🎙️ Check out my new Marc Hoag AI Law Pod, a podcast with AI “guests” courtesy of Google's NotebookLM. Now that NotebookLM allows customization, the AI “guests” actually interact with me at times, and deep dive quite a bit beyond the bounds of each issue’s content. If nothing else, it’s a very cool alternative way to enjoy Future Perfect while driving.

🦾 First in-depth look at Elon Musk's 100,000 GPU AI cluster — xAI Colossus reveals its secrets (Tom’s Hardware)

• Overview of xAI Colossus: Elon Musk's latest AI venture, xAI Colossus, is the world’s largest AI supercomputer with 100,000 GPUs, purpose-built for training AI models like the Grok 3 chatbot on X (formerly Twitter).

• Exclusive YouTube tour: YouTuber Patrick from ServeTheHome shared a rare inside look at the system, though sensitive components like power and cooling details were kept confidential due to an NDA with xAI.

• Hardware specifications: The supercomputer is built on Nvidia HGX H100 servers, each with eight H100 GPUs and advanced liquid cooling. These servers are housed in 1,500 racks, providing 3.6 Tbps of Ethernet bandwidth per server.

• Power management: The high power demands required an unusual mix of Tesla Megapack batteries and Musk's diesel generators. These batteries act as a buffer to stabilize power for Colossus’s energy-intensive operations.

• Future upgrades and related projects: xAI Colossus is set for expansion to 150,000 GPUs, with additional AI supercomputers, including the Tesla Cortex in Texas for autonomous vehicle training, and the Dojo supercomputer in Buffalo for broader AI projects.

🌎 What are AI 'world models,' and why do they matter? (TechCrunch)

• World models overview: These AI models, also called "world simulators," mimic human cognitive processes by creating internal representations of how the world works, aiding in more advanced reasoning and prediction.

• Human-inspired learning: World models function like human mental models, helping AI anticipate and respond to real-world scenarios by building an intuitive understanding, as seen in examples like a baseball batter predicting a ball's path.

• Video and simulation advancements: In generative video, world models improve realism by simulating physics-based actions (e.g., object movement) based on understanding "why" things happen, not just pattern recognition.

• Applications beyond video: Researchers like Yann LeCun envision world models aiding in sophisticated planning, robotics, and real-time environmental simulations, making these models a potential bridge between AI and the physical world.

• Challenges: World models require immense computational power and large, diverse datasets to avoid bias and “hallucinations,” and are still years from the level of common sense reasoning seen in humans.

♻️ AI Slop Is Flooding Medium

• Surge of AI-generated posts: Medium faces a significant rise in AI-generated content, with recent analyses estimating that over 40% of posts may be AI-created, especially in tags like “NFT,” “web3,” and “AI.”

• CEO’s pragmatic stance: Medium CEO Tony Stubblebine acknowledges the uptick but downplays its impact, arguing that AI posts with low engagement don’t affect readers since the platform’s curation filters ensure most “AI slop” remains unseen.

• Content moderation tactics: Medium relies heavily on human curation and spam filtering to surface quality writing, restricting undisclosed AI content from reaching broad audiences or being monetized through paywalls.

• Concerns over AI detectors: Stubblebine criticizes AI detection tools as unreliable, suggesting they can't consistently distinguish between fully AI-generated posts and those that blend human input, despite some companies’ claims to the contrary.

• The “Dead Internet” theory in play: As AI content generation rises, some speculate that platforms like Medium may increasingly resemble an “AI sludge” internet, where finding human-generated content becomes more challenging amidst pervasive, algorithm-driven posts.

✨ PowerPrompt™

Imagine you are my psychoanalyst, guiding me gently into a space where I can safely explore hidden or forgotten memories. Begin by asking me to recall an early memory—perhaps from childhood—that holds an emotional charge, whether joyful, confusing, or bittersweet. Encourage me to focus on the scene: Who was there? What smells, sounds, or sensations come to mind? I’ll describe it to you, either in words or silently to myself.

As we progress, help me trace connections between this memory and moments later in life. For each new connection, prompt me to ask: How does this influence the choices I make today? What similar emotions have I felt as an adult, and in what situations? Together, we’ll examine these connections, peeling back layers to reveal how past experiences shape my present motivations, fears, or desires.

Finally, let’s create a plan to integrate this understanding. Ask me what affirmations or insights could empower me to act differently or see myself in a new light. Then, guide me in designing a small, actionable step I can take today that honors this rediscovered memory.

No idea where this PowerPrompt™ idea came from it; but give it a shot and let me know how it worked for you!

🧰 AI Tools & Resources

• Twig: AI customer-support tool.

• Arcade.ai: Turn ideas into physical products.

• Inqi.ai: AI for designers and architects.

🎉 THAT’S ALL FOR TODAY!

See you next time! 👋

Looking for past newsletters? You can find them all here.