Steven Johnson has written 13 books, on subjects starting from a London cholera outbreak to the worth of video video games. He’s been a tv presenter and a podcast host. He’s a keynote speaker who doesn’t have to name himself that during his LinkedIn profile. And for over a yr now, he’s been a full-time worker of Google, a standing that’s transparent when he badges me into the quest massive’s Chelsea places of work in New York to turn me what his staff has been developing.
It’s known as NotebookLM, and the best way to think about it’s as an AI collaborator with get admission to to your whole fabrics that sits to your metaphorical shoulder to lead you via your venture. NotebookLM was once soft-launched to a make a choice team previous this yr however is now to be had to all as an “experiment”—that’s Google’s low-risk strategy to see how the app behaves and the way we behave with the app.
Johnson discovered his strategy to Google by means of a lifelong obsession with device as a “a dynamic idea spouse,” a device to hurry up and fortify the inventive procedure. When he was once in faculty he become obsessive about HyperCard, Apple’s device that broke wisdom into chunks and allowed you to navigate an information-space via hyperlinks. It expected internet navigation ahead of the internet existed. “I fought mightily to show HyperCard into that dream instrument, but it surely wasn’t slightly in a position,” he says. He ultimately become an fanatic of Scrivener, a mixture phrase processor and venture organizer well liked by e book authors. (I’m a fan too.)
When Johnson were given get admission to to OpenAI’s GPT-3 textual content generator in 2021, he known that AI may just degree up a brand new era of idea gear. Oh, wait, he mentioned to himself, this factor that has all the time been at the back of my thoughts is now going to be conceivable. Situations unthinkable even a yr ahead of had been all at once at the desk. Johnson didn’t but know that Google now not best had identical massive language fashions, however was once already running on a venture very a lot in keeping with his considering. In Would possibly 2022, a small staff within the experimental Google Labs department cold-emailed Johnson. They arrange a gathering by way of Starline, a Google Labs venture that permits for eerily intimate in-person conferences. “I principally had a dialog with a hologram who mentioned, ‘You understand, this factor you could have been chasing your entire existence? We will be able to in the end construct it,’” Johnson says. He become a part-time adviser to the small staff, in the beginning sharing the workflow of a pro author. “Here is 4 or 5 engineers, here is a real creator, let’s simply watch him,” is how Google Labs head Josh Woodward sums up the method. Ultimately Johnson were given concerned within the building of the product itself and was once sucked in to the purpose of accepting a full-time gig. His identify at Google Labs is editorial director.
NotebookLM, at first known as Undertaking Tailwind, begins via developing a knowledge set of your supply subject material, which you drag into the instrument from Google Doctors or the clipboard. After the app has digested all of it, you’ll then ask NotebookLM questions on your subject material, because of Google’s massive language style generation—partially powered via its just-released improve Gemini. The solutions replicate now not best what’s to your supply subject material but in addition the broader normal working out of the sector that Gemini has. A essential function is that each solution on your queries comes with a suite of citations reporting the place precisely the ideas got here from, so customers can take a look at the accuracy of its output.
Google isn’t the one corporate envisioning merchandise that permit folks create customized knowledge units to discover with LLMs. At OpenAI’s developer day remaining month, the corporate offered customized mini-GPTs that may be tuned to a selected job. Woodward recognizes a “core similarity.” However he argues that NotebookLM focuses extra on bettering a workflow, and is geared to supply awesome accuracy in its outputs. Additionally, he says that the OpenAI merchandise have extra of a character, whilst NotebookLM is designed to haven’t any such pretensions.