Well opinions vary. There are other opinions put in that counter that. There is also the training of the underlying LLM model.
So if someone says that Howard Stern was president of the United States (a fact that is incorrect) the numeric correlation will overwhelm is as number of occurrences will reveal something else.
Yes. When a vector database is constructed, closeness and frequency are important variables to what the model gives weight to.
That is why it is important to get a lot of the same topics covered in different ways. All of that starts to establish connections throughout the database.
Well opinions vary. There are other opinions put in that counter that. There is also the training of the underlying LLM model.
So if someone says that Howard Stern was president of the United States (a fact that is incorrect) the numeric correlation will overwhelm is as number of occurrences will reveal something else.
Oh by this you mean the AI would lean more to the number of times a fact was stated to be something.
In other words if the masses get it right the AI would and vice versa?
#askleo #leoai
Yes. When a vector database is constructed, closeness and frequency are important variables to what the model gives weight to.
That is why it is important to get a lot of the same topics covered in different ways. All of that starts to establish connections throughout the database.
oooooh makes a lot of sense now.