When OpenAI makes an improved model, do they just correct its “wrong answers” till it spits out more correct-ness?
In: Technology
No. They increase the size of the model, improve the training data, change the format of the model and such. Making the model bigger allows for more associations with words and means less mistakes. We don’t know at what point increasing the size of the model gives diminishing returns, but at the moment, increasing size is still giving good returns.
Latest Answers