Good morning from San Francisco, 🌉☕️🌤

Where today's AI news feels like a relationship update: complicated. 🤖❤️🤷‍♂️

OpenAI just dropped GPT-4.5—an emotionally intelligent giant that's empathetic enough to ask how your day's going, yet struggles to match its predecessors' brainpower. Think therapist, not genius. 🛋💬🧠

Meanwhile, Mark Zuckerberg wants a slice of ChatGPT's pie. Meta will launch its own standalone AI assistant app, promising 1 billion users by year's end. Zuck's employees now hustle seven days a week, presumably fueled by ambition—and maybe panic. 📱🔥😅

AI companies are doubling down, shifting strategies, and ratcheting up the stakes. Silicon Valley’s latest pivot: trading horsepower for heart, and likes for logic. 🚀💡💔

Stay curious,

Marcus Schuler 🧐🔍✨

OpenAI's Latest Model: Bigger, Warmer, But Not Necessarily Better

OpenAI's newest AI model takes an unexpected turn. GPT-4.5, released today as a research preview, excels at emotional intelligence but comes with notable limitations and a hefty price tag.

Early testing reveals GPT-4.5's standout feature: it's surprisingly good at human interaction. When faced with a user's test failure, it responds with empathy rather than rattling off advice. "Want to talk about what happened, or do you just need a distraction? I'm here either way," it offers.

Mixed Performance Reviews
Despite being OpenAI's largest model yet, GPT-4.5 falls short of benchmark expectations. It doesn't match the performance of the company's reasoning models (o1, o3-mini) or deep research. However, it does show improvement in reducing hallucinations, with a rate of 37.1% compared to GPT-4o's higher rate.

The Last of Its Kind
CEO Sam Altman confirms GPT-4.5 will be "our last non-chain-of-thought model." This signals a strategic shift away from traditional language models toward AI systems that think through problems step by step.

GPT-5, expected as soon as late May, will integrate OpenAI's o3 reasoning model. The company sees GPT-4.5 as a stepping stone toward more sophisticated AI that combines language understanding with logical reasoning.

Starting today, ChatGPT Pro subscribers ($200/month) get first access. Plus and Team users follow next week, with Enterprise and Edu users after that. The model supports file uploads and web search but lacks voice mode and screen sharing capabilities.

Why this matters:

• OpenAI's pivot from pure language models to reasoning systems signals a major shift in AI development strategy
• The costly, compute-intensive nature of GPT-4.5 suggests we're reaching the limits of scaling traditional AI approaches

Read on, my dear:

• implicator.ai: Beyond Size: GPT-4.5 Signals OpenAI's Shifting Strategy

AI Photo of the Day

Prompt:

3w
A breathtaking underwater photoshoot featuring a newborn baby floating gracefully in a crystal-clear swimming pool. The baby appears weightless, with tiny hands reaching out and eyes wide open, capturing a moment of pure innocence and wonder. Gentle ripples surround the baby, and the sunlight filtering through the water creates a magical, ethereal glow. The scene is serene and beautifully composed, emphasizing the softness of the baby’s skin and the tranquility of the water. High detail, hyper-realistic, cinematic lighting, 4K resolution

From smarter emotions to deeper thinking: Silicon Valley’s latest AI wave

Blink once, and a wave of AI updates hits you. This week, it’s from all the big names: OpenAI, Anthropic (Claude), Alexa, Perplexity, Eleven Labs, Microsoft, IBM, and Hume AI.

OpenAI:

• Launched GPT-4.5, their largest model yet, with improved emotional intelligence and reduced hallucinations (37.1% vs previous 61.8%)

• Rolled out Deep Research to all ChatGPT Plus subscribers

• GPT-4.5 marks their last non-reasoning model before GPT-5

• Access starts with ChatGPT Pro ($200/month), expanding to Plus and Team users next week

Microsoft:

• Released Copilot with FREE unlimited Advanced Audio conversations and Think Deeper feature

Anthropic & Amazon:

• Claude partnership announced to power Amazon’s next-generation AI assistant, Alexa+

ElevenLabs:

• Launched Scribe V1, a new foundational model for transcription

• Claims to outperform existing transcription products

• Supporting 99 languages

Perplexity:

• Announced Comet, a new agentic browser (waitlist open)

• Released new Voice Mode with real-time multilingual capabilities

• Made Deep Research available to developers through Sonar API

IBM:

• Research team revealed CUGA (Computer Using Generalist Agent)

• System capable of handling diverse tasks across multiple domains

Hume AI:

• Introduced Octave, a text-to-speech model designed to generate lifelike and emotionally nuanced speech

• Octave can adjust tone, rhythm, and cadence based on context, enhancing human-AI interactions 

The field is clearly moving fast, with a notable trend toward more sophisticated reasoning abilities and improved human-AI interaction.

