Epilepsy is one of the most common neurological diseases in the world while multiple sclerosis (MS) is a rarer condition that affects a relatively small number of people. The fact that both affect the biosignals one's brain emits makes them suitable for automatic detection based on electroencephalogram (EEG) readings through machine learning architectures. In this study, we take a look at the history of deep learning (DL) methods' use in both diagnosis and prediction of neural conditions in order to see how the field had evolved over the years and why certain architectures took precedence over the others. We also perform several tests such as seizure prediction and classification, using CNN, RNN and GAN models. We fine tuned the algorithms on open source datasets in order to train them to predict seizure occurrence for epileptic patients. We compare the results and the model that produces the best accuracy as well as the lowest number of false positives is then tested on a private dataset that consists of EEG readings from MS patients and healthy controls. The results are analyzed in order to analyze if our chosen architectures perform well on both types of datasets, what characteristics may influence their performance and how they can be further improved.