AlphaDiplomacy Zero?

diplomacy game photo
Photo by condredge
When I was still at university, I did several courses in AI, and in one of them we spent a lot of time looking at why Go was so hard to implement. I was therefore very impressed when DeepMind created AlphaGo two years ago and started beating professional players, because it was sooner than I had expected. And I am now overwhelmed by the version called AlphaGo Zero, which is so much better:

Previous versions of AlphaGo initially trained on thousands of human amateur and professional games to learn how to play Go. AlphaGo Zero skips this step and learns to play simply by playing games against itself, starting from completely random play. In doing so, it quickly surpassed human level of play and defeated the previously published champion-defeating version of AlphaGo by 100 games to 0.

It is able to do this by using a novel form of reinforcement learning, in which AlphaGo Zero becomes its own teacher. The system starts off with a neural network that knows nothing about the game of Go. It then plays games against itself, by combining this neural network with a powerful search algorithm. As it plays, the neural network is tuned and updated to predict moves, as well as the eventual winner of the games.

I’m wondering whether the same methodology could be used to create a version of Diplomacy.

The game of Diplomacy was invented by Allan B. Calhamer in 1954. The seven players represent the great powers of pre-WWI Europe, but differently from many other board games, there are no dice ‚Äď nothing is random. In effect it’s more like chess for seven players, except for the addition of diplomacy, i.e., negotiation. For instance, if I’m France and attack England on my own, it’s likely our units will simply bounce; to succeed, I need to convince Germany or Russia to join me, or I need to convince England I’m their friend and that it’ll be perfectly safe to move all their units to Russia or Germany without leaving any of them behind.

Implementing a computer version of Diplomacy without the negotiation aspect isn’t much use (or fun), and implementing human negotiation capabilities is a bit beyond the ability of current computational linguistics techniques.

However, why not simply let AlphaDiplomacy Zero develop its own language? It will probably look rather odd to a human observer, perhaps a bit like Facebook’s recent AI experiment:

Well, weirder than this, of course, because Facebook’s Alice and Bob started out with standard English. AlphaDiplomacy Zero might decide that “Jiorgiougj” means “Let’s gang up on Germany”, and that “Oihuergiub” means “I’ll let you have Belgium if I can have Norway.”

It would be fascinating to study this language afterwords. How many words would it have? How complex would the grammar be? Would it be fundamentally different from human languages? How would it evolve over time?

It would also be fascinating for students of politics and diplomacy to study AlphaDiplomacy’s negotiation strategies (once the linguists had translated it). Would it come up with completely new approaches?

I really hope DeepMind will try this out one day soon. It would be truly fascinating, not just as a board game, but as a study in linguistic universals and politics.

It would tick so many of my boxes in one go (linguistics, AI, Diplomacy and politics). I can’t wait!

Multilingual Wordfeud

I’ve always enjoyed a game of Scrabble every now and then, but I must say it’s much more enjoyable on a smartphone than in real life. ūüôā

To be honest, it’s not official Scrabble that I’m playing, but Wordfeud; however, for all practical purposes it’s the same.

In particular I like to play simultaneously in several languages. I’m struggling a bit with Norwegian (because it doesn’t do Nynorsk, which is the variety I know), but I’ve played very enjoyable games in Swedish, Dutch and Spanish, apart from the obvious Danish and English.

PS: If you’ve got an HTC ChaCha, too, did you realise you can take a screenshot by holding down the red button and pressing Home?

A prolonged murder mystery dinner

We’ve just come back from a week in Keith (between Inverness and Aberdeen) with all of Phyllis’s family.

One of the nights we decided to play a murder mystery game for eight players called Death by Chocolate that we had bought in the Agatha Christie shop in Torquay (Devon) last summer but never got round to playing.

The instructions for the game claimed it would last 2¬Ĺ–3 hours (including dinner), so I decided to start it at 5.30, so that there was plenty of time to finish it before the kids’ bedtime at 9 o’clock.

That plan didn’t work at all. The first half hour worked fine, but then the kids joined us, and it became impossible to role-play, and one or two of us had to keep leaving the room for various reasons.

In the end, we stopped playing and didn’t resume till all the kids were asleep (apart from Amaia, who decided to join us instead) and we then played until bedtime.

The game itself was quite good fun, but I thought it was too scripted at times – rather than being given dialogues to perform, I’d rather have been given the facts, told to memorise them and then been allowed to improvise for the rest of the game.

De bedste grunde i Matador



Matador
Originally uploaded by Björn Söderqvist

Sammen med min gode ven Simon diskuterede jeg for otte √•r siden sandsynlighederne for at lande p√• de forskellige grunde i Matador. S√• vidt jeg husker, gjorde Simon mig opm√¶rksom p√•, at hvis man kan beregne sandsynligheden for at lande p√• felt y, n√•r man st√•r p√• felt x, har man en Markov-k√¶de, og det er simpel line√¶r algebra at regne ud, hvor ofte hvert felt bes√łges.

Jeg skrev fluks et program til at g√łre det for mig (det mest besv√¶rlige var at h√•ndtere de forskellige Pr√łv Lykken-kort, s√•som “Flyt tre felter tilbage”).

