In a recent episode of the podcast, Joe Rogan asked OpenAI founder Sam Altman if society should be worried about fake AI content. The episode’s question, “How will we know what is real and what is not?” was particularly appropriate because the entire hour-long conversation had been generated by AI.
According to whomever you ask, artificial intelligence is either a threat to the creative industries, a technology that will advance humanity, or just a dumb toy. The creator of The Joe Rogan AI Experience – a kind of AI generated fanfic YouTube podcast which depicts imaginary conversation between Rogan and new guests every episode – believes in all three. He brought up all three during a recent conversation with Hot Pod.
Hugo is a creative executive at a VFX and ad agency based in Sydney, Australia. He asked that we only use his name as a way to protect his privacy. Hugo readily admits that his YouTube show is not meant to be a real threat to Rogan, or any other podcasters.
Hugo released the podcast’s first episode a month earlier. The AI clone warns, “This is fiction — for fun only — don’t make your little TikToks to fool the world into thinking I said something I didn’t say.” The fake Rogan has a ChatGPT scripted conversation with an equally fake Altman. The podcast that resulted was so eerie, Rogan himself addressed it.
Rogan, the real , wrote on Twitter: “This is about to get slippery.”
Hugo had the idea for The Joe Rogan AI experiencein February when the social media discussion around ChatGPT was at its peak. Hugo had been following AI developments over the last year, and watched “a few YouTube videos about voice-cloning”, when the idea came to him.
He thought, “What if I used one of these voice cloning platforms to create a whole the Joe Rogan Experience Podcast with ChatGPT?”
He found demos of Joe Rogan’s voice on a few AI voice software companies after a Google search. Unfortunately, many of them were not that good or publicly available. After a few weeks of searching Hugo finally found a platform for text-to speech that could clone any audio clip. He used the platform to clone the voices of Rogan, Donald Trump and Andrew Tate.
AI supporters claim that it is cheaper and faster to produce work than by humans. The work of creating a script that can be used on ChatGPT and then transferring it to a text-to speech program, as well as editing audio, takes hours.
AI supporters claim that it is cheaper and faster than human labor
Hugo wrote in an email: “The first […episode] only took me one week. The second took me one day and a quarter, and the third took me over three weeks.” It depends on many factors, but my critical opinion is the most important. If I don’t think it is good enough, I will put in a lot more effort to make it better.
Hugo spent hours gathering audio samples and perfecting the AI voices he created. Hugo spent many hours in the later episodes to perfect his voice clone. There were still moments where his speakers sounded like an AI or robot. It took a lot of work to edit the script so that it sounded like a human conversation.
Hugo said that “[…ChatGPT [is not really that great yet in terms of replicating speech patterns. ChatGPT does not use voice recordings, even though Rogan’s countless hours are available online. Hugo says that even feeding it transcripts from interviews is not foolproof.
When you listen carefully to a two-person conversation, you will notice that they often talk over eachother. He said that they answer each other, sometimes with very short answers.
Hugo’s AI podcast takes a lot longer to produce than the Joe Rogan himself. Why would you make an AI-generated version of one of the world’s most popular podcasters?
Hugo explained, “I wanted to make a version of Joe Rogan’s podcast that never happened or hadn’t yet occurred.”
I wanted to create an episode of the Joe Rogan Podcast that never happened, or hadn’t yet occurred.
Hugo thinks that the main attraction of the show is to create hypothetical conversations between Rogan, and other people. He wants to create episodes with dead guests such as Steve Jobs and Abraham Lincoln.
I listened to segments from The Joe Rogan AI Experience for about an hour. The quality of Hugo’s edits and the cloning of voices is impressive, but the content is lacking. The dialogue is bland and wordy in the way that we’ve come to associate with AI generated text.
“I can certainly understand people’s fear, because I personally don’t think that podcasts or real conversations will ever replace them. But I do not believe we are anywhere close.” Because I don’t think it’s super close — i mean — it’s close. If you have listened to Joe Rogan’s podcast for a while, you will know that [… [The AI Experience] does not feel real. It lacks emotion and the correct intonation,” said Hugo.
the first episode The Joe Rogan AI Experience attracted close to half a millon views, but subsequent episodes have seen fewer viewers. The first episode of The Joe Rogan AI Experience was viewed by close to half a million people (probably due to Rogan tweeting it), but viewership has dropped in subsequent episodes.
Other reasons exist to be concerned with the rise of AI podcasting. Digital fakes can make Rogan, or any other podcast host, say something offensive. Hugo claims he can already see how AI could replace certain jobs within the VFX sector.
It is not yet clear whether humans prefer AI recreations over the real thing.
It doesn’t necessarily mean that live-hosted programs are over. At one point I mentioned a Vulture Interview with Comedy Bang! Bang! Scott Aukerman is the host of this podcast, and he points out that podcasts are unique because most entertainment today has a prescribed format. It is rare to hear people laughing and “joyful in the moment” on podcasts.
Hugo does not believe that AI can replace spontaneity.
You know that thing you do when you are talking to a friend and then an hour later, you find yourself in a totally different place than where you began? You realize that AI cannot do this. “The way the brain works is very unpredictable.”