There has arguably never been a more exciting time in AI. Alongside the arrival of so much promise and potential, however, more attention is due to the ethics and trustworthiness of this powerful technology. It is a question of not just what can be done with AI but how it should be done—or whether it should be done at all. Because AI has been developed to this level of maturity, we must now grapple with some of the more complex ethical considerations concerning AI.
What does it mean for AI use to be ethical? How do we know if we can trust the AI tools we use?
The trajectory of AI can be conceived along three streams: research, application, and trust and ethics. Research concerns data science and AI engineering, which has characterized much of the history of AI to date. For several decades and accelerating, we have seen AI application in growing volume, impacting every industry and touching every organization.
What is emerging today, as AI scales, is the increasingly important third stream— determining how to use this technology in an ethical way such that we can trust it. For AI to reach its fullest and greatest potential, it must be trustworthy and aligned with enterprise values and human ethics.
There will never be just one approach or universal solution for AI trust. Just as every business is unique, so too are the tools they use. Each company must identify what trustworthy AI means for the enterprise and then design, develop, and deploy to that vision. In this way, striving for trustworthy AI is not a one- time engagement but a feature of the overall AI lifecycle.
When ethical considerations in AI are accounted for, addressed and treated in practice, stakeholders are informed and engaged. Documentation is comprehensive. Processes and strategies reflect business priorities and values. A sociotechnical system emerges that allows the organization to move confidently into the dazzling future with AI. This whole-of-enterprise activity requires consistent attention, and the reward is AI that can be trusted to perform as intended with maximum value and minimal drawbacks.
In pursuit of this vision, the challenge becomes defining trustworthiness, identifying the relevant fields and concepts, mobilizing people, amending processes, and implementing new technology. This is no small endeavor, and it is one that can be aided by the insights and leading practices explored in Trustworthy AI.
