to implement for small problems, but it’s not scalable. On the other hand, using science and engineering responsibly, although time consuming, provides a competitive advantage because people who ignore ethics will erode both trust and performance over time.

A commitment to innovation, positive impact, and community

Ethics-by-Design is an organizational approach to the responsible use of technology, and as such there needs to be deliberate steps organizations take to implement it. This development cannot be done in a vacuum; its social impact ought to be considered from the beginning. Organizations that use technology ethically consider not only the benefits but also the potential harm to society at every stage of their decision-making process, which is especially important for Axon and how we serve our public safety customers.

To define this impact, Axon has an independent AI Ethics Board that provides expert advice and guidance during product conception and subsequent development. The Ethics Board has developed a framework for analyzing the ethical costs and benefits associated with new products. This is an important framework within Axon, even for products that do not use AI.

To create with an Ethics-by-Design framework, it is important that the team composition reflect the diversity we see in the world. A diverse, and importantly, inclusive team is required so you can apply a spectrum of opinions, genders, ethnicities, income levels and skill sets to a challenge. This allows the team to consider all angles when working with new ideas and products in a way that truly creates groundbreaking technology.

A diverse team that espouses and builds upon those ethical framework guidelines leads to innovation focusing more on how products can positively impact a community at large. With AI employed into the right product, in the right way and at the right time, it can become a great catalyst that helps address tangible societal issues.

Innovation starts with building a world-class AI team

As we grow our AI team, Axon aims to attract a diverse group of talented engineers and scientists at all levels, who are as excited as we are about our Ethics-by-Design ethos. A great candidate for our team starts with the eagerness to understand the customer problem as they responsibly develop an AI solution to address it. We seek candidates who are interested in developing something meaningful and impactful, such as how to operationalize AI in a way that preserves privacy, maintains fairness and avoids bias. It’s not enough to implement the latest and greatest approaches in AI when building new products; a successful candidate has to show they are able to take these ethical considerations into account and address the corresponding design challenges at scale with the team.

At Axon, we focus intently on real problems while developing end-to-end solutions. This requires a multidisciplinary team with the right balance of AI Scientists, ML Engineers, Software Engineers and Product/Program Managers. As the team evolves, we’re looking to include subject matter experts in every AI modality to build proprietary models that include audio, video, text and images for our products.

In addition to building the right team, it is also imperative to keep pace with the ever-changing AI advancements. Still, as we add in novel AI layers to some of our products, we do so with the customer and impact to community in mind. Keeping Ethics-by-Design in the foreground of all we do helps Axon get closer to its goal of providing more effective tools in service of our mission to Protect Life and Preserve the Truth.

About the Author:

Yasser Ibrahim is Axon’s SVP of Artificial Intelligence and leads Axon’s AI Team from the Seattle office. Prior to joining Axon, Yasser served as Amazon’s head of distributed machine learning at Alexa AI. He led applied science and ML engineering teams developing state-of-the-art deep learning frameworks and distributed algorithms powering Alexa’s large scale ML experimentation and production platforms. He also led multidisciplinary engineering and science teams building computer vision systems and computational imaging algorithms for Amazon Go, the world’s first “Just Walk Out” store and technology platform. At Microsoft, Mr. Ibrahim led computer vision system teams to deliver technologies such as Pixelsense, and previously developed flight control and autopilot systems for flight simulators at CAE Inc., in Montreal.