Using oscillation to speed learning #

I've been reading Without Miracles. I had just read about children learning labels. The example was a child looking at a bird and then being told it was called a "bird", then looking at a different bird and being told it was a "bird". The child then has to generalize and consider that other small flying creatures are likely to be "birds" as well. However, upon seeing a butterfly and being told it was not a bird, the child then has to refine his understanding of what a "bird" is.

This form of learning, with overgeneralization followed by correction, may well be faster than undergeneralizing. For example, what would happen if the child only believed the two creatures he saw were birds, and did not label other small flying creatures as birds until explicitly told they were birds? He'd learn slower.

I was at the beach, looking at tide pools, when I realized there's a similarity to the child's learning. Rivers flow in one direction (usually) and have a certain amount of life in them. Beaches have water flowing in both directions, due to waves (short time scale), tides (medium time scale), and seasons (long time scale). The diversity of life in tide pools is far greater than what I've seen in rivers.

In the case of the child's learning, generalization (classifying everything as a bird) is one direction and correction (being told that some creatures previously thought to be birds are not) is the other direction. Could it be that oscillation leads to faster "learning" than a steady stream?

The advantage of learning and then correcting mistakes is that you can learn faster than if you were learning cautiously enough to avoid making mistakes. Many AI algorithms are of the cautious sort. It may be that it'd be better to have our games learn patterns very quickly and then keep a set of exceptions. For example, in Simulated Annealing, we slowly lower the temperature until the system stabilizes. It may be better to quickly lower the temperature, then raise it again, and continue oscillating for some time. In Neural Networks we change the neuron parameters very slowly. It may be better to change them quickly, then change them back if needed.

I'm enjoying the book a great deal but I haven't yet figured out how I might use this for my own games. My intuition tells me that there's something very useful here, but I haven't pinpointed anything specific.

Labels:

The Guild 2 #

Yesterday I took a look at the demo for The Guild 2. It's a strange game. It's somewhat like The Sims in that you have a character that has relationships with other people in the game, and you can start a family. It's somewhat like a role playing game in that you have classes, skills, abilities, attributes, experience points, combat, etc. It's somewhat like a city-building strategy game in that you can build mines, businesses, houses, and so on. The graphics and music are nice. The world feels alive—there are people wandering around, with jobs and relationships and government positions and alliances and feuds. While standing around, a random person came up to me and gave me a cake. I'm very impressed with the game world.

The problem is that someone forgot to design a fun game. In trying to do lots of different styles of games, it does none of them well.

The game is rather tedious. You have to keep track of your relationships with everyone else. These seem to be influenced by your position in society, your family, your religion, who you work for, and who works for you. It has more detail than The Sims (for example, you can buy objects and give them as gifts to someone) and more relationships to keep track of than in Tropico. The problem is that there's too much information, and you can't act on most of it. You only indirectly influence it. This is something that is impressive and thus appealing to game designers, but it's not actually any fun.

To run a business, you have to click on your cart, send it to the market, buy raw materials (prices are set by supply and demand) by dragging them into your cart, then send the cart back to your shop, drag the materials from the cart to the shop's inventory, click on an employee, and tell him to make an item. Then you wait for a while, and once the item is ready, you drag the item to the cart, send the cart to the market, and drag the item to the market stall to sell it (prices are set by supply and demand). It's impressive, but it's boring. Imagine if Warcraft had done this—you'd have to tell each orc to walk to the mine, pick up a tool, mine for metal, pick up the metal, carry it back to the storehouse, and then walk back to the mine. You have to do this for each business you run in The Guild 2. Although there is a way to set this on auto-pilot, there never should've been this much detail. The programmers wasted too much time with this.

There's also combat, positions in government, bribery, assassinations, thievery, a legal system (including court cases), and lots more complexity to this game. That's the real problem: there's too much to keep track of and do, and too little reward for doing it. You get to play out the drudgery of living as a serf in the Middle Ages. I can't even describe how bad the UI is.

There are also some amusing things in the game. Every building has a name: Better Homes, Lumpy & Liquid, The Hot Spot, Raw Iron Raw Power, A Waste of Paint, Roof Included, The Funny Farm. The help text is often strange:

... Scholars are also not adverse to compliments and gifts and will even let themselves be persuaded to dance.

Once the initial awkwardness has faded one can oft soften a reluctant scholar with soft embraces and gentle kisses, bewitch him in private conversation or even climb into a tub with him. But beware of using too much imagination in your courting, for scholars seldom appreciate variety.

There are odd player skills: Master of manure, Pack mule, Kama Sutra master, Exploiter, Great preacher, Deep sleep, Local club president, Face of innocence, Dorian Grayish. The description for the "Strong hair growth" skill starts out like this:

You have been blessed by nature with a curious biological phenomenon: whenever you must rot in the dungeon because of some misdeed, your hair begins to grow at a breathtaking speed. After a few hours, you look as though you have been mouldering for 30 years. ...

Even though the game doesn't look like it'll be any fun for most people, I do recommend that game developers try out the demo. Look for things that you might think would be cool in a game: multiple genres (from role playing to real time strategy to business simulation to action/combat), deep and complex simulation (relationships, businesses, employment, politics, law), open-endedness, etc. Try to figure out whether they are actually fun for a player or merely impressive to a developer. I would have loved to design and write a game like this. It's so impressive that it's sad to see that it's not at all fun.

Labels: