The online podcast recording platform Riverside recently launched its own AI-driven year-end review, dubbed "Rewind," drawing parallels to Spotify's popular "Wrapped" feature. While the personalized recaps offer a moment of amusement for podcasters, they also underscore a growing ambivalence about the increasing entanglement of artificial intelligence with creative culture and content production.
Unlike traditional year-end summaries that focus on statistics like total recording minutes or episode counts, Riverside's Rewind generates three custom videos. One compilation features a rapid succession of laughter clips from podcast sessions. Another creates a supercut of hosts repeatedly uttering "umm." The third video identifies the single word spoken most frequently by podcasters, excluding common filler words like "and" or "the," by analyzing AI-generated transcripts of their recordings.
For instance, on the author's podcast about internet culture, the most frequent word was "book," likely influenced by subscriber-only book club recordings and a co-host's upcoming publication. Similarly, another show on the same network, Spirits, found "Amanda" to be their top word, attributed to having a host named Amanda.
These Rewind videos, particularly the "umm" compilations, often elicit genuine laughter among creators. However, they also serve as a stark reminder of the pervasive integration of AI features into creative tools, many of which are perceived as unnecessary or unwanted. The author questions the practical utility of a video showcasing repeated words, suggesting such AI-generated content, while momentarily entertaining, lacks substantial value.
The Dual Nature of AI in Content Creation
The arrival of Riverside's AI recap coincides with a period where many industry professionals are experiencing reduced opportunities in podcast creation, editing, and production, largely due to the very AI tools that powered Rewind. While AI can automate certain mechanical tasks, such as generating transcripts for accessibility or removing filler words and dead air, the essence of podcasting remains deeply human and non-mechanical.
AI excels at quickly generating transcripts, a crucial feature for accessibility that significantly reduces tedious manual work. However, AI currently falls short in making nuanced editorial choices essential for effective storytelling. Unlike human editors, AI cannot discern when a tangential conversation is genuinely funny or when it should be cut for being boring. It lacks the contextual understanding and creative judgment required to shape compelling audio and video narratives.
High-Profile AI Failures in Media
Despite the rise of personalized AI audio tools, their application as creative instruments has recently seen notable failures. Last week, The Washington Post began rolling out personalized, AI-generated podcasts summarizing daily news. This initiative, likely driven by a desire for profit through automation, aimed to bypass the intensive human effort involved in researching, recording, editing, and distributing a daily show. Yet, the results were problematic.
The AI-generated podcasts were found to contain made-up quotes and factual errors, posing an existential threat to a news organization's credibility. According to Semafor, the Post's internal testing revealed that between 68% and 84% of these AI podcasts failed to meet the publication's journalistic standards. This highlights a fundamental misunderstanding of how Large Language Models (LLMs) operate: they are designed to produce the most statistically probable output to a prompt, which does not always equate to the most truthful or accurate information, especially concerning breaking news.
Distinguishing Useful AI from "Slop"
While Riverside has successfully created an entertaining end-of-year product, it also serves as a potent reminder of AI's pervasive infiltration across all industries, including podcasting. In this era of rapid AI advancement, as companies experiment with new technologies, it becomes increasingly vital to distinguish between AI applications that genuinely serve human needs and those that merely generate "useless slop." The challenge lies in leveraging AI for efficiency and accessibility without compromising the integrity, creativity, and human touch that define quality content.
