A (Beta) Audio Roundup of September’s WordPress Vulnerabilities
For those of you that want to stay abreast of the newest vulnerabilities in the WP ecosystem, but like to multitask, here’s an audio roundup of the vulnerabilities we published in the month of September. This is something new I’m trying. The conversation is AI generated by Google’s NotebookLM podcast generation feature which I find quite impressive. It’s designed to be accessible to the layperson.
Let me know what you think in the comments. I’ll be happy to share our own evaluation results but don’t want to bias you going into this. Here’s a downloadable link if you right click on this link and save-as. Enjoy.
Comments
1:44 pm
Holy cow, that conversation sounds so natural, it's hard to believe that it's AI-generated. Some filler words & phrases thrown in, dynamic pacing & tones. Very impressive.
1:59 pm
Agreed.
1:44 pm
Have to confess that was pretty good and I've heard worse from real people. I'd look at the interjections with the "yeah" parts as a few points there were too many but overall it was really good!
2:00 pm
For v1 I agree.
1:45 pm
What exactly is AI generated? The script? The voices? Both?
1:59 pm
Both.
2:59 pm
Holy F*xK I'm a geek and this scares the absolute fricking hell out of me. I just started listening, and then I read comments. The script I wasn't surprised by as I'm an author, and I've researched ChatGPT et.al. BUT THE VOICES!!!!! Are you saying you didn't have to do any manipulation for the emotions reflected in the creation of the podcast? The AI did it all? In my head I'm having a vicious argument of WOAH! Do you realize how great and cheap your AI audiobooks would sound? battling the Are you ready to have politicians, world leaders and the unscrupulous lead you into the valley of the Handmaiden's Tale and Dante's Inferno all at the same time. I'm amazed, horrified, enthralled, and cringing all at the same time.
1:52 pm
You are kidding, right? This is awful. It's terrible to listen to, and it takes longer and is less informative than just reading a list of vulnerabilities. Even laypeople deserve real information.
2:01 pm
Kidding? No. For v1 of their product it’s incredibly impressive.
2:09 pm
As an experiment, it is very interesting. As a delivery mechanism for content, I side with Clare.
1:56 pm
Firstly did a human write the dialouge? Because I noticed how one voice is uninformed then becomes the teacher, and vica-versa. The two voices do this all through the file. But, the voices are engaging. Now, if I could get AI to use my voice, I could convert 10 years of blogs into audio files that would bring my material to thousands. How do I do that??
1:58 pm
No this is AI written based on our blog posts. Yes the role reversal we’re seeing is a bug.
1:57 pm
Okay I heard a Notebook AI podcast before, I just wanted to quickly check this out and I ended up continuing to listen. I loved it. But is the information true and accurate? I have no idea. I'd love to know if you think it was all well explained. It's maybe a bit too casual and wordy, but it's so good to listen to.
2:07 pm
Yes we checked and it’s good enough. For some of the explanations - we may have clarified things differently, emphasized differently, but overall it’s quite I impressive. No glaring issues.
2:04 pm
How does the info in the podcast compare to the source articles? Mainly, is it accurate or are their factual discrepancies?
2:06 pm
Favorably. This is what we paid most attention to during evaluation. Not perfect - we didn’t love some of the dumbed down explanations - but good enough.
2:07 pm
Not a fan. Got through a few minutes but couldn't continue. I listen to a decent amount of podcasts. Knowing this is AI generated makes it cringy (for me) as the banter and repetitive interactions, podcast tactics to length episodes, means that this has been engineered to take more time than necessary to delivery information.
2:08 pm
I have a feeling knowing it’s AI may bias us. I found after evaluating it with a few of our team for 15 mins I was far more critical on second listen.
2:15 pm
I completely agree that knowing it is AI does add bias. Contextually, when I think of Wordfence, I think "excellent service, thought and security leaders, focused messaging," etc. IMHO (everyone's got one!), this does not have that same focus and clarity.
2:19 pm
Interesting. Thanks.
2:10 pm
As someone who has used WordFence on the website since 2012 for my NON-AI talk radio program that I then distribute AS a podcast, I am highly motivated to move fully to Sucuri or another solution with this move from WordPress. Why should ANY podcaster feel anything but betrayed by you for going this route?
2:17 pm
I wouldn’t worry too much. There’s a human connection that these generative products lack. Not sure what “route” you think I’m going btw. This is just a fun thing I thought I’d share. It’s googles product and was ranked #1 on hacker news earlier today so I’m afraid the cat is out of the bag.
2:13 pm
Woah! This is incredible!
2:20 pm
Yeah when I play this to friends and family I get that reaction.
2:16 pm
I'd have liked to listen to that first not knowing it's AI. I was definitely impressed by the voice quality and conversational tone. The excessive grunts and interjections might not have stood out if I wasn't listening for flaws.
2:19 pm
Yeah sorry Gabe I almost structured the post that way but decided against it because I figured some people would get mad at me. I’m already getting shouted at a bit in these comments so no regrets there. But agreed not knowing first listen is more fun.
2:34 pm
Interesting. So AI read through the vulnerabilities, you suggested two character types to discuss the vulnerabilities, and the result is the conversation between what sound like two humans, but the voices and conversation are all AI generated. Impressive, but freaky at the same time.
2:36 pm
It reads the docs and generates its own audio summary - not based on prompts I created but based on Googles own defaults. The podcast format with the two presenters is their default.
3:16 pm
This is impressive but the implications of it are dire for human work, creativity and dignity. The potential for exploitation is huge and skill devaluation is enormous.
Whose voices are these based on?
Did they give their permission? If they are voices generates by aggregate, was consent obtained by those whose voices were sampled?
Listening for longer than a minute, it becomes less and less human. A pattern emerges. The cadences are all the same.
I can’t stress enough how much this kind of project devalues and reduces the human element when it comes to *valuable* contributions to society.
This content needs, by law, to have a kind of audio watermark or disclaimer that it is auto generated. My hope is that people will value human efforts and ignore this reductive, commodified AI content.
3:17 pm
I've been using it myself: the guy says100% and the girl says exactly LIKE way too much Like Chuckle hahaha LOL
3:18 pm
Overall, given that this is an AI generated podcast, I find it fascinating for several reasons. The natural-sounding conversation, voice quality and modulation, and pacing all seem very natural. Some comments have been negative, but if you consider how this demonstrates how far AI has gone in a short period of time, this is eye-opening. If it gets to the point where there are zero factual errors, who knows how far this can be extended.
4:27 pm
Doesn't mean you should, just because you can.