Axios AI+

August 04, 2025
Last night was a rough one for the Valkyries, but they have still had an impressive road trip. Today's AI+ is 1,194 words, a 4.5-minute read.
1 big thing: AI slop's inescapable ooze
An AI-generated video of rabbits jumping on a trampoline that went viral last week — and was widely believed to be real — proved even cute animal vids aren't safe from convincing slop machines.
Why it matters: All the fake AI-generated content online is sapping the joy of casual scrolling.
- It's not just animal vids. Hobbyists say genAI is ruining gardening forums, knitting communities and the entire DIY aesthetic on Pinterest. AI-generated images lack the imperfections that make these community connections relatable.
- After Vogue ran an ad featuring an AI model in its August print issue, readers called the ads hollow and accused them of stealing jobs from models and photographers.
- Fake influencers have been slipping into our feeds for a few years, but they're getting harder to spot now that image generators have mastered fingers — though they still struggle with knees.
Driving the news: The latest viral video in question features outdoor night camera footage of the bounding bunnies, which was quickly debunked. The tell-tale signs: lack of continuity, defying physics, and a glitchy video timer.
- Young people online expressing their fear that falling for the bunnies made them feel old quickly became a trend.
Fun fact: A real coyote trampoline video sparked this craze, according to YouTuber Jeremy Carrasco.
Between the lines: AI-generated oddities are showing up in our feeds more often because they're lucrative for creators.
- The weirder it is, the more time we spend looking at it, signaling the algorithm that we want to see more like it.
- If we watch content over and over again or stop scrolling to see if an image or video shows the signs of being AI-generated, that's more fuel for the algorithm.
- Engagement — even confusion-driven engagement — is valuable.
Catch up quick: Images and videos generated with OpenAI, Google and Meta's sophisticated free or cheap AI tools are swamping the internet and fooling even the savviest scrollers.
- This particular brand of synthetic content is called AI slop because it oozes into and potentially suffocates human-made media. Others argue it's more nuanced.
Slop isn't new either, Ben Kusin, founder of AI studio Kartel, told Axios.
- He points to shows like "America's Funniest Home Videos" as evidence that what some call "slop" has existed long before generative AI.
- "AI has democratized the ability for people to make that at scale," Kusin said.
Zoom out: More users appear to be adapting to and growing comfortable with AI-generated content overall.
- Case in point: In 2023, the musician Nick Cave called AI "a grotesque mockery of what it is to be human." This week, Cave released a video featuring an AI Elvis.
What to watch: Platforms themselves must stop incentivizing the slop, Kusin explained.
- If you open up Instagram and all you see is AI slop, you might stop opening up Instagram. This, Kusin said, is "an existential threat to the platform."
2. Blue-collar revenge against the machines
AI is supposed to displace millions of workers in the coming years — but when your toilet won't flush at 2am, you're not going to call ChatGPT.
Why it matters: The reshaping of the American economy promises to offer a kind of revenge for the blue-collar laborer, as white-collar workers become largely dispensable, but the need for skilled trades only grows.
The big picture: Companies are already boasting of saving hundreds of millions of dollars a year by using AI instead of humans. The stock market rewards are too enticing for the C-suite to ignore.
- But ask those same executives who's going to run the wiring for their data centers, or who's putting the roof on the building, and just how well those skilled technicians are getting paid.
- It's become a key Trump administration economic talking point: Blue-collar wages are rising faster now than at the start of any other administration going back to Nixon.
Driving the news: A recent Microsoft paper analyzing the most "AI-proof" jobs generated a list of the work most and least vulnerable to the rise of the LLM.
- The 40 most-vulnerable jobs (translators, historians, sales reps, etc.), basically all of them office work, employ about 11 million people.
- The 40 least-vulnerable jobs (dredge operators, roofers, etc.), just about all of them manual labor, employ around 5.5 million.
- All those extra folks have to go somewhere.
What they're saying: "We've been telling kids for 15 years to code. 'Learn to code!' we said. Yeah, well, AI's coming for the coders. They're not coming for the welders. They're not coming for the plumbers. They're not coming for the steamfitters or the pipe fitters or the HVACs. They're not coming for the electricians," Mike Rowe, the TV host and skilled-trades philanthropist, said at Sen. Dave McCormick's (R-Pa.) AI summit last month.
- "There is a clear and present freak-out going on right now," Rowe said, as everyone from politicians to CEOs recognizes just how badly they need tradespeople to keep the economy running.
Yes, but: While the AI boom will create lots of jobs for skilled trades, eventually there'll be less demand to build more data centers, which may in turn sap demand for those tradespeople too.
The intrigue: There's already a labor shortage in many of these blue-collar professions, one that AI will, ironically, only make worse (think the electricians for the data centers, for example). Factories alone are short about 450,000 people a month, per the National Association of Manufacturers (NAM).
- "We're really talking about high-tech, 21st century, rewarding, well-paying jobs," Jay Timmons, the CEO of the NAM, tells Axios. "Manufacturers are really embracing what's coming, and they accept the responsibility."
Training is the answer, but that will require a large-scale, national effort — not just for up-and-coming students, but for mid-career workers forced into a pivot.
- "Everybody needs these roles, they're high-security roles," says Carolyn Lee, president of the NAM-affiliated Manufacturing Institute.
- She points, for example, to a program already in 16 states to train maintenance technicians to keep factories running — precisely the kind of job people like Commerce Secretary Howard Lutnick have said are the future of the workforce.
- Students in an early cohort of that program, on average, were earning $95,000 a year within five years of graduating.
- One of the challenges, Timmons notes, is selling that to people who may not understand how lucrative these careers can be: "You have an economy-wide perception problem."
The bottom line: Next time you have to call the plumber, you might want to ask if they're hiring.
3. Training data
- Spending on AI infrastructure is masking weakness in the rest of the economy. (Axios)
- Apple is in the early stages of developing an AI "Answers" tool designed to work like ChatGPT. (Bloomberg)
- People who need help writing an obit for their loved ones are turning to AI. (Washington Post)
4. + This
Over the weekend I got to see Melissa Etheridge and the Indigo Girls together in concert. My 21-year-old self would have been very jealous. My 50-year-old self was very happy to share the moment with my partner, kid and mother-in-law.
Thanks to Scott Rosenberg and Megan Morrone for editing this newsletter and Matt Piper for copy editing.
Sign up for Axios AI+