Zuckerberg Takes Aim at ChatGPT with New App

Meta plans to release its AI assistant as a standalone app in Q2 2025, joining the company's family of major apps like Facebook and Instagram. The move signals Mark Zuckerberg's aggressive push into AI.

Already serving 700 million users across Meta's platforms, Meta AI has yet to match ChatGPT's standalone success. The current Meta AI website attracts less than 10 million monthly views, trailing far behind major AI services.

Zuckerberg aims high. He expects Meta AI to reach over 1 billion users this year and become "the leading AI assistant." The company will follow ChatGPT's model with a paid subscription tier.

The pressure is intense. Meta employees report seven-day work weeks to meet deadlines. OpenAI's Sam Altman responded to the news with a playful tweet: "ok fine maybe we'll do a social app."

India leads Meta AI usage, particularly on WhatsApp. The assistant currently powers search across Meta's platforms and enables users to question recommended content.

Why this matters:

• Social media and AI are converging, creating new competitive battlegrounds
• Meta's massive user base could give it an edge in AI adoption

AI & Tech News

Europe's Most Valuable Tech Company Cuts 10% of Staff

SAP CEO Christian Klein slashed 10% of his workforce while collecting a €19 million paycheck - a 165% raise that towers over the company's 2.4% average staff increase. Despite becoming Europe's most valuable tech company and riding the AI boom to record profits, SAP's vision of efficiency apparently requires fewer humans to execute it.

Nvidia CEO Warns: Next-Gen AI Will Be Extremely Power-Hungry

Jensen Huang says upcoming AI models will need up to 100 times more computing power as they shift toward step-by-step reasoning—much like your math teacher always hoped. Nvidia is poised to profit handsomely, assuming customers can handle the eye-watering bills for their increasingly hungry artificial intelligences.

Amazon Debuts Quantum Chip “Ocelot”—Late but Cost-Effective

AWS enters the quantum race fashionably late, launching its first quantum chip dubbed “Ocelot”—a playful nod to cats and oscillators that might puzzle even Schrödinger himself. The kicker: Amazon claims its simple, sandwich-like design could slash quantum computing costs by up to 90%.

HP Shifts Production Away From China—90% Moved by October

HP CEO Enrique Lores announces a dramatic supply chain overhaul: by October, only 10% of products destined for North America will be sourced from China, describing it as a “very, very big change.” CFO Karen Parkhill adds that HP has already boosted inventory levels to mitigate potential U.S. tariffs, temporarily impacting the company’s cash flow.

Microsoft Launches AI Assistant Copilot for Mac

Microsoft today released a native macOS version of its AI-powered Copilot app, featuring dark mode and quick activation via Command + Space. Users can log in with their Apple ID, upload documents, and generate AI-powered summaries—now with unlimited usage following Microsoft’s recent removal of all usage limits.

Google’s Sergey Brin Calls for 60-Hour Weeks to Win AI Race

Google co-founder Sergey Brin has urged employees to spend at least 60 hours per week in the office to reclaim leadership in developing artificial general intelligence (AGI), reports the New York Times. Internally, Brin criticized colleagues doing the bare minimum but also cautioned against burnout from excessive hours.

AI Decoded 🔓: PowerPoint

Your guide to mastering AI tools - no tech degree required.

How to Outsmart PowerPoint (Without Losing Your Mind)

Create Stunning Presentations in No Time

Let’s be honest: making PowerPoint slides is about as exciting as watching paint dry. But what if you could get it done in minutes? ChatGPT can generate structured, well-organized slides, so you focus on the fun part—presenting.

Here’s how:

Method 1: Direct Approach

🚀 Prepare Your Content

• Gather source material (blog posts, reports, or notes).

• Decide on slide count and key topics.

🎯 Generate the Outline

• Ask ChatGPT to create a slide-by-slide breakdown.

• Request titles, bullet points, and speaker notes.

📊 Convert to PowerPoint

• Copy ChatGPT’s text into PowerPoint manually or use AI tools like Gamma.

• Add visuals, refine the text, and apply a design theme.

Method 2: Using Microsoft 365

📄 Refine in Word First

• Paste the ChatGPT-generated outline into a Word document.

• Adjust fonts, colors, and formatting.

📂 Export to PowerPoint

• Use Word’s “Export to PowerPoint” function.

• Select a template for a professional look.

💡 Final Touches

• Add transitions, animations, and speaker notes.

• Insert logos, headers, or footers if needed.

With ChatGPT, you’ll spend less time formatting slides and more time dazzling your audience. Try it and thank me later! 😉