Griffin on Tech: Google’s NotebookLM and the US ‘podcast election’
As I’ve been scrolling through Youtube videos over the past couple of weeks I've been surprised to see US presidential election hopefuls Donald Trump, JD Vance, Tim Walz and even the interview-shy Kamala popping up in video podcasts.
Busy jetting around the country to political rallies in swing states as they try to appeal to voters, why are these politicians devoting precious time to often lengthy podcast interviews, rather than stacking up more media interviews likely to reach a larger audience via CNN or Fox News?
I was surprised, for instance, to see Trump appear on Lex Fridman’s podcast, where the host seemed fascinated by Trump’s psychology more than anything. If Trump had expected it to turn into a sort of therapy session, I doubt he’d have agreed to go on the pod. But the podcast has notched up over 5 million views. There’s a similar trend here, with the likes of Between Two Beers and Grey Areas with Petra Bagust attracting high-profile guests.
As Bloomberg reported this week, 135 million Americans listen to a podcast monthly, while 98 million listen weekly, according to Edison Research. That’s a big jump since 2016 when Trump first ran for president.
But there’s another important factor explaining the attractiveness of podcasts, particularly to politicians.
“A few different studies suggest people trust podcasters more than other media personalities, particularly when it comes to hearing about the news,” Bloomberg’s Ashley Carman explained this week.
A matter of trust
“Research from Deloitte last year found that 75% of surveyed respondents said they agreed with the statement, ‘I trust the podcast hosts I listen to’. Pew Research found last year that most people who get news from podcasts either say they trust that news more than the news they receive from other sources (31%) or trust it about the same (55%).”
In an era when trust in the mainstream media is declining virtually everywhere, people are placing higher trust in podcast hosts, many of whom have no media background. With podcasts able to appeal directly to listeners with product endorsements and product codes, that higher level of trust also translates into dollars, as listeners and viewers take the hosts’ advice on what to buy.
Maybe that’s why we seem to have reached ‘peak podcast’. The tech world in particular is full of them - I have my own own, The Business of Tech, which I co-host with Ben Moore. This week we looked back on ten years of deep tech investment in New Zealand and heard from Sir Peter Beck about Rocket Lab’s deal with NASA to scope out a mission to retrieve rock samples from Mars.
My favourite podcasts include Kara Swisher’s On, Hard Fork from the New York Times, Techemem’s Ride Home, and Decoder from The Verge.
I can tell you from having made hundreds of podcasts that it's an exhausting job trying to line up guests, research topics, produce the thing and then promote it. But that may be all set to change.
As my mate Mike ‘MOD’ O’Donnell wrote in The Post this week, Google’s AI audio tool NotebookLM is doing an incredible job of auto-generating convincing-sounding podcasts. You can feed in research documents or news reports, and two hosts will have a natural-sounding discussion about it and do an uncanny job of getting to the heart of the matter.
I don’t see Joe Rogan, the reigning podcast king, losing his job anytime soon. He’s too idiosyncratic to be replaced by an AI bot, though one that is trained on his past shows will probably be able to do a fair job of replicating Rogan before long. The AI works well when it is generating both sides of the discussion, but I haven’t seen a compelling example of one holding a convincing conversation with a human podcast guest in real time.
Auto-generated tech tutorials
But as MOD explains, NotebookLM is a boon for any organisation that wants to easily communicate and gain insights into what it is doing.
“So if you feed in your business plan, trading results from the last month and the year previous, it will produce a broadcast quality podcast that not only summarises your delivery against plan; it also scores it,” he writes.
“Remarkably it only takes about four minutes of ‘thinking’ to then produce that 10 minute podcast.”
Think of fast-growing tech companies that want to keep everyone on the same page about what the company is doing. The management and board could turn their briefing papers into short podcasts for internal communication. A complex new product launch could be explained and strategised based on the documentation. Notebook LM could largely automate training tutorials.
The result, according to MOD, is a tool “so powerful that it could kneecap entire segments of the podcast industry”.
“But for small creators, it’s a godsend, levelling the playing field by allowing them to produce polished, professional content without the overhead of staff or time-consuming processes.”
Another example of the speed at which generative AI is developing and the disruptive potential it has across numerous industries. Maybe by the time the next US election rolls around, the AI bots will be dominating the podcast conversations.