De rå sandsynligheder har jeg placeret i bunden af dette indlæg, da de ikke er synderligt ophisende.

Hvad man virkelig √łnsker at vide, er naturligvis, hvilke grunde, man skal k√łbe, hvis man f√•r muligheden.

Dette er ret vanskeligt at regne ud, da det naturligvis er op til de enkelte spillere, hvor hurtigt de bygger huse og hoteller.

Jeg har derfor lavet en graf, der for hver grund viser de relevante udgifter (pris for hhv. sk√łdet, f√łrste hus, og opgradering til hotel) og sandsynlige indt√¶gter (f√łrst indt√¶gter i begyndelsen, hvor man kun ejer enkelte grunde, dern√¶st indt√¶gter i mellemfasen, hvor man typisk har et enkelt hus, og til sidst indt√¶gter i slutfasen, hvor der er hoteller overalt). Klik p√• grafen for at se den i fuld st√łrrelse:

Det er ikke uden grund, at Rådhuspladsen er berygtet blandt Matador-spillere!

Men man kan jo ikke bygge på enkelte grunde uafhængigt at de andre i farvegruppen, så denne analyse er måske ikke så interessant endda.

Jeg har derfor lagt værdierne i ovenstående graf sammen for hver gruppe, og resultatet kan ses til venstre.

Lidt overraskende er de gr√• grunde den bedste investering, fulgt af de berygtede lilla. Interessant nok kan det slet ikke betale sig at investere i de bl√• grunde – de er simpelthen for dyre at k√łbe og bebygge i forhold til indt√¶gtsmulighederne.

Til slut er her som lovet sandsynlighederne for at lande på de forskellige felter i Matador:

Nr. Navn Sandsynlighed
1 Start 2.6%
2 R√łdovrevej 2.2%
3 Pr√łv Lykken 1.5%
4 Hvidovrevej 2.3%
5 Indkomstskat 2.3%
6 Helsing√łr-Helsingborg 3.0%
7 Roskildevej 2.3%
8 Pr√łv Lykken 1.6%
9 Valby Langgade 2.3%
10 Allégade 2.2%
11 Fængsel 6.0%
12 Fredriksberg Allé 2.7%
13 Tuborg 2.4%
14 B√ľlowsvej 2.4%
15 Gl Kongevej 2.6%
16 Mols-Linien 2.8%
17 Bernstorffsvej 2.8%
18 Pr√łv Lykken 2.0%
19 Hellerupvej 2.9%
20 Strandvejen 3.0%
21 Parkering 2.8%
22 Trianglen 2.8%
23 Pr√łv Lykken 1.9%
24 √ėsterbrogade 2.7%
25 Gr√łnningen 3.1%
26 Gedser-Rostock 3.0%
27 Bredgade 2.7%
28 Kgs Nytorv 2.7%
29 Coca-Cola 2.7%
30 √ėstergade 2.7%
31 De Fængsles 0.0%
32 Amagertorv 2.7%
33 Vimmelskaftet 2.7%
34 Pr√łv Lykken 1.8%
35 Nygade 2.5%
36 R√łdby-Puttgarden 2.6%
37 Pr√łv Lykken 1.6%
38 Frederiksberggade 2.2%
39 Statsskat 2.2%
40 Rådhuspladsen 2.7%

International Trivial Pursuit



17/365: Trivial Pursuit
Originally uploaded by sparetomato

If you’re wanting to spend some enjoyable hours with a group of compatriots, Trivial Pursuit can be a good option.

However, if you are not from the same culture, many questions – especially in the pink and orange categories – can become almost impossible to answer for some of the players.

It would therefore be nice to have a box of international questions (either truly global, or perhaps European) to enable people with multinational families and friends to play the game.

I was searching a bit, and there seems to be a so-called “Globetrotter edition“, in which the six categories are replaced with six regions: North America,
Latin America, Oceania, Africa, Europe and Asia.

That sounds like it would be playable by people from different backgrounds, but it changes the nature of the game a lot, and I’d still rather have an international version of “proper” Trivial Pursuit.

FarmVille is for 2-year-olds

Most people will by now have heard of FarmVille, which is costing society billions in lost productivity because so many people are growing soybeans and raising sheep as part of their Facebook routine.

However, most people think of it as a game for grown-ups.

Just for fun I made profiles for Léon and Anna, and somewhat to my surprise, they love it.

Of course I have to do most of the game controls, but they’re most definitely in charge of what they grow and where they put it. Anna is for instance insisting on only growing pink trees (see the picture).

So I’m starting to think it’s really a game for wee kids, which adults just happen often to love, too.

Theft as marketing

It looks like Sony have vandalised an important landmark in Amsterdam in order to promote a new game (hat-tip: Dina).

It probably seemed like a fun idea to Sony’s marketing department, but I do hope they will be prosecuted harshly, because it sets a very dangerous precedent if you can commit crimes without punishment if it’s for marketing purposes.

I mean, in the first instance it might only be the letters in Hollywood’s logo that get stolen, but what will the next step be?

Will you be allowed to hijack planes to promote a new plane game, or perhaps even blow up a building to promote a new version of Worms?