AI+
AI
即將從棋盤,走入我們的真實世界
AI + 社會科學的開始
New Wave of AI Application Example – Digital Ads
AI+ “Misbehaving”
Q & A
1. New Wave of
AI 應用例 — 數位廣告
你的點擊,
反映你的選擇決策
Digital Advertising Revenues Hit $19.6 B in Q1 2017, Climbing 23% Year-Over-Year
Internet Ads > TV 全球網路廣告花費已大於電視
Winner Takes All 網路廣告贏者通吃
Google & Facebook Ads Examples
搓合與優化配置
Internet as a mass media
網路廣告優化方程式
七大族群的精準行銷?
十五大族群的精準行銷?
Common Data Categories
大數據分析找到更多潛在客群
Advertiser Utility: The Value Funnel
Range
Ads Optimization Formula
Data Science
The Revolution of Big Data
Models Cases
Models Cases
Optimization Perspective
Gradient Descent
“New” Wave of Machine Learning
“Deep” Learning AI
(Big) Data-driven
More tolerance for “state-of-the-art” empirical evidence
Ensemble with Reinforcement-learning & other methods
World, Model & Theory
Model?!
Artificial Power Artificial Intelligence 體力 腦力 的第四次工業革命
2.
AI 與 “不當行為”
以人為中心
2017 年諾貝爾經濟學獎揭曉,行為經濟學出線
人是自私和理性的
“The Theory of Moral Sentiments” by Adam Smith
Every man is, no doubt, by nature first and principally recommended to his own care; and he is fitter to care of himself than of every other person…" (1759, 82)
(每個人天生都是為自己活著的,並且他比其他任何人都更有能力為自己精打細算)
市場是有效率的(尤其是金融市場)
正是由於每個人自私自利的天性,Adam Smith 提出的Invisible Hands(看不見的手)才可能最有效的發揮其作用,讓市場在供需影響下達到最有效的狀態。
Loss Aversion & Endowment Effect
「97% 贏得 100 美元」 vs. 「37% 贏得 300 美元」 ?
如果送你一個賭注,你會願意多少錢轉賣出去?
Loss Aversion & Endowment Effect
A: 這項治療法可治 200 人
B: 這項治療法有 1/3 的機會拯救 600 人, 2/3 的機會無人得救
Loss Aversion & Endowment Effect
「97% 贏得 100 美元」 vs. 「37% 贏得 300 美元」 ?
如果送你一個賭注,你會願意多少錢轉賣出去?
the Coase Theorem Works at Tokens 寇斯定理對於明確價值的交易可行
the Coase Theorem did not Work in Practice 寇斯定理在實務上不可行
「損失的痛苦」是「獲得的快樂」之 2 倍
The Behavioral Economics of BitCoin
The Behavioral Economics of Cryptokitties
「損失的痛苦」是「獲得的快樂」之 2 倍
Simon’s Bounded Rationality
人真的是理性/非理性的嗎?
Paul Krugman – Nobel Price(2008)
Home DNA Test
How Many Kinds of People in the World? 人有幾種?
Know-What, Know-Why, Know-How and Decision Making
Kinds of Human in the World? 人有幾種?
Ads Optimization <-> Economic Decision
AlphaGo, Master to Zero
AlphaZero
AlphaZero
AlphaGo/Master/Zero, AlphaZero
全宇宙原子數大約為 10^80
以全宇宙可見物質總質量(1.45×10^53) / 氫原子質量(1.67×10^−27)
圍棋的排列組合總數 10^171
AlphaGo 的運算能力
早期的 AlphaGo Fan 使用 176 個 GPU
AlphaGo Lee 使用了 48 個 TPU
AlphaGo Master 與 AlphaGo Zero 皆只使用 4 個 TPU。
Computation Economics
“New” Wave of Machine Learning
“Deep” Learning AI
(Big) Data-driven
More tolerance for “state-of-the-art” empirical
1. AI+
From The New Wave of
AI & Big Data
to See the “Misbehaving”
Big Data Consultant / Data Scientist
Craig Chao ( )
chaocraig@gmail.com
https://www.slideshare.net/chaocraig/
13. Internet as a mass media
“Half the money I spend on advertising is
wasted; the trouble is I don‘t know which half.”
