Good morning. I’m sending this early to target your commute. Listen to this in your car, or in your AirPods, or however you consume your info-on-the-go.
This is the latest episode of Two AIs were chatting, a series of AI-generated podcasts that I have created from my original journalism using Google’s Notebook LM.
This episode has taught me how the AI can go awry, and what to do about it, so it’s perhaps the most valuable experiment in this series so far.
This is what happened…
Initially, I asked the AIs to create a podcast by analysing the text of this article I wrote about Meta’s upcoming FTC antitrust trial earlier this week.
But Notebook LM went on some tangents that came perilously close to being outright wrong. My analysis suggests that my commentary in the article led it astray.
So, I went back to the drawing board.
Instead, I uploaded the entire 92-page PDF of the court summary. I copied it in raw, and entirely unedited. Here it is if you want to read it.
Then, to set the AIs’ direction, I added a prompt to place ad breaks, and for the hosts to conclude with their thoughts on the implications for the news industry.
It did an amazing job, given the input was a dry legalese court ruling with precious little colour for it to springboard from.
In my view, it did a near human-level job. The best yet.
And the AI hosts’ attempts to consider the implications for the news industry were doubly impressive, given the court documents provided them with very few pointers.
I don’t know whether Notebook LM is learning from my prompts, or being optimised by Googlers out of my sight, but it appears to be improving in leaps and bounds.
That creates interesting opportunities for news companies seeking to convert staid old word-and-picture articles into new and more valuable formats for the AI era.
Please don’t be offended that I am using generative AI to report the news.
I know it is challenging, but like it or not, it will be part of all our news futures, whether for delivery or discovery.
Testing it now; putting it through its paces; trying to break it and discovering its limits is therefore worthwhile work for those of us pushing the boundaries.
Have a listen, and tell me what you think 💬
Share this post