A.I. needs greater scrutiny and more thoughtful consideration of impacts on society

Amid the current furor over rogue AI bots and the concerns about establishing guardrails, we need to keep the other threats from AI in perspective, says Shankar Narayan. We also need to keep in mind the threats to privacy, the threats to democracy, the erosion of human skills and creativity, and the potential for malicious use of the technology.

"These technologies are inseparable from systems of power," Narayan said, "so it's not can AI help us but will the benefits of AI built by powerful and unaccountable tech companies exploiting our data for profit outweigh the costs?"

Narayan presented his talk on Artificial Intelligence at the Kenmore Community Club's Dinner and Discussion at Oct. 14, 2026, the nearly monthly discussions that feature experts talking about a variety of timely topics that society is facing today.

Following a brief recap of his Kenmore Club presentation in July, Narayan outlined the current state of AI highlighting the problems and limitations of the technology. He noted that the technology combines ubiquitous data collection, sensors, and machine learning that have normalized mass surveillance.

With algorithms built on aggregated, massive datasets that reflect existing human biases they amplify these biases with adverse effects in policing, hiring, job performance, housing and medical insurance. The technology companies' over reliance on these flawed models result in AI predictions based on historical correlations and such correlations do not imply causation. 

He drew comparisons of the historical pseudo-science of phrenology that tried to draw correlations between facial and cranial features and criminal behavior and current facial recognition software. Such software has been shown to have much higher rates of misidentification among people of color compared to white people. 

Other examples of poor performance of AI models included performance scoring of employees by employers and when used by insurance companies to judge medical necessity, or issue claims denials. In these cases AI has proven to be highly unreliable, resulting in massive error rates, wrongful denials, and mounting class-action lawsuits.

Moving on to efforts to regulate or otherwise impose guardrails, Narayan said most efforts have largely failed as the wealth of these tech companies exceeds the wealth of many countries, and the companies have used that wealth to successfully thwart most efforts to regulate the technology, while promising various forms of self-policing. 

Nevertheless some regulations have been implemented, but mostly in a patchwork manner that have largely made little difference in curbing the relentless rush to further develop the technology regardless of consequences. Narayan says that existing tools that could make a difference, such anti-trust, and environmental laws have been underutilized.

Narayan emphasized that the national conversation we need to have must center on questions about who owns the technology, who does it benefit and who will be hurt. What is the purpose of this technology, and how will it affect human creativity, human cognition and human morals.