Jump to content

  • Log In with Google      Sign In   
  • Create Account






Some results

Posted by polyfrag, 13 August 2014 · 290 views

Working on economics. AI, to be specific. Competition now works using an inverse utility function and profit maximization.

Profit maximization is: given a set of "demand points", each with quantity demanded and maximum acceptable price, adjust the price level such that the maximum profit (revenue, actually, so far) is attained. A lower price might capture more of the market segment and thus get a greater quantity, but it will be multiplied by a lower price. A higher price might get less of the market but it will get a greater return per quantity. Kind of like intersecting supply and demand. But it tries all combinations.

The inverse utility function is used to calculate the maximum price that would be acceptable to a demander. Assuming it is a physical resource, it is a trade-off between price and distance; a different supplier might be just as acceptable as another (equal utility rating) depending on its price and distance. To steal a customer from a competitor, your building's utility rating has to be greater. This leads to price competition as different AI's keep lowering their prices until they reach the lowest profitable price or reach their building's production capacity.

So far there's no lower limit on price that would be due to cost of inputs and operation.




As a fellow dev that has begun work on AI design, I am interested in knowing more about some of the pitfalls y0u have ran across.

I have run across getting buried too deep in complexities. In moments of clear thinking I see the light eventually, making strides and milestones. I thought that it might be better to abandon trying to logically think through how the AI should play and try to use neural nets. This however, upon consideration, would require elaboration over how the neural net would be hooked up in such a way as to have enough information to make informed decisions and not be too complex. And, dealing with freezing lag right now, a neural net implementation would be at least a magnitude greater in processing complexity due to extra neurons or layers.

Lag is a point of worry for me as well. I don't want people to feel that the AI takes a year to make a single move.

September 2014 »

S M T W T F S
 123456
78910111213
141516171819 20
21222324252627
282930    

Recent Entries

Recent Comments

PARTNERS