How to outsmart the Prisoner’s Dilemma - Lucas Husted

4,647,583 views ・ 2020-08-27

TED-Ed


请双击下面的英文字幕来播放视频。

翻译人员: Joyce He 校对人员: Yolanda Zhang
00:07
Two perfectly rational gingerbread men, Crispy and Chewy,
0
7218
4469
两个绝对理性的姜饼人, 嚼嚼和脆脆,
00:11
are out strolling when they’re caught by a fox.
1
11687
3340
在外面闲逛的时候 被一只狐狸抓住了。
00:15
Seeing how happy they are, he decides that,
2
15027
2586
狐狸看到它们这么开心,
就决定比起简单地把它们吃掉,
00:17
instead of simply eating them,
3
17613
2000
00:19
he’ll put their friendship to the test with a cruel dilemma.
4
19613
4087
它要通过一个残酷的窘境 来测试它们之间的友情。
00:23
He’ll ask each gingerbread man whether he’d opt to Spare or Sacrifice the other.
5
23700
6590
它分别问每只姜饼人是 选择解救还是牺牲对方。
00:30
They can discuss,
6
30290
1430
它们可以进行讨论,
00:31
but neither will know what the other chose until their decisions are locked in.
7
31720
5290
但直到它们的决定被确认后 才会知晓对方的选择。
00:37
If both choose to spare the other, the fox will eat just one of each of their limbs;
8
37010
6681
如果它们都选择解救对方, 那么狐狸只会吃掉每人一肢;
00:43
if one chooses to spare while the other sacrifices,
9
43691
4000
如果只有一个人选择解救对方,
00:47
the sparer will be fully eaten,
10
47691
2140
那么选择解救对方的人会被吃掉,
00:49
while the traitor will run away with all his limbs intact.
11
49831
4224
而叛徒却可以完好无损地跑掉。
00:54
Finally, if both choose to sacrifice, the fox will eat 3 limbs from each.
12
54055
7260
最后,如果它们都选择牺牲对方, 那么狐狸会吃掉每人三肢。
01:01
In game theory, this scenario is called the “Prisoner's Dilemma.”
13
61315
4717
在博弈论里,这个情况 被称之为“囚徒困境”。
01:06
To figure out how these gingerbread men will act in their perfect rationality,
14
66032
4780
为了搞清楚这些姜饼人 在绝对理性的情况下会怎么选择,
01:10
we can map the outcomes of each decision.
15
70812
3370
我们可以把每种情况的结果写出来。
01:14
The rows represent Crispy’s choices, and the columns are Chewy’s.
16
74182
4792
每一行代表的是脆脆的选择, 每一列代表的是嚼嚼的选择。
01:18
Meanwhile, the numbers in each cell
17
78974
2320
同时,每个单元格中的数字
01:21
represent the outcomes of their decisions,
18
81294
2460
代表的是每种选择所对应的结果,
01:23
as measured in the number of limbs each would keep:
19
83754
3459
通过每人残留的肢体数量来表示:
01:27
So do we expect their friendship to last the game?
20
87213
4379
你觉得它们的友情在游戏 结束后还能完好无损吗?
01:31
First, let’s consider Chewy’s options.
21
91592
2730
首先, 让我们来考虑嚼嚼的选项。
01:34
If Crispy spares him, Chewy can run away scot-free by sacrificing Crispy.
22
94322
5175
如果脆脆选择解救他,那么嚼嚼 就可以通过牺牲脆脆来逃脱惩罚。
01:39
But if Crispy sacrifices him,
23
99497
2030
但如果脆脆选择牺牲他,
01:41
Chewy can keep one of his limbs if he also sacrifices Crispy.
24
101527
4773
那么嚼嚼可以通过同时 牺牲脆脆来保留自己的一肢。
01:46
No matter what Crispy decides,
25
106300
2360
不管脆脆如何选择,
01:48
Chewy always experiences the best outcome by choosing to sacrifice his companion.
26
108660
6254
嚼嚼选择牺牲它的同伴 总能达到最优的结果。
01:54
The same is true for Crispy.
27
114914
2480
这一结论对脆脆来说也成立。
01:57
This is the standard conclusion of the Prisoner's Dilemma:
28
117394
3340
这就是囚徒困境的标准结论:
02:00
the two characters will betray one another.
29
120734
2620
两人都会选择出卖对方。
02:03
Their strategy to unconditionally sacrifice their companion
30
123354
4240
它们选择无条件牺牲对方的策略
02:07
is what game theorists call the “Nash Equilibrium,"
31
127594
4198
被博弈理论家称为 “纳什平衡”,
02:11
meaning that neither can gain by deviating from it.
32
131792
3948
意思是任何一方只要 背离这一策略都会有所损失。
02:15
Crispy and Chewy act accordingly
33
135740
2390
脆脆和嚼嚼遵照 这一理论做出决定
02:18
and the smug fox runs off with a belly full of gingerbread,
34
138130
4000
让沾沾自喜的狐狸 得以吃了一肚子的姜饼,
02:22
leaving the two former friends with just one leg to stand on.
35
142130
4434
而两位昔日好友都只剩下 一肢在支撑着它们的身体。
02:26
Normally, this is where the story would end,
36
146564
2770
通常情况下, 故事到这里就结束了。
02:29
but a wizard happened to be watching the whole mess unfold.
37
149334
3850
但有一个巫师 恰巧见证了这一切。
02:33
He tells Crispy and Chewy that, as punishment for betraying each other,
38
153184
4789
他告诉脆脆和嚼嚼, 作为背弃彼此的惩罚,
02:37
they’re doomed to repeat this dilemma for the rest of their lives,
39
157973
4131
它们余生都将注定 要一直重复这一窘境,
02:42
starting with all four limbs at each sunrise.
40
162104
4492
每天日出的时候 都将重新获得四肢。
02:46
Now what happens?
