[ExI] DeepMind AI teaches itself about the world by watching videos

spike spike66 at att.net
Fri Aug 11 16:13:58 UTC 2017

-----Original Message-----
From: extropy-chat [mailto:extropy-chat-bounces at lists.extropy.org] On Behalf Of BillK
Subject: [ExI] DeepMind AI teaches itself about the world by watching videos


“We want to build machines that continuously learn about their environment in an autonomous manner,” ...


>...What is the AI going to select as the 'best' role models?  Trump??
Stalin????  Pop Eye????  BillK


Popeye, now there's an idea.  A few hours of those things it would make comments such as:

AI yam what AI yam.

Hey, don't blame me, ExI-ers.  BillK set me up for that punchline; there is no way I could resist.  Such a great straight man is that lad.

Imagine an AI being trained on a typical 1930s era Popeye cartoon titled I Yam What I Yam.  The final scene is described in a fashion we might today consider a bit politically incorrect:

>...Meanwhile, Olive Oyl is successfully fighting off dozens of the Indians piling into the cabin, still screaming for Popeye's help. Popeye arrives at the cabin, dodging a barrage of arrows and fighting off dozens of Indians, even grabbing an entire group of them and delivering a punch that transforms them into a pile of Indian Nickels. With the bulk of the Indians defeated, one last batch of them is left to deal with, who shoot Popeye full of arrows, which he is completely unharmed by. Popeye promptly pulls out a can of spinach, eating both the vegetable and the can itself. He punches the row of remaining Native Americans, causing a domino effect. The final gag shows Popeye punching out the giant tribal chief, causing him to lose his outfit and become another type of Indian, Mahatma Gandhi. The cartoon ends with Popeye singing "I'm Popeye the Sailor Man!".

Ok then.

It is easy enough to imagine early AIs being insulting as all hell.  If they are in the form of words on a computer, we will deal with it the way we laugh off 1930s era Popeye cartoons.  But what if the AIs are put in androids that look like humans?  We need a paradigm shift.


More information about the extropy-chat mailing list