Red Flag: Google’s CoHosted-Podcast Maker Not Always Accurate
Google’s new NotebookLM — which has gone viral with its ability to auto-script and auto-produce a co-hosted podcast in minutes — is unfortunately also very good at making things up.
The new AI research tool — which uses two, extremely natural-sounding robot voices to discuss text, audio or video that you input into NotebookLM — is currently wowing the Web.
But reviewer Matt Derron has found that like all generative AI tools, the two robot voices can occasionally get it wrong.
“In trying to make the hosts sound natural with their back-and-forth banter, it sometimes misses the original intent of the source material,” Derron says.
Of course, that flaw is no reason to toss NotebookLM on the digital trash-heap: The AI tool’s ability to auto-generate an extremely lively, co-hosted podcast using numerous forms of media is still truly remarkable.
But to be on the safe side, you may want to do a little editing on any co-hosted podcast that’s auto-generated by NotebookLM before publishing it live — or suffer the consequences.
For an excellent, in-depth look and critique of how NotebookLM auto-produces co-hosted podcasts, check-out Derron’s in-depth video.
In other news and analysis on AI writing:
*In-Depth Video Guide: ChatGPT’s New Onboard Editor, ‘Canvas:’ If you’re looking for a crystal clear, extremely informative demo on ChatGPT’s new onboard editor, Canvas, this video is the ticket.
Produced by the ‘Productive Dude,’ channel, the 12-minute, video guide offers great insights into the editor, whose coolest feature is its ability to highlight and quickly change portions of a text in ChatGPT, on-the-fly.
Canvas is already available to paying users of ChatGPT Plus and ChatGPT Team, is currently being rolled-out to ChatGPT Enterprise and ChatGPT Education users.
It also may be offered to free users at a later date.
Other key tools included by the Canvas editor — which are operated with a click — include the ability to:
~adjust the word length of a document
~adjust the reading level of a document
~solicit editing suggestions for a document
~add final polish to a document’s wording
~add emojis to a document
Plus, while you’re using Canvas, you can also work with the document the ‘old fashioned’ way by using prompts to alter the document’s text.
*Gmail’s Upgraded Auto-Replies: For When “K” Isn’t Enough: Gmail aided by Google’s Gemini AI is now able to offer auto-replies to emails that are much more in-depth.
Observes writer Mike Moore: “After selecting to reply to a message, users will see several response options at the bottom of their screen, which now analyzes the full content of the email thread to provide more detailed, richer responses.
“Users can hover over each response to get a quick preview of the text, then select the one that feels right for the situation.
“You will be able to edit the pre-written message if needed, or send immediately.”
*When Less is More: New Frase Upgrade Cuts Clutter, Keeps Magic: Frase — an AI writer that specializes in auto-producing search engine optimized (SEO) copy — is out with an easier-to-use version.
Observes Matt Hurley, co-founder, Frase: “We removed the functionality that caused clutter and confusion and focused on building more of what truly mattered to our users.
“The result? A simpler yet more powerful tool that ensures you don’t need an army of specialists or endless training to create content that drives results.”
*Microsoft’s Upgraded Copilot: Part Assistant, Part Thinker, Always on: Microsoft Copilot — a key competitor to ChatGPT — now offers enhanced functionality, including:
~An audio-driven, daily news summary
~Natural voice interaction
~The ability to act as a companion when you browse the Web
~’Think Deeper,’ available via the Copilot Lab module, which enables Copilot’s AI to ruminate carefully before responding to a user question
*Google’s Podcast Creator: Instant Banter, Now With Audio and YouTube Inputs: Google’s NotebookLM — an AI research assistant that
auto-creates co-hosted podcasts from text — can now also ingest audio and YouTube videos to auto-create podcasts.
Observes Raiza Martin, a Google product manager: “Today, you can now add public YouTube URLs and audio files directly into your notebook, alongside PDFs, Google Docs, Slides, Web sites and more.”
In practice, this means you can add a bit of text to NotebookLM, a few links to some YouTube videos, a few more links to some audio podcasts — and the tool will auto-create a co-hosted podcast for you based on those inputs.
*Automated Blogging: Who Needs Quality When You Can Have Quantity?: Marketers and others using AI to auto-generate endless posts for their blog could be playing with fire, according to writer Sandra Dawson.
Specifically, Dawson says such automated blogging can lead to:
~Misinformation and low quality content
~Auto keyword stuffing
~Generic-sounding posts
~A slew of other downsides
*Using ChatGPT? Congrats, You’ve Mastered Most AI Writers Already: While there are hundreds of AI writers, just a few companies — including ChatGPT’s maker Open AI, Anthropic and Meta — actually power those auto-writers, according to Ryan Doser.
The reason: Most AI writers are simply software interfaces that sit atop the powerful AI engines that actually do the real work of auto-generating writing, according to this 12-minute video by Doser.
Plus, the few AI titans who own those AI engines currently all use the same technology: Generative AI.
A key takeaway: This is why it makes sense to stay well-acquainted with ChatGPT, whose AI engine — and underlying technology –serves as the foundation for many other AI writers.
Essentially: If you know how to use ChatGPT, you already know — in a general way — how to use all those other AI writers that are powered by ChatGPT or powered by other generative AI.
*New ChatGPT Challenger: Free, Open — and Ready to Rumble: ChatGPT has another challenger lurching for its throne: A new AI engine just released by Nvidia.
Interestingly, the new AI engine is open source, meaning anyone can download its software, tinker with it and/or build applications atop it, free-of-charge.
The reason why this particular AI engine is so notable: Most of today’s generative AI is powered by Nvidia chips, which heavily dominate the world as the go-to hardware for powering AI.
Plus, Nvidia also has extremely deep pockets to continue competing with ChatGPT: It’s currently one of the top five companies in the world and worth about $3.4 trillion.
*AI Big Picture: AI Engine Building: For People Who Use Moons as Paperweights: The power to build AI engines — the underlying software that powers today’s AI writers and similar apps — is being concentrated in fewer and fewer hands.
The reason: It takes enormous amounts of capital to build such engines — also known as Large Language Models.
Case in point: Character.AI, an AI startup, just abandoned its efforts to enhance its own AI engine, given that such building has become incredibly expensive, according to writer Sage Lazzaro.
Share a Link: Please consider sharing a link to https://RobotWritersAI.com from your blog, social media post, publication or emails. More links leading to RobotWritersAI.com helps everyone interested in AI-generated writing.
–Joe Dysart is editor of RobotWritersAI.com and a tech journalist with 20+ years experience. His work has appeared in 150+ publications, including The New York Times and the Financial Times of London.
The post Dream a Little Dream for Me appeared first on Robot Writers AI.