|
TGIF!!!! It’s time for the weekly roundup.
In the podcast this week, we talk about how stupidly easy it was to make a ChatGPT-powered news site that ripped off our work and that of WIRED, The Verge, and others, and we listen to some of the AI-generated music that has major record labels suing some startups. In the subscribers-only section, we discuss how a bunch of chatbots suddenly changed personality overnight. Sign up to get access to that portion of the show.
Listen to the weekly podcast on Apple Podcasts or Spotify, or watch it on YouTube
Okay, let’s go over what happened this week.
 Photo by 404 Media
ABANDON ALL HOPE
Meta has fucked around, and now is finding out. Jason wrote about how Facebook and Instagram have been overrun by AI-generated spam and scams and “people” engaging with this content are bots who themselves spam the platform. Experts in moderation, who Meta once consulted for help, say they haven’t heard a peep from the company in a long time — and it shows.
 Screenshot via 404 Media
WE’RE RETIRING NOW
How hard could it be to set up one of those click-mill ChatGPT-generated news sites that steals from other journalism outlets? “It wasn’t hard to set up, and didn’t require one of the most advanced large language models in the world, but since this is the second technology news and investigations website I’m running these days, I outsourced its creation to a Fiverr freelancer in Turkey,” Emanuel wrote of his experiment. “I told him what I wanted, picked a layout, and two days later got a fully operational website.”
 Screenshots via Youtube / Amazon, collage by 404 Media
OKAY DEEP BRONZE QUEEN
A slice of Rudy Giuliani’s Amazon purchase history is now public thanks to his bankruptcy proceedings in court, and it turns out he has been racking up some questionable and sad purchases, including self-tanner, cheap ties, and knock off Airpods. I checked out what he’s been buying and only have one thing to say about it: Log off Rudy.
 Unsplash
HUGE SECURITY FAIL
A company that verifies the identities of TikTok, Uber, and X users—including using photos of their faces drivers’ licenses—exposed a set of administrative credentials online for more than a year, potentially allowing hackers to access that data, according to screenshots and data obtained by 404 Media. The company was “entrusted with people's identities and it failed to implement simple measures to protect people's identities and sensitive ID documents,” the cybersecurity expert who alerted us to the news told Joseph.
 Screenshot via eBay
WHAT A DEAL
eBay took down a listing for an apparently original StingRay cell phone surveillance device after 404 Media contacted the platform, as Jules reported. The device, which was labeled as used, was up for sale for $100,000. “Powers on. Not fully tested,” the seller wrote in the listing details. “Only pictured parts are included in listing.”
READ MORE
After reading “Perplexity’s Origin Story: Scraping Twitter With Fake Academic Accounts” L.K. wrote:
“The more I learn about this company, the more I think it’s just a 3 article spinners in a trench coat.”
Nothing to see here!
In response to “eBay Removes Listing for StingRay Cellphone Spying Tech” Sammy Wylie writes:
“With a bit of investigation I think you will find that, more than likely the unit was purchased at a local Government Surplus Auction by the seller. Governmental agencies of all shades (local, State and Federal) tend to be lax in screening equipment deemed surplus or obsolete; throwing it into the property disposal pipeline as a matter of course and convenience, where it can be purchased by the general public. If there are any aspersions to be cast in this scenario, the blame most likely lies upstream of the seller, who is probably just trying to make a living selling surplus items. As they say, ‘The fish stinks from the head’...”
And replying to “Has Facebook Stopped Trying?” Alex H. comments:
“This is why I've been frankly extremely confused by Meta's enormous investments into AI. I simply don't understand what their business plan is anymore. Last I checked they still make the vast majority of their revenue on advertising. At all times their overriding concern should be ‘how do we get advertisers to spend more money on us?’ That should either take the form of service improvements for the advertisers themselves directly or general platform improvements that increase the important user activity metrics that fuel a successful advertising business. But they're letting their user experience become absolutely horrible, and dumping billions and billions of dollars into a new technology that has a very unclear path towards making it any *better*. Everybody I know who uses Facebook advertising has noticed a decline in performance over the last 3-4 years, but people keep spending money anyway because, along with Google, they've basically captured the market. I guess they're just arrogant and assume that users will continue hanging around the sewer and advertisers will keep desperately throwing dollars at diminishing returns?”
I am in the sewer, splashing around.
BEHIND THE BLOG
This is Behind the Blog, where we share our behind-the-scenes thoughts about how a few of our top stories of the week came together. This week, we discuss exciting new updates to the podcast, having too many tabs open, and the generative AI boom.
JOSEPH: I’m gonna give you two Behind the Blogs for the price of one. First off, about the podcast: as you may have read or seen this week, we’ve made some major changes to the podcast. Nothing to do with format, I think we’re all very happy with that and I think listeners are too (I hate it personally when a podcast shakes up its format too much, throws me off). But we’re putting a bigger emphasis on video and improving the overall quality of the audio.
Up until this point, we’ve been using Riverside to record. This is a browser-based tool we all log into, it records our audio and video locally, and uploads it. That sounds great, but Riverside has a lot of issues. Sometimes the tracks don’t sync. The audio level normalization can introduce weird artifacts that actually make the voices sound worse. The company is introducing a bunch of AI features that look like they’ll save time, while not improving the tool’s general reliability. We have been using Riverside to make it easier for us to go from recording to finished product, but the output we’re automatically getting from them still requires a fair bit of work. On top of that I’m not a trained audio engineer, and sometimes mistakes are made in the podcast mostly during recording itself. Read the rest of Joseph's Behind the Blog, as well as Jason, Emanuel and Sam's, by becoming a paid subscriber.
Okay, I need to do some online shopping. See you next week.
|
