This is why things like Google Pathways are so important. (Note: the following is based only on my understanding of what Pathways is and how it works, which may

This is why things like Google Pathways are so important. (Note: the following is based only on my understanding of what Pathways is and how it works, which may be incomplete or inaccurate.) As far as I understand, Pathways allows exceptionally parallelizable training of neural networks. That, in turns, means the networks can be bigger and bigger, with more and more training data, and still stay within reasonable training times. If scale truly is one of the most important factors in generalizability and performance of AI (which it seems to be), optimizing the training of larger and larger models will bring us to that AGI future sooner.