AI’s Teenage Years: Popular with Friends, Awkward at Work
Plus, the Pepsi Navy saga and why your favorite cake might taste different.
Hey there, friend!
I know, I know, the whole AI thing is getting a bit old, right? It’s like we’re drowning in news, updates, and developments. It’s like we need AI just to keep up with AI! It’s like we’ve entered a whole new world where tech giants are pouring hundreds of billions of dollars into data centers, GPUs, TPUs, training, and developing better (and new) algorithms. With so much at stake (you don’t invest that kind of money if you’re the CEO of Microsoft, Google, OpenAI, or any other tech company in the AI race), expect the barrage to keep coming for a while.
With that, here’s your Weekend Briefing.
Have a beautiful day!
P ツ
The Thin Wisps of Tomorrow
AI’s Enterprise Problem — Here is something which should give you pause: “Some 75% of OpenAI’s revenue comes from consumers paying for its AI-powered chatbot and other products, a company spokesperson said on Monday. That figure highlights the uphill battle OpenAI and its rivals face in getting businesses to pay for their AI software.”
Everybody is using AI for their personal stuff (and surely to help with their jobs), but seemingly few companies have figured out how to use it productively (otherwise they would spend much more money on it). Which stands in stark contrast to claims from IDC / Microsoft who claimed that in a study IDC did (sponsored by Microsoft, a company which clearly has something to gain from AI’s rise), that companies see $3.5 of return for every dollar invested. Let me call BS on this: If the IDC study were a faithful representation of reality, OpenAI wouldn’t be able to keep up with demand for their enterprise/API side of the business.
The (AI) Agents Are Coming — Last week saw the announcement of Anthropic’s Computer Use, as well as an early announcement of Google’s Jarvis (seriously, who comes up with names like this?), both AI-powered tools that essentially allow a LLM to navigate your computer by clicking on things the AI sees on your screen. It’s an (early) step into the exciting world of agentic AI, where our machine overlords can (and will) actually do things for us and not just merely provide insights that humans then have to turn into mouse clicks and keyboard typing. Alas, the current generation of these systems is too slow and too expensive to be truly useful. So, let’s wait.
The Gestalt Returns — Here is an interesting development highlighting the effect of what we like to call the gestalt: the effect of surrounding factors that typically aren’t directly related to the progression of the development of a weak signal but can affect it positively or negatively (we write about this in our book “Disrupt Disruption”): Indonesia made the sale and possession of an iPhone 16 illegal in the country over disputes between the country and Apple regarding investment subsidies.
What We Are Reading
🎨 This AI Painter Has Sold $4 Million In Artwork. Now Sotheby’s Wants A Piece Of The Action An AI artist called Botto has already generated $4 million in art sales and is now partnering with Sotheby’s to bring its decentralized AI-created artwork to a mainstream art market audience. @Jane
🍰 Skimpflation Is Hard To Spot, But The Berry Chantilly Cake From Whole Foods Can Help Explain You’ve likely heard of shrinkflation—selling smaller versions of products at the same price; now there’s skimpflation: reducing quality or changing ingredients to cut costs. Think replacing cocoa butter with palm oil. If and when customers notice, it could result in some risky business for the companies selling. @Mafe
🤖 AI And The Myth Of Personalised Learning One of my favorite thinkers at the intersection of emerging tech and education argues that the dream of AI-enabled “personalized learning” is a profitable myth for tech companies far more than it is a bright future for learners. @Jeffrey
🌍 The U.N.’s Verdict On Climate Progress Over The Past Year: There Was None The UN report concludes that only a fraction of the nations that had committed to lowering their emissions as early as Paris have come through with adequate policies. Nevertheless, each policy that slows fractions of a degree has crucial benefits. @Julian
👃 How Chain Restaurants Use Smells To Entice Us Some smells can be as distinctive as an impeccably designed logo. Cinnabon, Subway, and Auntie Anne’s continuously capitalize on their very identifiable aromas to lure customers. @Pedro
🚢 The Doomed Voyage Of Pepsi’s Soviet Navy The truly crazy story of how Pepsi, the maker of your favorite beverage and snacks, ended up owning one of the largest military naval fleets in the world. @Pascal
Headlines from the Future
🚗 Maybe it’s not a good sign: The CEO of Ford says he’s been driving a Xiaomi EV for the past 6 months and doesn’t want to give it up.
👨🏼🎨 Answer: Only 200-600 samples: How many Van Goghs does it take to Van Gogh? Finding the imitation threshold.
🏥 Hallucinations in LLMs continue to be a real-world problem: Researchers say an AI-powered transcription tool used in hospitals invents things no one ever said.
💬 Truly astounding progress in voice to text AI technology: Introducing Moonshine, the new state of the art for speech to text.
🤜🏼 Maybe that’s what we need AI for: Google’s DeepMind is building an AI to keep us from hating each other.
🔋 Electric Vehicles were the linchpin; now the revolution is in full swing: The Battery Revolution Is Finally Here.
🗺️ Still confused how to properly work with your AI of choice? Here is a handy Prompt Engineering Roadmap.
🛋️ Urban planners better pay attention: The COVID-19 pandemic has turned us into homebodies.
🔊 The magic behind Google’s NotebookLM podcast-generating AI: Pushing the frontiers of audio generation.
👨🏼💻 So, that is what AI is good for: Google CEO says more than a quarter of the company’s new code is created by AI.
💰 In case you’re curious about the magnitude of Meta’s AI endeavors, here’s Ben Thompson’s perspective: Meta’s AI Abundance.
The Fun Stuff
🧑🏼🚀 Ever wondered how to put on an Apollo spacesuit? Here is how.
🎃 Supersize Me: 2,000 pound pumpkins are a thing now. There is (a lot of) science behind it.
🧑🏼💻 Mac-only, but I’ve been using this meeting summarizer for a little while, and it it’s truly useful (although a little buggy from time to time).