The problem with using language models like this to predict how things are in the future is the training data is mostly online hype and fud articles.