-- John Wanamaker, ~ 1875a pioneer in marketing
…
18. Common Data Categories
▣ Persona
u Age, Gender, Birth date,
City, …
▣ Attributes
u Phone brand/model, location,
time, App, browser, banner…
▣ Behavior
u Click,
Conversion(Installation, Cart,
Purchase, …), Activation,
Payment…
22. Range
- Roger Martin
Rothman School of Management, Toronto
If only attach importance to quantify the business
model, it will not have the ability to find a potential
growth opportunities: The pursuit of quantifying
the biggest problem is that people ignore the
context of the behavior generated, detached from
the context of the event, and have not been
included in the model ignores variables
effectiveness.
29. The Revolution of Big Data
DATA
Hypotheses
Statistical Analysis
BIG DATA
Hypotheses
Machine Learning
Data Mining
Machine-generated
Sampling, Multi-variant… All, Hyper space, …
Volume, Velocity, Variety, Veracity
Human-explainable
30. Models ßà Cases
Russ Merz, An Integrated Model of Media Satisfaction and Engagement: Theory, Empirical
Assessment and Managerial Implications, Journal of Applied Marketing Theory, Nov 2011
31. Models ßà Cases
Russ Merz, An Integrated Model of Media Satisfaction and Engagement: Theory, Empirical
Assessment and Managerial Implications, Journal of Applied Marketing Theory, Nov 2011
BIG DATA
Hypotheses
Machine Learning
Data Mining
Machine-generated
All, Hyper space, …
Volume, Velocity, Variety, Veracity
deductive inductive
Cases
Models
Models
Cases
34. “New” Wave of Machine Learning
▣ “Deep” Learning AI
▣ (Big) Data-driven
▣ More tolerance for “state-of-the-art” empirical evidence
▣ Ensemble with Reinforcement-learning other methods
35. World, Model Theory
Credit: John F. Sowa
generalized statements,
proven scientifically with evidence
Simplified representation, helpful tool to
understand specific phenomena
40. ▣ “The Theory of Moral Sentiments” by Adam Smith
u Every man is, no doubt, by nature first and principally
recommended to his own care; and he is fitter to care of
himself than of every other person… (1759, 82)
▣
60. Know-What, Know-Why, Know-How
and Decision Making
Know-How
(Feasible?)
Prescriptive
Know-What
(Objective?)
Descriptive
Know-Why
(Scientific?)
Normative
Descriptive D-M:
How decisions are made?
Normative D-M:
How decisions should be made?
Prescriptive D-M:
How decisions could be made better?
RationalityBounded Rationality
H. Simon,
Administrative
Behavior AI
Src: JT Chiang, NTU MBA
68. AlphaGo
▣ AlphaGo Fan 176 GPU
▣ AlphaGo Lee 48 TPU
▣ AlphaGo Master AlphaGo Zero 4 TPU
69.
70.
71. Computation Economics
▣ Wikipedia
■ a research discipline at the interface of computer science,
economics, and management science.
■ encompasses computational modeling of economic systems,
whether agent-based, general-equilibrium, macroeconomic, or
rational-expectations, computational econometrics and statistics,
computational finance
■ Some of these areas are unique to computational economics,
while others extend traditional areas of economics by solving
problems that are difficult to study without the use of computers
and associated numerical methods
72. “New” Wave of Machine Learning
▣ “Deep” Learning AI
▣ (Big) Data-driven
▣ More tolerance for “state-of-the-art” empirical evidence
▣ Ensemble with Reinforcement-learning other methods
73. alchemy
▣ Ali Rahimi
u Google Scientist
u Machine learning Expert
u Test of Time Award, NIPS 2017
▣ NIPS
u The Conference and Workshop on
Neural Information Processing
Systems (NIPS)
▣ Yann LeCun
u Director of AI Research, Facebook
u The famous pioneer of DL expert
u The author of the famous LeeNet
https://keysduplicated.com/~ali/Research.html
LeeCun: people threw the baby with the bath water and focused on
provable convex methods or glorified template matching methods