5 ms·
No, the unsolved problem of AI is continuous learning. We never stop learning, we don't have a "training phase". You are always updating your world model even w
by 10xDev 15d ago
No, the unsolved problem of AI is continuous learning. We never stop learning, we don't have a "training phase". You are always updating your world model even when you sleep. Also more quality training data does lead to greater learning efficiency as you have more priors to work with.
- suddenlybananas 15d agoWhy not both?
- matteoraso 15d ago>We never stop learning, we don't have a "training phase". We kind of do, if you count critical periods in childhood.
- nomel 15d agoProbable mechanism: "What we found is a rapid increase in GABA in children, associated with learning, ..." [1] [1] https://www.brown.edu/news/2022-11-15/children-learning https://www.brown.edu/news/2022-11-15/children-learning