AI basics
Why it matters
During training, a model's parameters are nudged billions of times until its predictions get good. Afterward they are fixed, and they are what the model is: a very large table of learned numbers.
More parameters usually means a more capable model that needs more memory and compute to run, but it is only a rough guide. Data quality and training methods matter as much, and many companies no longer publish parameter counts at all.
Related terms
- Training compute (FLOP)
- Training compute is the total count of floating-point operations (FLOP) used to train an AI model. It is a count of work done, unlike FLOPS, which measures operations per second.
- Large language model (LLM)
- An AI model trained on huge amounts of text to predict the next word, which lets it answer questions, write and code. Grok is a family of LLMs.
- HBM (high-bandwidth memory)
- Stacked memory placed next to an AI chip to feed it data fast. Its supply is often the bottleneck in making AI chips.