The AI-generated hosts don't have names, but they're a team of men and girls who sound great and described my features in a way that just about made me blush. In my case, I included the net address of my online bio, but you can even load PDFs, YouTube videos, audio files, Google Docs or Google Slides.
When you load an internet address, the 2 fake hosts not only inform you what they learned out of your biography, but additionally have a look at all the website. I learned this by providing links to 2 almost equivalent biographies. The one hosted on Larrysworld.commy personal website, contained references to a lot of my newspaper articles. The one at ConnectSafely.org focused totally on my and the organization's online security work.
The fake moderators injected some humor into the conversation and made fun of a few of the things the AI had learned about me on my website. They also touched on some things that made them sound like live podcast hosts.
Comparison
To get a feel for the way it’d describe someone with a very different job, I downloaded a sample electrician resume. The two fake hosts said they’d base their podcast on his resume. They commented positively on his qualifications and talked just a little about their very own struggles with electrical problems.
Like my biography, this podcast was based on the loaded material published in each cases by the person described. These fake hosts apparently didn't do any cross-research to seek out out if the knowledge was true.
Once you've created the podcast, you’ll be able to download it as an MP3 file and share it with others. When you do that, you find yourself telling the listeners that it’s generated by AI. Ethics and honesty aside, every podcast uses the identical two fake hosts. They sound real, but eventually listeners will discover they're not.
You can try NotebookLM free of charge at notebooklm.google.
Practical applications
Aside from being a parlor trick, there are also some practical uses for this tool. I loaded a Google presentation and heard that the fake hosts gave me an accurate summary. And even the biographies, while subservient, corresponded to what the AI tool had gleaned from the source material, although that doesn't mean the source material was accurate.
I also included the address of a PDF from the Pew Research website a couple of recent survey of public opinion on President-elect Trump's plans for the country. The AI-generated podcast, which lasted 14 minutes, was based on the facts from the survey report, however the fake moderators also added their evaluation. According to the survey results, they covered each positive and negative features of Trump's characteristics, especially in the general public's perception.
NotebookLM also provided written materials that were more detailed and useful to anyone who wanted to know the survey. These included a 115-word summary and AI-generated pages on key topics, “Views on Trump and the nation after the election,” and “Public confidence in a smooth transition,” amongst others.
Based on the PDF, NotebookLM also allowed me to create an FAQ, study guide, table of contents, timeline, and briefing document. The study guide asked questions on key parts of the document and provided answers based on the PDF. The timeline focused on the events before and after the 2024 US presidential election, based on the knowledge within the document. The 740-word briefing document covered key topics.
In case you're wondering, it concluded: “Republicans greet Trump's victory with enthusiasm, Democrats view it with deep concern.”
Answers your questions
In addition to the materials provided, you can even ask questions on the document. What I like about it’s that the answers to your questions are based on the content of the document, unlike many generative AIO tools that depend on all the web. The undeniable fact that the knowledge may be very limited is each a bug and a feature. This is a mistake because there may be probability you won't receive a response. This is a feature because what you get is predicated on the document or web page you’re analyzing. So for those who trust this source, you usually tend to get a solution you’ll be able to trust. However, while NotebookLM interprets answers when it doesn’t know them, it makes clear the difference between what it knows from the document and what it concludes from it.
Needs work
NotebookLM remains to be in its infancy and wishes work. For one thing, the interface is a bit confusing each on the internet and on a smartphone. I needed to click around a number of times to seek out what I needed, including the power to create this podcast or find summary information. Additionally, like all generative AI products, it just isn’t infallible. There's a disclaimer on every page that claims, “NotebookLM may make mistakes, so please double-check.” Despite its rough edges, this is really a remarkable tool and example of how generative AI goes beyond the unique offerings OpenAI, Microsoft and Google can transcend.
Originally published:
image credit : www.mercurynews.com