AGI is all about predicting/generating the future data from old data. Lossless 
Compression/decompression is all around you and allows you to extract free 
insights/energy from the fuel/knowledge. It allows you to understand the data 
and generate more data like the data you want. The Hutter Prize shows they can 
compress and, get it all back perfectly. So the compressor/predictor is 
probably more likely to generate also unseen data, too, not just the dataset we 
use in the Hutter Prize. Currently it does work but loses long term meaning if 
you feed it a prompt to generate a story/sequence/data (or transform data, 
data>data, same thing), unlike GPT-2 which can do it super well. If they add 
that in then it gives better compression/prediction, but fewer have that 
mechanism in as of currently.
------------------------------------------
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/T887757e45bfd1342-M8ef6339e15c2ac5d3ea82e84
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Reply via email to