On a less serious note, it looks like the model is making mistakes, then good outputs, then bad, then good stuff, repeat. We all do that. Half of what I say/do is no good. Then there is the good outputs.
More seriously, more, diverse, data/context IS better, bigger model IS better. More sensor types (infrared, night-vision, vision, ultra-sound) IS better. It leads to better generalization. But if you you take it all from one spot like from a small dataset, it will spoil. You need to lower cost (compression type) and NOT error rate or error level! So that it learns all the general fundamental basic lowest parts to build larger features and have a tighter small world predictive model. That's the best thing you can have, it will work better at compressing if you give it more data, but works better if has most diverse data. ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Tc1f2c133ae3e4762-M134b9935ea7ab23f0d525c61 Delivery options: https://agi.topicbox.com/groups/agi/subscription
