I Finally Got a Chance to See DeepSeek's R1

Let me explain why DeepSeek's AI innovations are blowing people away (and possibly threatening Nvidia's $2 trillion market cap) in simple terms.

1/ First, some context: Training the best AI models is extremely expensive right now. OpenAI, Anthropic, etc. spend over $100 million on compute alone. They need massive data centers with thousands of $40,000 GPUs. It's like needing an entire power plant to run a factory.

2/ DeepSeek came along and said "LOL, what if we did this for $5 million?" And they didn't just talk the talk, they DID it. Their models match or beat GPT-4 and Claude on many tasks. The AI ​​world is (as my teenagers say) shaken.

3/ How? They rethought everything from top to bottom. Traditional AI is like writing every number with 32 decimal places. DeepSeek thought: "What if we just used 8 decimal places? That's still pretty accurate!" Boom - 75% less memory needed.

$FET

FET
Created with Highcharts 9.1.1
1.072
+4.79%