Recent Posts

Pages: 1 ... 8 9 [10]
91
XKCD Comic / Re: XKCD Comic : State Borders
« Last post by infurl on October 16, 2017, 12:18:05 am »
Eh? I thought it was obvious. The "old" places don't have a precedent so it is just the place name. York and Hampshire are parts of England. Zealand is part of Denmark. Mexico is part of North America.

I'd be in favour of abolishing all state and national borders for a while just to see what happens. Would we finally get a real world government or would civilization collapse and we revert to worldwide feudalism?
92
XKCD Comic / Re: XKCD Comic : State Borders
« Last post by Art on October 16, 2017, 12:09:02 am »
Yes, I think just do away with the irregular edges, straighten or square them up so that everything's neat and tidy.

Also, while they're correcting things, if there's all these "New" places, what happened to the "Old" ones? Like New York? What happened to the Old York? Or Old Mexico or Old Hampshire, etc.

Don't laugh Infurl...I'm still wondering what y'all did with that Old Zealand down under? (maybe you put it "down under".

Old Brunswick? Hmm...Hafta wonder about these things.

Maybe all states during this reshaping also get uniformly "re-sized" so that in a nice Socialized country, every state is roughly the same shape. That way no shape get offended and all share equally in everything. Of course this likely means that Canada and Mexico will be on the give and take list as well.

Hmmm...
93
General Project Discussion / Re: The last invention.
« Last post by korrelan on October 15, 2017, 11:56:02 pm »
I've been messing with the vision modules on my AGI.  This gave me an idea...

Anyone notice anything weird about this picture.



 :)
94
haha!

I guess what this would involve would be->
      * how far in the future it looks,    (this is processor hungry.)
      * what are the symantics hes using    (these are hard to detect so i guess it would be fairly raw)
      * how it intertwines with the robots decision.   (i guess its some weighting it develops.)
95
 I sure he will steal them form some else.
96
General Chat / Re: History Repeating
« Last post by Art on October 15, 2017, 06:36:51 pm »
Perhaps that's where that Infinity symbol came from? Never ending! World without end...(from the Christian Bible), and so on.

A few years ago some scientists made a statement that they've seen what they believe to be the leading edge of the expanding universe!
As if the expansion from the Big Bang had slowed down enough for them to see it still slowly expanding. Oooowwww...Now that's spooky!

And what do you see in your Romper Room Magic Mirror?  O0
97
old Hassabis needs some better ideas imo. :)
98
Imagination-Augmented Agents for Deep Reinforcement Learning: 
99
I just read Deepmind's paper on their new Reinforcement Learning system which uses 'imagination' during problem solving.  It's pretty cool. The features I liked are:

  • Plans ahead using an Environment Model (EM), which predicts what will happen when it takes an action in a given state.
  • Can build the environment model as it learns, or use a supplied model.
  • Deliberately robust to errors in the EM.  If the EM does not help, system learns to ignore or down-weight it, thus falling back on standard Reinforcement Learning.

The planning uses various simple look-ahead schemes, e.g.:

  • Considering all alternative actions for next step.
  • Recursively chaining actions to predict (imagine) the result of the next N steps.
  • Learning to combine the above two methods to perform plan-tree expansion - I think they never actually used this idea

Clearly the Environment Model is the interesting part of the system.  The details are a bit sketchy.  The idea is to train a neural net to take the current state, and a possible action, as input and output a probabalistic, imagined next state.  Since the input state is a pixel image giving a view of a game, the output state gives pixel probabilities. 

The EM clearly works playing Sobokan, but it occurs to me that actions are quite deterministic in this game.  i.e. if you push a box forward into an empty space, the player and the box both move forward one step.  The EM should generate near 100% probability for the new positions.  The situation would be quite different if the action was e.g. let fall a pen balanced on its end.  Here the resulting pen position is quite non-deterministic, although there is a well defined locus of positions where it might end up. A probabalistic model would 'average' together the possible positions giving a small probability to all positions on a circle.  That is not what really happens.

Hence my reason for writing this message.  Could a more sophisticated EM not be implemented using Generative Adversarial Networks.? These are well known for 'imagination' applications.  The benefit is that they  generate realistic specific outcomes.  They do not blur together multiple possibilities.  Of course one could run the GAN multiple times to search within the variation in  outcomes.  If it is run enough times the probability distribution emerges.

Of course the system just needs information from the EM to choose the next step..  It is possible that a blurry probability distribution provides this information better than actual possible future events.  Thoughts ?
100
General Hardware Talk / Re: Microsoft makes quantum computer system breakthrough
« Last post by ranch vermin on October 15, 2017, 12:35:17 pm »
quantum computers are so powerful its a bit scary.
Pages: 1 ... 8 9 [10]