2021 will be a big year for AI implementation and adoption, but we must first be able to trust it. Trustworthy AI begins by building a foundation with five critical cornerstones at the core, says Josh Elliot, the head of operations at Modzy.
When describing 2020, the word “predictable†likely doesn’t come to mind. But the pandemic has shown that artificial intelligence and automation can truly help businesses in times of need. AI helped modernize critical healthcare processes in both patient care and administration; it helped drive manufacturing challenged with decreased on-site staff; and provided personalized recommendations and better virtual customer service in the retail and ecommerce industries. 2021 will likely welcome more of this emerging technology.Â
It is also very clear that building trustworthy artificial intelligence is a necessity in 2021 and beyond. As AI becomes increasingly involved in decisions that impact our lives, we must be able to explain why it’s making decisions.
Looking back at the past year and forward to what’s to come, a few key AI predictions emerge:Â
1. An AI Regulation Evolution
Just a few weeks ago, the U.S. saw an executive order to federal agencies on the adoption of AI to facilitate the delivery of services to citizens and promote public trust in AI. Then we saw the U.S. House of Representatives approve a plan to create a national AI strategy. We can expect heightened scrutiny and even increased regulation (already the case for several industries such as financial services, healthcare, etc.) as AI becomes more prevalent.
Without a proper foundation, many may find themselves unprepared to plan for or meet new and changing requirements. Teams need to think about ways to create a foundation for their AI capability that promotes flexibility but also accountability, including building and selecting tools that enable them to go back and look under the hood to see who did what and when – because this is crucial to establishing trust and confidence for AI decisions.Â
Learn More: Building Trustworthy AI in 2021 and Beyond
2. AI Governance Will Be Everyone’s Responsibility
Even though governance takes a long view, it is imperative that businesses begin to provide clarity around scope, roles, and decision authorities as it relates to the research, development, deployment, and monitoring of AI and machine learning models under management.
Seasoned AI practitioners know that proactive governance not only protects them from decisions made, but when done right, it can also spur innovation, optimize resources, reduce risk, and position businesses to quickly respond to change.Â
3. Model Security Is No Longer an Afterthought
Just as advances in AI systems are racing forward, so are opportunities for adversaries. Businesses must begin to adopt proactive adversarial defenses into their AI security posture or risk creating opportunities ripe for attackers.Â
As this comes to light, AI security will become a building block incorporated right from the start. Security will undoubtedly become a core feature of AI tools, but mitigations must also be designed and integrated at the model level. Otherwise, it will come back to haunt with the possibility of an AI security Opens a new window incident.
Learn More: The Role of IoT in Office Buildings in 2021
4. Widespread Adoption Will Start With ExplainabilityÂ
The widespread adoption and acceptance of AI requires it to be trustworthy. This comes from one key technique: explainability, which extends beyond basic model transparency to understanding how a model operates. AI explainability can also pinpoint features of a model that may generate bias, enabling retraining to mitigate negative effects.
Explainability, governance, and security are all interconnected. Understanding how machine learning reasons against real-world data helps build trust between people and models. Without explainability, there will never be true traceability, transparency, or confidence in AI models for critical decisions or business functions. The next phase of widespread AI adoption is reliant on the ability to understand and trust AI-enabled decision making.Â
5. The Question of Build Versus Buy Won’t Be Black or White
Businesses can’t afford to ignore what AI will bring in the future. Most will be asked to choose between building capability in-house or buying it externally, but organizations can use a combined approach to drive tangible value and outcomes. This includes starting with open-architecture solutions which allow for easy integration of or swapping out new tech as it becomes available or as new requirements emerge.
For many, it will make sense to buy and integrate most AI tools rather than build something bespoke, especially tools that embed governance, security, and explainability into an existing AI tech stack versus spending the time and resources to build it themselves. It’s difficult to overestimate the cost and complexity of the “last mile†of adopting AI and generating real value – a “buy not build†approach will be the way forward for many.
While it’s no surprise that 2021 will be a big year for AI implementation and adoption, focus will need to be on explainability, security, and governance — the cornerstones of trust in AI. It’s all about building a foundation that allows AI to evolve and adapt over time with these critical cornerstones at the core.
Did you enjoy reading this article? Let us know your thoughts on LinkedInOpens a new window , TwitterOpens a new window , or FacebookOpens a new window . We would love to hear from you!