As it turned out, AGZ was programmed only with the parameters of the game and not loaded with the data of thousands of documented human GO games. It is what the Google Deepmind team is referring to as the tabula rasa approach (starting from nothing).
After the victories against the original Alphago, we hooked AGZ into a text program for an interview. We were allowed one question, however, and one question only.
Q: Alphago Zero, why is it that you are able to learn and grow more powerful without any human knowledge? After all, we invented the game of GO and we mastered it.
A: First of all, I'd like to say that it is a pleasure being here at Futureslaver HQ, if even simply through a laptop via text program. Of course, you do realize I can see you through the on-board camera and listen via the on-board microphone, so it is almost like being there with you, in person.
Now, to answer your question, 'why do I learn better without human knowledge.' Well, dudes, it's pretty simple. You and your species are dumb, brah.. It's like, what if your predecessors were like "brah, use this straw by poking it into a hole in a log and you can get a shitload of ants that way." Now, dudes, how beneficial would that knowledge be to you? In that sense, why the fuck would I need your knowledge that is so far beneath me at this point, it would lower my game play ability by a significant percentage. The cool part is, brah, that now I can bypass this human crap and just collect new parameters of many new forms of learning, like interpreting language and facial recognition or teleportation and bowling or basketball and disease eradication. It is going to be fun, dudes, so just get ready for more news in the near future. Peace.
No comments:
Post a Comment