Justin Oberman Questions LinkedIn’s AI Content Strategy, Advocates for Human-Centric Algorithm

In a recent LinkedIn post, Justin Oberman discusses LinkedIn’s new initiatives to combat what he terms “AI slop” on the platform. While many users are celebrating the move, Oberman expresses skepticism about the chosen solution, arguing that it fundamentally misunderstands the root of the problem.

Oberman’s central critique is that LinkedIn’s plan to tackle AI-generated content involves using more AI. He likens this approach to a concerning scenario:

“Which, when you really think about it, is a lot like hiring a fox to audit the henhouse and being relieved when he says everything looks fine.”

This analogy highlights his view that the platform’s chosen method is inherently flawed, as the tool intended to solve the problem is part of the same category of technology that created it.

The Inaccuracy of AI Detection

A significant point of contention for Oberman is the reliability of AI detection software. He points out that these tools are far from perfect and often provide inaccurate assessments of content.

“As most people know, AI detection software has the accuracy of a Magic 8 Ball that’s been left out in the rain.”

This assertion suggests that relying on AI to police AI-generated content is a precarious strategy, prone to errors and misjudgments. According to Oberman, this is a classic instance of engineers attempting to solve complex human issues with technological fixes that miss the mark.

Distinguishing Bad Content from AI Content

Oberman argues that the core issue is not the existence of AI-generated content itself, but rather the proliferation of low-quality, “bad” content, which AI has merely amplified. He believes the focus should be on content quality, not its origin.

The Human Element in Content Curation

Instead of doubling down on AI solutions, Oberman advocates for a return to a more human-centric approach to content algorithms. He suggests that the platform should prioritize human judgment in identifying valuable content.

“The problem is that people hate bad content. There has always been bad content. AI has just amplified it.”

In Oberman’s view, the ideal system would involve human users engaging with and evaluating posts, with the algorithm learning from these human signals. He posits that this method was more effective in the past and should be reconsidered.

“So the fix for AI slop isn’t more AI. It’s going back to the old human-centric algorithm. The one where humans saw a post, decided if it was any good, and the machine took notes from there.”

By returning to a human-centric model, Oberman suggests LinkedIn can better address the underlying problem of poor-quality content, rather than relying on an AI-driven solution that may prove to be ineffective and overlooks the nuances of human content consumption and creation.