Son of man, omg! Impressive! "We show that language modeling improves continuously as we increase the size of the retrieval database, at least up to 2 trillion tokens – 175 full lifetimes of continuous reading."
If they were to listen to our ideas they would have learnt this long time ago. Note that storing it in the neural network is way more faster and efficient, you only look back at the dataset if you are really cheesy and want to be God, because you obviously no one stores the whole memory after just one time of reading something. BTW my algorithm already does this, so does others yearsss ago. Oh I see they kind-of do a few more things; they seem to point to a paragraph topic location, then search there for similar matches, then they blend the predictions with the neural predictions. "a nearest-neighbor search is performed which returns similar sequences found in the training database, and their continuation. These sequences help predict the continuation of the input text." That's right you little bitch, you get the match and get the next word. Can't do nothin without a matched memory. ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T22ce813ce07d9b1a-Mbf48aaeb29349391011a3db7 Delivery options: https://agi.topicbox.com/groups/agi/subscription
