Who Needs Transparency, Trust and Accountability? A Perspective on the Hidden Cost of AI

George Freeman, PMP, is a seasoned IT project manager and leader who has worked in the software industry for nearly four decades, including over 25 years of project management. He has significant experience and expertise in enterprise information systems, data, and business architectures, and is an advocate for “business and technical architectural awareness” among all project team members. Mr. Freeman has international and remote team experience, and has a passion for meta-modeling, domain-driven design, and “all things architecture.”

  Artificial Intelligence   Knowledge Management  

One of the few things we can all agree upon in life is the need for solid foundations, whether it’s related to engineering, our families, careers, finances, or belief systems. This underpinning is self-evident, as none of us wish to succumb to instability and find ourselves taking a fall. Hence, we build and seek transparent foundations whose exposed structures yield trust and accountability.

If the above assertion is true, how can we assign observable fact-based trust to the outputs of a function whose foundation is obscured (i.e., non-transparent)? One cannot, so we extend an alternative form of trust based on the visible knowns, stated narratives, and our due diligence. However, absent a transparent foundation, this lesser form of trust resides in the domain of skepticism.

Transparency à la AI
The phenomenon we know as artificial intelligence (or more specifically, generative AI) has entranced society through its novelty and sensationalism, creating a veil that reflects transparency to the masses while obfuscating elements of concern such as proprietary algorithms, control metadata (i.e., instructions), scoring, and decision models. In other words, transparency and AI have become distant cousins when, at one time, they were devoted companions.

Consider OpenAI, who, on the release of GPT-4, stated that its system was 40% more likely to provide accurate …

Please log in or sign up below to read the rest of the article.