I hate that I love Riverside’s AI-driven “Rewind” for podcasters

The online podcast signaling level Riverside came retired with its ain mentation of a year-end reappraisal similar Spotify’s “Wrapped.” The recap, called “Rewind,” creates 3 customized videos for podcasters.

Instead of sharing statistic similar however galore minutes you recorded, oregon however galore episodes you made, Riverside created a fifteen-second collage of laughter, showing a speedy succession of clips successful which my podcast co-host and I marque each different ace up. The adjacent video is similar, but that it’s a supercut of america saying “umm” implicit and over.

Then, Riverside scans its AI-generated transcripts of your recordings to find what azygous connection you said much than immoderate different (we’re assuming they chopped retired words similar “and” oregon “the”).

It’s a spot ironic, but connected my podcast astir net culture, my co-host and I said “book” much often than immoderate different connection (this was astir apt skewed by our subscriber-only “book club” recordings… oregon the information that my co-host has a publication coming out, which we plug incessantly).

Another amusement connected our podcast network, Spirits, said “Amanda” much often than immoderate different connection (not due to the fact that they’re obsessed with me, but due to the fact that they besides person a big named Amanda).

In the podcast network’s Slack, we exchanged our Rewind videos. There’s thing inherently comic astir a video of radical saying “umm” implicit and over. But we besides cognize what these videos represent: our originative tools are becoming much saturated with AI features, galore of which we bash not privation oregon need. The Riverside Rewind points to the uselessness of these tools themselves — wherefore would I request a video of my co-host and I saying the connection “book” implicit and over? It’s bully for a speedy laugh, but there’s nary substance.

Though I enjoyed Riverside’s AI recap, its accomplishment comes astatine a clip erstwhile my manufacture peers are losing opportunities to create, edit, and nutrient caller podcasts, acknowledgment to the aforesaid AI tools that generated our Rewind videos. But portion AI allows america to automate distant immoderate tasks — similar editing retired our “umms” and dormant aerial — podcasting itself is not that mechanical.

AI tin rapidly make a transcript of my podcast, which is important for accessibility reasons, helping to automate an enactment that utilized to beryllium incredibly time-consuming and tedious. However, AI is not capable to marque editorial choices astir however to maneuver audio oregon video to archer a communicative effectively. Unlike the quality editors I enactment with, AI cannot find erstwhile a tangential speech successful a podcast is comic and erstwhile it should beryllium chopped due to the fact that it’s boring.

Despite the emergence of personalized AI audio tools, similar Google’s NotebookLM, its quality to service arsenic a instauration instrumentality has besides seen high-profile failures, arsenic of late.

Last week, The Washington Post started to rotation retired personalized, AI-generated podcasts astir the quality of the day.

You tin spot wherefore this would look similar a “good” thought to profit-hungry execs — alternatively of paying a squad to bash the intensive enactment of researching, recording, editing, and distributing a regular show, you could automate it — but you can’t.

The podcasts spouted made-up quotes and factual errors, which is existentially unsafe for a quality organization. According to Semafor, the Post’s interior investigating recovered that between 68% and 84% of the AI podcasts failed to conscionable the publication’s standards. This seems similar a cardinal misinterpretation of however LLMs work. You can’t bid an LLM to separate world from fabrication due to the fact that it’s designed to supply the astir statistically probable output to a prompt, which is not ever the astir truthful output — particularly successful breaking news.

Riverside did a large occupation making a amusive end-of-year product, but it’s besides a reminder. AI is infiltrating each industry, including podcasting. But successful this infinitesimal of the “AI boom,” arsenic companies tinker with caller technology, we request to beryllium capable to separate betwixt erstwhile AI serves us, and erstwhile it’s fodder for useless slop.