Classic Retro Video Game Blog

8 Bit Central - Retro Gaming Blog

May 2015 Retro Gaming Article


May 15, 2015 Retro Gaming Blog Post:

I want Daimler's self-driving truck to bring Deep-Q to my house for some retro gaming

As a kid I fully expected the Jetson's lifestyle to be a reality by now. However, as I navigate the landscape of today's technology, the idea of self-driving cars is frightening on several levels. Part of my doubts come from disbelief it will work when highways are clogged with self-driving vehicles yelling "I am here" into a wifi grid full of HD porn downloads. This leads to my second issue - Skynet.

Human demise may not result from an inbound asteroid or water shortage. We may succumb to Boston Dynamics' robot dogs chasing us into traffic where autonomous vehicles will steamroll us. Solar power seems great until the only things taking advantage of it are transformers roaming the otherwise vacant planet. Don't worry, animals will be just fine - until Apes learn to ride horses and use rifles.

Daimler's self-driving truck In other less frightening tech news, an AI venture has taught a computer to play Atari games without instruction. Scientists have developed a problem solving algorithm, named Deep-Q, that operates on a reward system. Allegedly, Deep-Q tries to solve problems in a way that maximizes reward. They call it a bio-inspired approach which sounds swell, but I'm not sure I understand how you reward an algorithm for good achievements. A computer won't eat a dog-treat, so I'm stumped. Perhaps this is why I'm not a scientist.

Interestingly, this Deep-Q system learned how to play 49 Atari 2600 games simply by keeping track of the pixels on the screen and using the achieved scores as a success metric. As with computers that can "play" chess, I've always felt that we mistake speed for intelligence. With chess, there are finite moves and a computer can evaluate each one very quickly.

As processor speeds increase, computers are able to analyze more info in shorter times. I've never believed that AI is real in the way that the human brain learns and adapts. I've always thought that computers can simulate intelligence, but they don't really know anything they are not told.

Atari 2600 quarter view My son figures out how to play Atari games quite quickly. By the second or third round, he knows the basics and how to maximize his score. Since Deep-Q would have to learn at the pace at which a game plays, there may not be a way to analyze every scenario in an accelerated manner.

Therefore, where my son figures out how to play a game in a matter of minutes - it may have taken Deep-Q weeks of playing 8-bit games to encounter, store, and analyze enough scenarios to figure out how to play.

Sorry, I'm very skeptical of AI, because those I've spoken with about it have never explained to me how a computer learns versus being fed every outcome along with a way to rank them. Here's my analogy- A computer can be programed to understand the syntax of written communication and can probably be programed to "write". A computer will never write the kind of story Stephen King can write. Mr. King is intelligent, while a computer is artificial.

On the other hand, I'm all for Daimler's kooky truck jumping the rails and bringing Deep-Q to my place for some retro gaming. We can start with Yar's Revenge and Space Invaders, but lets see how fast that bucket of bolts can figure out Zaxxon, Tempest 2000, and Tekken in the course of a night!

« Return to the main Retro Gaming Blog 2015



Printed: