What is the difference between a ‘small AI model’ using a lots of data and a ‘large AI model’ ? Or what do you mean by small AI model vs large AI model?
A large AI model means a model with many layers of neurons and perhaps with skip connections, dropouts, and a complicated, large architecture in general. A model that has millions to billions of parameters, perhaps even trillions.
1 Like
Thank you for your answer.
Just to be cristal clear. There are 2 different concepts here:
1.the AI model based on the number of layers of neurons, connections, dropouts, size of the architecture etc…
2.The amount of data used to train the AI model
Is that correct?
Yes they are entirely different and if there is a difference in any of the two among two AI models then we can surely say that they are different and will perform differently.
1 Like
The model doesn’t contain the data - only the weight matrices, and a definition of the dataflows within the model (the number and size of the layers, the interconnections, activations, etc).
With the same model, you could use a large dataset for training, but only use small datasets for making future predictions.
So the size of the model does not imply anything about the size of the data set.
3 Likes