41
166596
1810
现在该如何是好?
02:48
This is called an Infinite Prisoner’s Dilemma, and it’s a literal game changer.
42
168406
5699
这被称为“无限囚徒困境”, 它颠覆了之前的结论。
02:54
That’s because the gingerbread men can now use their future decisions
43
174105
4481
这是因为姜饼人 可以用未来的决定
02:58
as bargaining chips for the present ones.
44
178586
3220
作为现在讨价还价的筹码。
03:01
Consider this strategy: both agree to spare each other every day.
45
181806
4694
让我们考虑下这个策略: 每人每天都同意互相解救对方。
03:06
If one ever chooses to sacrifice,
46
186500
2920
但凡有任何一个人选择牺牲对方,
03:09
the other will retaliate by choosing “sacrifice” for the rest of eternity.
47
189420
4219
那么另一人就可以通过余生 一直选择牺牲它来进行报复。
03:13
So is that enough to get these poor sentient baked goods
48
193639
4000
这样就足够让这些 可怜的,有意识的焙烤食品
03:17
to agree to cooperate?
49
197639
2120
同意合作了吗?
03:19
To figure that out, we have to factor in another consideration:
50
199759
4675
为了弄清楚,我们得 将另一因素考虑进来:
03:24
the gingerbread men probably care about the future
51
204434
3170
比起将来,这些姜饼人
03:27
less than they care about the present.
52
207604
2830
应该会更重视现在。
03:30
In other words, they might discount
53
210434
2480
换言之,它们可能会将
03:32
how much they care about their future limbs by some number,
54
212914
4000
自己所在乎的未来的 肢体数量换算成一个数字,
03:36
which we’ll call delta.
55
216914
2410
我们将其称为 δ 。
03:39
This is similar to the idea of inflation eroding the value of money.
56
219324
4818
这个点子类似于通货膨胀 会降低金钱的价值。
03:44
If delta is one half,
57
224142
1900
如果 δ 是 1/2 ,
03:46
on day one they care about day 2 limbs half as much as day 1 limbs,
58
226042
5486
那么第二天的每两个肢体对它们来说 都相当于是第一天的一个肢体,
03:51
day 3 limbs 1 quarter as much as day 1 limbs, and so on.
59
231528
4832
第三天的肢体是第一天肢体 价值的四分之一,以此类推。
03:56
A delta of 0 means that they don’t care about their future limbs at all,
60
236360
5021
δ 等于 0 则意味着它们根本 不在乎未来的肢体数量,
04:01
so they’ll repeat their initial choice of mutual sacrifice endlessly.
61
241381
4958
所以它们将会无止境地 重复最初的选择:互相牺牲。
04:06
But as delta approaches 1, they’ll do anything possible
62
246339
4264
但当 δ 趋近 1, 它们将会尽己所能地
04:10
to avoid the pain of infinite triple limb consumption,
63
250603
4000
避免自己每天无止境地 失去三肢的痛苦,
04:14
which means they’ll choose to spare each other.
64
254603
2820
于是他们会选择互相解救。
04:17
At some point in between they could go either way.
65
257423
3360
当 δ 取这两个值之间的某个点时, 任何一种选择都有可能发生。
04:20
We can find out where that point is
66
260783
2120
我们可以通过写出 代表每种策略的无穷级数,
04:22
by writing the infinite series that represents each strategy,
67
262903
4235
来找到那个点的位置,
04:27
setting them equal to each other, and solving for delta.
68
267138
4000
设它们的数值相等,来求解 δ 。
04:31
That yields 1/3, meaning that as long as Crispy and Chewy care about tomorrow
69
271138
5666
结果是 1/3 ,说明只要脆脆和嚼嚼 认为明天的重要性
04:36
at least 1/3 as much as today,
70
276804
3110
至少占今天的 1/3 ,
04:39
it’s optimal for them to spare and cooperate forever.
71
279914
4363
那么合作:互相解救 是对他们最有利的。
04:44
This analysis isn’t unique to cookies and wizards;
72
284277
3750
这个分析并不只 适用于饼干和巫师这则故事,
04:48
we see it play out in real-life situations
73
288027
2820
在现实生活中也经常出现于
04:50
like trade negotiations and international politics.
74
290847
3790
像是贸易谈判和国际政治的形势下。
04:54
Rational leaders must assume that the decisions they make today
75
294637
4210
理性的领导者必须假定 它们每天所做的决定
04:58
will impact those of their adversaries tomorrow.
76
298847
3240
会影响他们竞争对手明天的决定。
05:02
Selfishness may win out in the short-term, but with the proper incentives,
77
302087
4573
自私自利也许在短期内能带来利润, 但只要有恰当的激励措施,
05:06
peaceful cooperation is not only possible, but demonstrably and mathematically ideal.
78
306660
6583
和平的合作不只是可能的, 而且也被数学推导证实是更理想的。
05:13
As for the gingerbread men, their eternity may be pretty crumby,
79
313243
4000
对于姜饼人来说,它们 无穷无尽的故事看起来很糟糕,
05:17
but so long as they go out on a limb,
80
317243
2300
但只要它们肯为对方担风险,
05:19
their friendship will never again be half-baked.
81
319543
2999
那么它们的友谊就能地久天长。
关于本网站

这个网站将向你介绍对学习英语有用的YouTube视频。你将看到来自世界各地的一流教师教授的英语课程。双击每个视频页面上显示的英文字幕,即可从那里播放视频。字幕会随着视频的播放而同步滚动。如果你有任何意见或要求,请使用此联系表与我们联系。

https://forms.gle/WvT1wiN1qDtmnspy7