Logo

CS 188's Personal Meeting Room - Shared screen with speaker view
Mengying Ju
14:46
Are all the coming written assessments going be due 11:59pm?
Mengying Ju
15:18
Thanks
Joshua Wu
15:25
thanks
Zhe Zhao
16:58
Can anyone provide the link of ppt?Ty
Joneil Manansala
17:05
Slides link for today: https://drive.google.com/file/d/17J9EfT-dAZHh-Ow5MRMWM7MgRHh8h_R7/view?usp=sharing
Zackoric
23:54
couldn't the blue ghost have gone up and then down?
Sunay Poole
26:09
Is each child equal probabilities?
Zackoric
26:27
expectimax pruning?
Miran
26:31
24/3?
Yewen Zhou
26:34
8?
Chenyue Cai
26:35
8
Ferdie Taruc
26:35
8
Tyler Nunez
26:43
8
Mengying Ju
26:46
8
Zhe Zhao
26:47
8
Sunay Poole
26:53
8
Israel Rodriguez
26:58
8
Zackoric
34:53
A minimax nested in expectimax
Julius Tereck
35:02
expectimax
Zackoric
39:22
So pacman moves first?
Yewen Zhou
39:59
why expectimax is considered optimistic?
Vishal Raman
40:14
It’s the average case vs the worst case
Yewen Zhou
40:31
Yeah, but why is average case considered "optimistic"?
Zackoric
41:23
Because the ghost could be mastermind
Zackoric
41:51
We are being optimistic assuming the ghost is random
Yewen Zhou
42:17
hmm makes sense
Zackoric
42:38
or are you thinking that optimistic means the ghost actively avoids us?
jimwang
43:05
Why isn’t expectimax pacman vs adversarial ghost deterministic?
david
43:59
So expectimax, in this example, assumes the ghost has equal probability to move in any possible direction? ( ie 1/4 N 1/4 W 1/4 S 1/4 E )
allen
44:00
why that same value node won't happen on random ghost
Vishal Raman
44:00
You can have two equally good options in minimax though
Mengying Ju
44:28
Why does the minimax pacman VS adversarial ghost gets lower average score than the minimax pacman VS random ghost?
david
44:32
So expectimax does not necessarily know what the ghost is “running”?
Zackoric
45:13
wdym @david
Yewen Zhou
45:35
wait, how can minimax pacman make decisions without knowing probabilities?
Sunay Poole
45:43
I think pacman never knows what the ghost is running @david, but assumes whether the ghost is adversarial or not based on whether paceman is optimistic or not
david
45:48
We said earlier that we could have minimax within expectimax, and act accordingly
Zackoric
45:51
minimax assumes worst case @yewen
david
45:58
I see thank you @sunay
Yewen Zhou
46:20
I see
Yewen Zhou
46:24
thanks
david
46:41
Would it be the most optimal then, even if it is the least efficient, to assume the opponent is "smart"
Zackoric
47:13
Not if you know for sure the opponent is not optimal
Julius Tereck
49:49
When was this?
Vishal Raman
50:15
https://en.wikipedia.org/wiki/TD-Gammon
Sunay Poole
53:13
lol
Miran
01:01:50
is that like a dutch book argument?
Zackroic’s phone
01:01:50
so we can think of preference as a value?
Yewen Zhou
01:02:16
how to understand continuity intuitively?
Julius Tereck
01:08:06
So it's kind of like the Scoville scale?
Zackoric
01:09:11
why not just U of EMV?
Calvin Wong
01:10:50
EMV changes based onthe lottery you define
Calvin Wong
01:10:59
lottery being L in this case
Julius Tereck
01:11:02
I don't even think we can say it's monotonic tbh. Winning the lottery seems like it usually improves peoples lives
Julius Tereck
01:11:14
*doesn't
Calvin Wong
01:12:56
sick economics in a cs class
Zackoric
01:13:12
So insurance companies make money by collecting the expected value margins?
Yewen Zhou
01:14:33
what's the purpose of this example?
Zackoric
01:15:04
given you only play the lotto a few times?
Calvin Wong
01:15:44
"humans are not risk neutral"
Zackoric
01:16:17
but if you play it all day every like Geico, they would be equal?
Calvin Wong
01:16:40
law or large numbers yeehaw yep
Calvin Wong
01:16:54
of*
Yewen Zhou
01:17:26
in this slide why isn't it U(L) < EMV(L) instead of U(L) < U(EMV(L))?
Yewen Zhou
01:17:48
got it
Yewen Zhou
01:18:00
thanks
Tyler Nunez
01:18:21
how is a lottery different from a probability distribution
Calvin Wong
01:18:30
LOL
Zackoric
01:18:56
lottery has monatery value or utility associated with each probability
Estea
01:19:20
A
Ferdie Taruc
01:19:42
risk seeking lets go s/o the 17%
Calvin Wong
01:19:46
I think lottery is a specific version of a distribution
Zackoric
01:19:51
I am really risk averse
Calvin Wong
01:19:55
as zackoric said
Manav
01:20:29
We could have all picked a and divyed up the proceeds 3.2 k each
Estea
01:20:40
D
Calvin Wong
01:20:41
lol just make an insurance company
Zackoric
01:20:43
lol, you need really solid coordination
david
01:21:07
Go big or go home
Abayjeet Singh
01:21:30
lol
Zackoric
01:21:40
not worth the 5% risk
Danny Sallurday
01:23:04
lol
Daisong Pan
01:23:11
lol
Zackoric
01:23:34
I think humans collaboratively are rational
Ferdie Taruc
01:23:41
different utility functions based on probabilities
Calvin Wong
01:24:12
ehhh debatable zackoric
Calvin Wong
01:24:27
salem witch hunts
Chudi Nnorukam
01:24:28
too real
Zackoric
01:24:37
lol
Julius Tereck
01:24:39
So in the backgammon example, the randomness meant that a shallow search was good enough. Or rather, a deeper search was pointless. So could the preference for certainty be a strategy to let you improve the quality of' planning?
Danny Sallurday
01:24:45
it also depends on current wealth though, right? would a rational agent consider that?
Zackoric
01:24:50
woo hoo!