Hi guys, my name is Raj Singh, and let’s start today’s tech news.
The first update is about DeepSeek.
One of Deep-Seek’s engineers has revealed how their AI model was trained. What makes Deep-Seek special is that they managed to build a high-quality model at a very low cost, and yet it performs on par with some of the top AI models in the industry. Without using massive computing power, they’ve achieved genuinely impressive results. To me, this feels like a real game-changer. check the official CNN news.
how did they train it?
First, they started with a base model, and then trained it with well-refined data. Think of it like teaching a child. At first, you give the child a proper curriculum. As the child keeps learning, he grows and becomes smarter. Similarly, the language model grows as it’s trained
At the beginning, Deep-Seek fed the model with clean, structured data—the “curriculum.” After that stage, they let it explore, play, and experiment on its own, just like a child figuring things out. The model would make mistakes, but from those mistakes it learned.
To guide this process, they built a rejection system—separating good answers from bad ones. Wrong answers were discarded, and right answers were kept. This way, the model understood what was correct and what wasn’t.
Next came supervised learning. At this stage, it’s like teaching a kid to walk properly after he’s already taken his first steps. You guide him on how to do it better. This helped the model improve its performance even further.
Why DeepSeek AI Training Is a Game-Changer?
Finally, they applied reinforcement learning, where the model learns by trial and error—falling, getting up, and improving on its own.
And here’s the most shocking part: the entire training process cost under $300,000. When you compare that to other AI companies spending billions of dollars, it’s clear why this approach is being seen as a breakthrough.
Link to my MacBook Air M4 Review → “Read our detailed MacBook Air M4 review
Link to my Nothing Phone 3 Review → “Check out how AI is shaping smartphones in our Nothing Phone 3 review
