The most recent innovation in synthetic intelligence has made podcasting simpler, however made differentiating between people and synthetic intelligence tougher. A brand new text-to-voice characteristic of Google‘s AI-powered software program NotebookLM can remodel written phrases, akin to information articles or weblog posts, into an artificially generated dialog between two AI hosts.
The result’s “so pure and sensible that there isn’t a method you’d imagine you were not listening to 2 actual folks speaking”, stated TechRadar senior editor Graham Barlow, who generated an eight-minute podcast from a weblog put up. After listening, “the world merely wasn’t the identical any extra; I now not had confidence that I may inform what was actual and what wasn’t”.
‘That is the satan’s work’
Google launched NotebookLM final 12 months as a analysis instrument, to assist customers summarise notes and different paperwork. It already makes use of Google’s AI mannequin Gemini 1.5, stated The Verge. This newest experimental characteristic is “like an audio model of that”, having AI hosts “summarise your materials, make connections between matters, and banter forwards and backwards”.
Subscribe to The Week
Escape your echo chamber. Get the details behind the information, plus evaluation from a number of views.
SUBSCRIBE & SAVE
Join The Week’s Free Newsletters
From our morning information briefing to a weekly Good Information E-newsletter, get the very best of The Week delivered on to your inbox.
From our morning information briefing to a weekly Good Information E-newsletter, get the very best of The Week delivered on to your inbox.
However the speedy emergence of AI-generated audio has sparked fears that it may upend the human business. “I’m not in any respect spiritual, however after I found this instrument, I needed to scream, ‘That is the satan’s work!’,” stated ZDNET contributing editor David Gewirtz. The standard “verges on making creators and content material producers like me start to really feel the warmth”. The “voice constancy”, the suitable use of colloquialisms, the “utterly natural nature of their banter and the truth that there even was banter”… the ultimate product is “indistinguishable from an actual broadcast”.
But it surely is also a “watershed” second in democratising the business. It price Google “billions of {dollars} to show my article right into a podcast”, stated Gewirtz. “But it surely price me nothing. It additionally took moments. That is an enormous discount within the barrier of entry to content material manufacturing.”
Future variations may develop into extra sensible and more and more personalised. Presently, NotebookLM generates a dialog solely between a male and a feminine host with commonplace American accents. However think about for those who may select the kind of speaker, the fashion, the accent, and “possibly edit a bit of the AI-generated script”.
‘One-to-one relationship’
NotebookLM is not the one software program providing human-free podcast companies. Simply hours after exploding pagers and radios killed dozens in Lebanon, a “peculiar new present appeared on podcast apps”, stated Wired. “‘Pager Protocol’ is not in regards to the assaults aimed toward Hezbollah members or the Israeli operatives believed to be behind them. It is an ongoing AI-generated 10-part sequence created to quickly flip these unprecedented occasions right into a fictional podcast.”
Launched on the day of the explosions, the podcast was created by feeding details about the information occasion into Claude, a ChatGPT-like generative AI instrument. Claude “spat out a top level view of a narrative”, stated Mark Francis, co-founder of Caloroga Shark Media, the corporate behind the podcast. The staff wrote a script and “put it again into the AI for much more massaging”. The script was then fed into AI voice-generating software program to create an audio model narrated by an AI host.
However others warned of dangerous implications for the business. The true energy of podcasts is “the host-audience relationship”, stated Jason Saldanha, chief working officer of the non-profit digital radio firm PRX. Probably the most profitable have a “one-to-one relationship with their audiences”.
An AI host might sound tempting within the brief time period, to avoid wasting time and sources, however “flooding the market with content material to get the bottom degree of engagement” isn’t a “long-term technique”, he informed Wired. “It is gross and it is unhealthy, and, in the end, you are chopping off your nostril simply to make an additional greenback.”
