👋 Hey, Fellow Humans! Welcome to this week’s Practical AI News, where we break down the wildest AI updates with a mix of hard facts, witty banter, and just a touch of existential dread.
This week? Elon Musk’s Grok-3 is here—but is it really the best in the world? Meanwhile, quantum computing took a major leap, Nvidia bounced back from its historic crash, and AI tools are getting more or less censorship, depending on where you look. Let’s dive in!
Take a listen to our podcast here and on YouTube. Here are the main chapters:
00:00 Grok-3 Dropped Hard
01:10 AI Models Benchmark
03:40 Grok's Unique Features and User Experience
04:48 Deep Search
07:23 Perplexity R1-1776 with DeepSeek
08:29 DeepSeek banned in South Korea
09:34 Quantum Computing Breakthroughs
11:06 Wrap-Up and Future Insights
🏆 Grok-3: The Best in the World? Or Just the Best at Gaming the System?
Elon Musk and xAI just dropped Grok-3, their latest ChatGPT competitor, and according to their own charts, it’s on top. Best in math, best in coding, best in science… basically the AI version of a valedictorian.
💡 The twist? They asked Grok the same question 64 times, then took the most common answer as the final score. Meanwhile, competitors like OpenAI’s models were tested once. Fair play? You decide.
🔍 Reality check:
Grok-3 is solid, but not a ChatGPT killer—OpenAI’s newest models still outperform it.
Less censorship than others, but not totally uncensored (Molotov cocktail DYI instructions not available anymore!).
Deep Search looks promising, like pulling in live data from tweets for fresher responses.
Still free for now! Test it while you can at Grok.com.
⚛️ Quantum Computing Just Got a Plot Twist
Big moves in the quantum world! This week we saw:
🚀 Japan launched the world’s first quantum-supercomputer hybrid. A step toward making AI even smarter (or at least more confusing).
🧪 Microsoft accidentally discovered a new state of matter (Majorana fermions). Who knew building better quantum computers would lead to redefining physics?
💭 What does this mean for AI? Still debatable, but we’re inching toward a future where quantum + AI = mind-blowing breakthroughs.
💸 Nvidia’s Wild Comeback: From $600B Loss to “Just Kidding”
Remember when Nvidia lost $600 billion in a single day after DeepSeek released an open-source AI? Well, investors have stopped freaking out—Nvidia’s almost fully recovered.
Why?
🔹 Musk’s Grok-3 was trained on a Memphis supercomputer, called Colossus, packing 200,000 Nvidia GPUs, with plans to buy 100K more every quarter.
🔹 AI companies still can’t get enough of Nvidia’s chips, keeping demand sky-high.
Lesson learned? Never bet against the GPU king. And “Shoulda bought those shares!”
🔥 OpenAI Drama & Perplexity’s Freebies
Elon Musk tried to buy OpenAI for $97 billion. They said no.
OpenAI quietly admitted that Anthropic’s Claude 3.5 is better at coding than their own models. Awkward.
Perplexity released R1-1776 (US Declaration of Independence year) —an uncensored DeepSeek R1 AI model, minus China’s restrictions. You can grab it on Hugging Face, but it’s too big for most GPUs unless you have a data center in your garage.
🏁 Quick Hits: AI Roundup
📂 Gemini now lets you upload documents for free. Need an AI to summarize your Excel sheets? Go for it.
🛑 South Korea banned DeepSeek AI over privacy concerns. Meanwhile, users are loving it.
🤖 Try It Yourself!
Grok-3 is still free (for now)—give it a spin at Grok.com and tell us what you think! Better than ChatGPT? Overhyped? Let us know in the comments.
Practical Help From Humans on AI—when opening many tabs of DeepSeek you can’t tell which tab is which because they are all called the same. We‘ve got a fix for you! The titles of your chats with DeepSeek will be visible as the tab titles.
🚀 Want more AI news like this every week? Hit Subscribe, and we’ll keep you updated—minus the fluff. See you next time! 🎙️
