Papież napisał o AI lepiej niż niejeden CEO Doliny Krzemowej. | AI WEEKLY #41

Hi! This is Krzysztof Wojewodzic from Escola AI 👋 New episode of AI Weekly #41 is now live. This time, 7 news stories that I had a particularly difficult time choosing. From the papal encyclical on AI, to the Polish language model with 70 billion parameters, to the three largest IPOs in history, all of which are scheduled to land in the same year. In this episode: The 7th Pope issued an encyclical on AI. And he wrote things in it that are hard to argue with. Magnifica Humanitas is 200 pages of reflections on what AI is and what it should be. AI for people, not people for AI. Technology is not neutral. AI as a weapon, new forms of slavery, the risk of concentration of power in the hands of a few. Krzysztof has read it and is frank: he finds few passages there to disagree with. 6️⃣ Human Capital Balance Report: Polish companies struggle with AI competencies. Only 25% of companies say they have access to qualified AI experts. Nearly half of companies are inactive in the face of this problem. And one in six companies that tried to recruit encountered excessively high financial requirements for candidates. The conclusions are clear, only few want to hear them. 5️⃣ Musk lost to OpenAI. The court dismissed the lawsuit. An Oakland jury dismissed the case as time-barred. For OpenAI, this cleared the way for an IPO. For Musk, it was a defeat, but not the end of the story. Krzysztof explains why the outcome of this dispute was less important than the speed at which the court resolved it. 4️⃣ Uber burned through its annual token budget in four months. The phenomenon of token maxing—using AI for everything without considering the costs—has reached one of the world's largest companies. The result: companies are starting to look for alternatives and are increasingly turning to Chinese open-source models. 60% of OpenRouter's token consumption is already Chinese models. 3️⃣ Claude Opus 4.8: Greater token control and better context management. Anthropic solves two of its users' biggest problems. Effort control allows you to control how many tokens the model uses for a given task. Better context compression means the model is less likely to lose track of long, multi-step tasks. And Anthropic confirms that it will be profitable this quarter. 2️⃣ The Polish Plum model has 70 billion parameters and is fully open source. Plum is a government-backed model, compliant with copyright law, and available for commercial use. It can be downloaded through Hugging Face and run locally. Krzysztof sees it as the ideal engine for chatbots in public administration and government agencies that don't need to use American models. 1️⃣ The three largest IPOs in history. In one year. What will this do to the market? SpaceX, OpenAI, and Anthropic. Each with valuations exceeding $1 trillion. For comparison, Orlen is valued at $45 billion. Krzysztof explains why these companies' inclusion in indices like the S&P 500 could paradoxically lower the prices of other tech giants. If you want to know where AI is truly changing the economy, law, and the way we think about the future, this episode provides a very concrete picture. 🎥 Watch on YouTube 🎧 Listen on Spotify ✍️ Read on Substack Let us know which topic from this episode resonates with you most. #AI #ArtificialIntelligence #AIWeekly #EscolaAI #Business #Technology 0:00 Introduction 0:48 Pope Issues Encyclical on AI 6:33 Report on AI Competencies in Polish Companies 9:57 Musk Loses to OpenAI in Court 12:05 Uber Burns Through Annual Token Budget in 4 Months 17:43 Claude Opus 4.8 and Control of Tokens 22:48 Polish Plum Model with 70 Billion Parameters 26:48 The Largest IPO in History and What They Will Do to the Market 31:40 Summary