Parameters (கட்டுப்பாட்டு சுவிட்சுகள் / Control Knobs)

Let's mix music on a giant DJ Console! (டிஜே கன்சோல்ல மியூசிக் மிக்ஸ் பண்ணுவோமா!)

Technical Meaning: அளபுருக்கள் (Alapurukkal) - The internal variables (weights and biases) that an AI model adjusts during training to learn.

The Core Idea

When you hear an AI model has "70 Billion Parameters," it refers to the total number of internal connections (weights and biases) the network uses to make decisions. Parameters are the actual "knowledge" of the AI. When an AI learns a new fact, it isn't saving a text file; it is slightly tweaking a few thousand parameters to reflect that new knowledge.

The Origin Story

Early neural networks had only a few thousand parameters, limiting them to simple tasks like recognizing numbers. As hardware (GPUs) became more powerful, researchers scaled models into the millions, then billions, and now trillions of parameters. The more parameters a model has, the more nuanced patterns it can memorize—but it also becomes massively expensive to run.

The Tamil Analogy

Imagine a massive DJ Mixing Console (ஆடியோ மிக்சர்) at a grand concert.

If you have a small radio, you only have 2 knobs (Volume and Tuning). You can't customize the sound much.

An LLM is like a DJ console with 70 Billion tiny knobs.

• One knob slightly increases the bass.
• Another knob ensures the singer's voice doesn't echo.
• Another knob controls the background drums.

During training, the AI acts like a sound engineer, constantly twisting and tweaking all 70 Billion knobs until the music (the answer) sounds absolutely perfect. The final position of all those knobs is the trained AI model!

Try It Yourself