GENAI MEDIA

AI bunnies on trampolines are causing a "crisis of confidence" on TikTok

TikTok users are having existential breakdowns over AI-generated videos of adorable bunnies bouncing on trampolines. The videos look real enough that viewers are questioning their ability to distinguish reality from artificial content, with comments like "I can't tell anymore" and "this is terrifying."

What we're seeing goes far beyond bunny videos. AI-generated "slop" is flooding social platforms, with Facebook's top posts now regularly featuring obviously AI-generated content—from bizarre religious imagery to impossible scenarios that nonetheless rack up thousands of likes from unsuspecting users.

In this newsletter, we've run an "AI or Not" section for over two years as a way to showcase how impressive these models are getting. What started as a fun experiment has become genuinely challenging—we've seen voting accuracy trend significantly toward an even 50/50 split on our daily polls. People simply can't tell anymore.

The warning signs have been building for years:

• The viral Pope Francis in a Balenciaga puffer jacket from 2023 fooled millions of users, including celebrities like Chrissy Teigen

• Fake Trump arrest photos sparked online threats and widespread confusion

• Deepfake videos of Ukraine's Zelensky appeared to show him telling troops to surrender during wartime

AI-generated content played a significant role in 2024 election disinformation, with Trump posting AI images falsely showing Taylor Swift endorsing him. Research shows people can't reliably detect deepfakes even when warned to look for them.

Meanwhile, social media platforms are financially incentivizing this chaos. Creators from developing countries are monetizing bizarre AI videos that exploit platform algorithms, earning money from TikTok and Facebook's creator funds for generating "AI slop"—low-quality AI content designed purely for engagement.

We're witnessing the collapse of visual truth online, and it's accelerating faster than anyone anticipated. When people can't distinguish AI bunnies from real ones, we've crossed a threshold. The "crisis of confidence" on TikTok is a preview of how completely unprepared we are for sophisticated political deepfakes and coordinated disinformation campaigns. Platforms are choosing short-term engagement profits over long-term societal trust, creating a system where seeing is no longer believing.

TOGETHER WITH MOMENTUM

Try This AI Prompt Cheat Sheet

A solid prompt can be the difference between confused agents and seamless workflow – but what constitutes a good one when it comes to AI? Great question.

These best-in-class examples from Momentum’s AI Prompt Library are the perfect start. 

With more than 200+ go-to-market-ready prompts specific to sales, customer success, marketing, or beyond, this library is your AI prompt cheat sheet. Use it to improve your existing prompts, uncover powerful new ones to immediately uplevel your agent output, or simply get a feel for what makes a great prompt successful. 

It’s all free, and it’s all courtesy of Momentum — plus, you’ll get access to a free Live Prompt Training.

Access their AI Prompt Library right here.

FRONTIER AI

OpenAI's research chiefs drop major hints about GPT-5

Mark Chen and Jakub Pachocki were practically giddy when MIT Technology Review caught up with them this week. OpenAI's twin research heads had just scored back-to-back victories: second place at the AtCoder World Tour Finals on July 16, followed by gold-medal results at the International Math Olympiad three days later.

Chen seemed most excited about the programming competition, where their model lost to Przemysław Dębiak, a legendary coder known as "Psyho." Dębiak used to work at OpenAI and personally knows Pachocki from their competitive coding days. "Psyho resists for now," Pachocki admitted while watching the late-night livestream from Tokyo.

The interview offered rare insights into GPT-5 development. While reports suggest an August launch, Pachocki revealed that they’re "probably still at the very beginning of this reasoning paradigm. You have these models which know a lot of things but can't chain that knowledge together."

When pressed about safety team departures and Jan Leike's criticism that "safety culture has taken a backseat to shiny products," their responses were revealing:

• Chen dismissed it as researchers whose work became "less consistent with the way the field is evolving"

• Pachocki was blunter: "Two years ago the risks were mostly theoretical. The world today looks very different"

The superalignment team that once consumed 20% of company resources? Disbanded. Safety research focused on hypothetical superintelligence? Replaced by deployment challenges.

Most tellingly, when asked whether focusing on mathematical intelligence was problematic, Chen responded: "Why not?" to the suggestion that maybe we don't want AI replacing politicians. That response reveals OpenAI's current mindset—they're building for maximum capability, not careful integration.

