- AI Fire
- Posts
- 👁 GodmOm's GPT SpatialBlind
In partnership with
Fei-Fei Li (that godmather of AI) just called out the biggest blind spot in AI. Plus, OpenAI’s new data center project is bigger than the Manhattan Project (literally).
What are on FIRE 🔥
IN PARTNERSHIP WITH SECTION
On Jan 29 from 12 - 4 PM ET, join an afternoon of AI marketing strategy sessions, from content creation to creative brainstorming.
Section is brining together the leading voices in AI for marketing to share what’s hype, what’s table stakes, and how the marketer’s playbook has changed with AI.
AI INSIGHTS
Source: Fei-Fei Li
You know Fei-Fei Li, the Stanford professor who built ImageNet and basically helped launch the deep learning revolution? She just dropped a manifesto saying LLMs have hit their ceiling. The next frontier is Spatial intelligence.
In AI terms, that’s like Steve Jobs dropping a blog post about the future of computers. So spatial intelligence is the stuff humans and animals use daily without thinking:
Catch keys tossed across a room
Read a room full of chaos (like a firefighter in smoke)
Build models like Watson & Crick did to crack DNA’s shape
Li’s argument is that if we want AI to do more than just talk, we’ll need “world models” with 3 big capabilities:
Generative: They can create 3D environments that follow real physics
Multimodal: They understand text, images, video, depth maps, and actions
Interactive: They predict what happens when you do something
Fei-Fei’s World Labs is already shipping tools like Marble, which turns text prompts into explorable 3D scenes.
Soon, you’ll prompt it to build a world you can explore. Because like Fei-Fei says, “the limits of my language mean the limits of my world.”
PRESENTED BY HUBSPOT
Want to get the most out of ChatGPT?
ChatGPT is a superpower if you know how to use it correctly.
Discover how HubSpot's guide to AI can elevate both your productivity and creativity to get more things done.
Learn to automate tasks, enhance decision-making, and foster innovation with the power of AI.
AI SOURCES FROM AI FIRE
1. Vibe coding 2025: Most AI tools are bad. Use these instead. I tested them all. Learn which platforms are great for pros and which for beginners.
2. How to master 90% of building AI agents in 30 minutes. This is the SIMPLE, 3-step launch, focusing only on components that actually matter.
3. AI creativity is trapped. Use this 8-word key - A new Stanford paper shows how to get truly varied, creative ideas.
4. How I run powerful LLMs on my laptop. This setup gives you the easy path to free, private, and unlimited AI power on your Mac or Windows.
TODAY IN AI
AI HIGHLIGHTS
😂 Talking about physical AI, there’s the funniest robot video we’ve ever seen. He DIY’ed his genius robot halloween, how excited he’s by the creation. See the video.
🧠 If you're still wondering why AI struggles with facts vs. beliefs, even on your simple questions? Then this report is the perfect answer for you. It’s just the missing piece.
🎮 Someone built an AI Agent called Steve for Minecraft. Just say “mine iron” or “build a castle” and it spawns multiple agents that self-coordinate like pros. Try for free here.
🎬 Amazon’s “House of David” used over 350 AI shots in Season 2? Showrunner calls it “magical filmmaking” but many call it cheap. See this AI scene, what do you think!?
📚 Wikipedia just told AI companies to stop scraping its content. It’s urging everyone to use its paid API for fair access and proper credit. Is it talking to Grokipedia!?
🛡️ Google launched Private AI Compute, Gemini’s full cloud power, but with locked-down privacy. No one, not even Google, can see your data. Listen to the article here.
💰 AI Daily Fundraising: Crusoe secured $1.38B, bringing its valuation to $10B. Leading investors include NVIDIA, Mubadala Capital, and Valor Equity Partners.
HOT FREE EBOOK OF THE WEEK
This book offers a hands-on guide to building GPTs that learn, adapt, and respond in your unique style.
NEW EMPOWERED AI TOOLS
✍️ UndetectedGPT humanizes AI writing and bypass detection
🍌 6 new things you can do with Nano Banana in Google Photos
🧾 MentorSpace gets personalized mentorship for your journey
🔈 Auphonic gets the best sound quality for your videos and more
AI CHART
According to the new report from Epoch AI, AI data center spend is on track to pass $300B by end of 2025, nearly 1% of US GDP, topping the Apollo program and the Manhattan Project combined.
The headline example is obviously OpenAI’s Stargate Abilene with:
$32B capex
1 GW power draw
250× GPT-4 compute
450 soccer fields of land
Thousands of workers
2-year build timeline
It’s so big that latency doesn’t matter anymore. You could bounce data off the Moon and still be bottlenecked by model generation time, not transfer time.
AI model inference takes 100× more time than sending data around the globe. Data centers go where the power is cheap and available, not where the users are.
Most projects start with on-site natural gas, then layer in solar + wind via grid interconnects. Only a few countries can handle many >1 GW AI sites.
For reference: 1 GW = power for ~1 million homes. AI might still decentralize in the long run, but right now → The future is shaped by who can build, and power faster.
We read your emails, comments, and poll replies daily
How would you rate today’s newsletter?Your feedback helps us create the best newsletter possible |
Hit reply and say Hello – we'd love to hear from you!
Like what you're reading? Forward it to friends, and they can sign up here.
Cheers,
The AI Fire Team
Reply