“Artificial intelligence. Here’s a little synopsis of alphago vs. Alphago zero. The original alphago was able to beat the best human go players in the world. It used a combination of human strategies and its own algorithms.
Alpha go zero was invented soon after. With the advent of alphago zero it used Tabula rasa, meaning, from scratch, ie. self learning. Alphago zero beat alphago 100 games to 0. Alphago zero taught itself in 72 hours by playing 4.9 million games against itself.
The average human can play a game of go in about 45 minutes to an hour. That means a human would have to play Go for 570 years to play 4.9 million go games to catch up to alpha go zero speed. This means alphago zero is about 70000 times more powerful than a human mind currently.
Neuroscientist Sam Harris in a podcast suggests that in the future a General AI (GAI) could in one week process the equivalent of 20000 years of computations. Meaning the AI is a million times more powerful than one human mind. Currently we’re at about 70000 to 1 human mind. So for AI to reach a million times that of a human would be a factor of 14.
Moores law states computing capacity doubles every 2 years. So, 70000 doubling every 2 years we may reach 1000000 times human level in 8 years. General artificial intelligence maybe reality by 2027. *note: that GAI internal algorithms/adjustments/optimizations may quicken or backfill Moores law rate of progression.” ~Carl Seppanen
NOTE: AI Race: 6-months GAI start edge equals 500,000 years advantage
Good convo – GAI,etc
https://www.youtube.com/watch?v=tAGeN5lrxws&feature=youtu.be
https://deepmind.com/documents/119/agz_unformatted_nature.pdf