TOGETHER WITH MISO

“One of the Largest Industries Ever”

That’s what NVIDIA’s CEO said about robotics – and the numbers back it up. In Q4 2024, the sector raised nearly more capital than in all of 2023. 

So it’s a big deal that both NVIDIA and Amazon helped Miso Robotics perfect Flippy, its new fully commercial kitchen robot; no wonder initial units sold out in one week.

Miso already has 200K+ hours for brands like White Castle under its belt. With the $1T fast-food industry facing 144% turnover rates, Flippy provides 24/7 labor and can boost profits by up to 3X.

Now, Miso is scaling its US-based manufacturing to seize a $4B/year revenue opportunity – and you can get up to 8% bonus stock as an investor. But don’t wait too long – Miso’s share price changes on August 14.

RESEARCH

Google's AlphaEarth turns Earth into a real-time digital twin

Google DeepMind just unveiled AlphaEarth Foundations, an AI system designed to tackle one of satellite imagery's biggest challenges: processing the vast amount of data streaming down from space daily.

The system functions as what Google calls a "virtual satellite," stitching together optical images, radar data, 3D laser mapping and climate simulations into a unified digital representation of Earth's surface. AlphaEarth compresses each 10-meter square of Earth into a 64-dimensional "embedding" that captures a full year of observations, allowing it to track changes over time and fill gaps where cloud coverage blocks traditional satellites.

Google's pitch centers on efficiency gains:

• The system requires 16 times less storage than competing AI models while achieving a 24% lower error rate in testing.

• The company has already generated embeddings for over 1.4 trillion footprints annually, working with partners including the UN Food and Agriculture Organization and Stanford University.

This builds on Google Earth Engine, the company's 15-year-old platform that already processes petabytes of satellite data for researchers worldwide. But AlphaEarth represents a strategic shift from hardware-dependent to AI-dependent Earth observation, potentially reducing reliance on expensive satellite launches.

As climate reporting requirements tighten globally, governments and corporations need reliable environmental monitoring. Google is positioning itself as the infrastructure provider for this data-hungry future, competing against not just traditional players like Planet Labs and Maxar, but also Microsoft's partnership with SpaceX and Amazon's Project Kuiper satellite constellation.

The success of AlphaEarth ultimately hinges on whether it delivers actionable insights that justify its computational footprint or becomes another impressive technical demo struggling to find real-world traction.

LINKS

• Apple reports biggest revenue growth since December 2021

• AI researchers are negotiating $250m pay packages. Just like NBA stars

• OpenAI launches Stargate in Europe with Norwegian deal

• Uber Eats is adding AI to menus, food photos and reviews

• Apple, Google, OpenAI to work with federal agencies to improve health data

• Army plans shakeup in software buying, starting with new $10b Palantir deal

• Why DOGE’s Luke Farritor followed Elon Musk to DC

• On AGI, mass automation, and what the Luddites really fought against

• Meta is on the hunt for AI video deals

• Meta brought AI to rural Colombia. Now students are failing exams

• Droidrun: Grant AI native control of virtual & physical phones

• Warp: AI-native HR & payroll platform that runs itself

• Cline: Open source AI coding agent right in your IDE

• Pally: AI relationship management for all your pals, from all your socials

• Machine Learning Engineer – 7 yrs, ex-Google, LLM deployment expertise

• AI Product Manager – SaaS growth, launched successful AI features

• Generative AI Engineer – Multimodal models, RLHF, prompt engineering

(Sponsored)

GAMES

POLL RESULTS

What's your take on the Musk v Altman lawsuit?

Musk is genuinely defending OpenAI’s original nonprofit mission (22%)

• “Although I believe there is a Billionaires game shop show at hand. I do believe that OpenAI was originally a free-for-all project.”

Musk’s suit is mainly a tactical move to slow a rival (32%)

• “Elon is all about doing massive Takeovers. He's done this all his adult life, a perfect example is Twitter. He thought he would be the king of social media until Threads, BlueSky and others that thwated him and his pirate mindset.”

Both Musk and Altman are equally at fault; this is billionaire gamesmanship (37%)

• “There is a mix for both of these guys. They do want to change the world for the better, and think they are the best ones to do it. And they want to make a killing along the way, and think it is deserved. There is a degree of boys on a play ground competition involved as well. In all of that, these two men are similar. And I probably wouldn't be against either of them.”

Something else? (9%)