Exactly. I've already hit the dreaded "knowledge cutoff" of 2021.
What happens when this training data is 10 years old? Where will we get new data?
In the meantime we'll get used to functioning with these models with assumption the "knowledge cutoff" will just be moved. But if there is little new data, how? Furthermore, how do we prevent feeding ML generated data to ML training in the first place?
What happens when this training data is 10 years old? Where will we get new data?
In the meantime we'll get used to functioning with these models with assumption the "knowledge cutoff" will just be moved. But if there is little new data, how? Furthermore, how do we prevent feeding ML generated data to ML training in the first place?