1) If the LC/Perplexity contest used more data, it'd be more accurate yes, but 1GB is "ok for now" purposes.
2) If you want to sell a product, train the fricker on 100TBs if you can, was all I'm saying. It'll cost a billion $ but is worth it. +3) As for what you just said, yes if we have 7 billion humans on Earth and all their data, and think of the humans (or their data) as nodes in a neural network, and try to change their beliefs until the network is aligned and predicts "better", we'll end up with a high compression or low perplexity score. Yes we could use this to see how misaligned we are, though it is not AGI, but now politics and law... ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T358f938c1cfb5c51-M5035e4ed54f76646d436e1bc Delivery options: https://agi.topicbox.com/groups/agi/subscription
