DeepMind claims its AI called RETRO matches the performance of neural networks 25 times its size, cutting the time and cost of training large language models (Will Douglas Heaven/MIT Technology Review)

https://ift.tt/3rNDgry

Will Douglas Heaven / MIT Technology Review:
DeepMind claims its AI called RETRO matches the performance of neural networks 25 times its size, cutting the time and cost of training large language models  —  RETRO uses an external memory to look up passages of text on the fly, avoiding some of the costs of training a vast neural network

from Techmeme https://ift.tt/3pA0kay