If they want to actually challenge the machine, have it understand a Warhammer 40K codex.

Yeah, let's teach a computer to play as the necrons.

Don't blame the Japanese. Sure, they're also trying to make self-aware AIs, but for different reasons: so that they can make hot robot chicks that only want...to be loved. That may almost be as effective at stopping the robots with the monkey army. Peace and love, man!

"Never trust a computer you can't throw out a window." -Steve Wozniak


...you would actually make love to a piece of plastic? :huh:

Now that is impressive! However it's still just a computer, it cannot understand what it "learns", it just executes it's programming. There's no real, in-depth understanding of how human brains work, so how could anyone possibly synthetize that with a computer?



One doesn't "make love" to a six toy. ...I hope :unsure:

http://channel.nationalgeographic.com/series/taboo/4599/Overview :facepalm:

I doubt we will have computers that become "Self Aware" because they are still built upon common processes of If, What-if, Else-If, etc... . Also computer code doesn't have the means to modify itself, but it does have the capabilities to write code that it can follow. However once a computer program learns how to "re-write" itself fixing mistakes in its coding then we should be alarmed. Thing that unnerves me is that there was a virus out there that started mimicking biological viruses, but trying to evolve itself.

It'll try to win the game... :ermm:

That's pretty cool, maybe it could be the next step to self aware AI, since it can learn and act upon what it learns. If it read about computer programming could it write code?

My gods, the amount of misinformation in this thread is staggering! Some points for you to consider:

• The game was neither Civ 5 nor Civ 2, but rather the open source FreeCiv.
  
  
• The research and http://people.csail.mit.edu/regina/my_papers/civ11.pdf are misattributed in the OP. Both Branavan and Barzilay are from MIT. Only Silver is from UCL.
  
  
• We still don't have an objective definition of what self-awareness is, so any claim for or against computers being self-aware is bunk. The closest we've gotten is http://en.wikipedia.org/wiki/Mirror_test, which is trivial for a computer program to pass, and makes no sense in this context.
  
  
• Self-modifying programs are fairly common, especially in the malware industry. In fact, executable code is just data, and modifying this data is technically trivial. Professional programming rarely involves this, as it makes the program much harder to debug and maintain, and does not help solve typical computer science problems.
  
  
• Modern AI, including the techniques used in this study, are very different from the typical logic-driven program. Nowadays AI is "soft". It uses http://en.wikipedia.org/wiki/Artificial_neural_network, http://en.wikipedia.org/wiki/Statistical_learning and various other data-driven techniques to find patterns, which is very similar to how animals and humans learn. A good practical example is Google Translate. Sure, it's nowhere near a human, but it does produce quite meaningful results. What you may not know is that Google Translate is completely data-driven. Google does not employ a single linguist. In fact for some of the supported languages, no one at Google even speaks them. Instead, the program is trained by being given a text that has been previously translated to several languages, and the program infers meanings and grammatical rules by itself by finding patterns of similarity between the pairs of texts (the original and its translation). If you'd like to read more about it, http://en.wikipedia.org/wiki/Google_translate#Translation_methodology is a good place to start (follow the reference links!).
  
  
• Someone mentioned that the manual is a database of all that can happen in the game. Aside from the fact that it's untrue, that's not the point. The points is that the manual is written in English, a natural language, and was not preprocessed when given to the program. The program itself was not given any rules about English either - all it saw was sequences of tokens (words), each token as opaque as any other. It was able to recognize causes and effects from simply matching words to its experiences (aka, its feedback function).
  
  
• WTF is wrong with people being creeped out by advances in AI and machine learning? Sure, most of it is joking, but this comes up every time there's an AI discussion, and the underlying thought is damaging. This kind of thinking is what kills scientific progress.

Also, http://xkcd.com/386/.

The vary paper you just linked to says it was Civ II.

It returns a logic error.

They showed an excerpt from the Civ II manual, but further on you can see that the actual implementation was done on FreeCiv (search the PDF for "free"). I do agree that their repeated references to Civ II are a little confusing though (they consider FreeCiv an exact clone of Civ II, which it isn't).

Oh come on, man. It's the open source version of Civ II. That still makes it Civ II.

Excerpt from the section titled "Experimental Setup":


I guess they refer to Civ II in their paper since it's immediately more recognisable and less confusing to the reader than first explaining what Freeciv is.

Thanks for pointing out it wasn't Civ 5 (it doesn't really help that the articles I read didn't mention what version it was and all used Civ5 screenshots, but hey that's journalism for ya).

If that database bit was pointed at me I'd like to make clear that I didn't mean to say it shows everything that can happen in the game, just what "pieces" are available and what tech will be available + all the requirements for them. You were right about it not being a database though, while the manual is rather extensive, it doesn't offer a way to individually access the pieces of data, so I should have known that I was making a flawed statement, sorry 'bout that. :wink_smile: