AI+
Craig Chao
chaocraig@gmail.com
AI Introduction (02)
圖靈測試(Turing Machine)
Computational Equivalence
圖靈測試(Turing Test)
望梅止渴
Src: Fortune, 2016/09
Natural Brain Activities
Deep Neural Network
Src: Fortune, 2016/09
http://104.155.212.49:5000/
Src: Fortune, 2016/09
Training with 2 hidden layers
Training with 4 hidden layers
Src: Fortune, 2016/09
AI Timeline
Ada
(1842)
Alan
Turing
(1950)
The first
conference
on
AI by John
McCarthy,
Marvin
Minsky
(1956)
Demonstrated
by Newell
(1957)
Unimation
s working
on GE
(1961)
Joseph
Weizenba
um (1965),
E.
Geigenba
um (1965)
Chess-
playing
program
by
Greenblatt
at MIT
(1968)
Jack
Myers
Harry
Pople
(1979)
1980s Ian
Horswil
l
(1993)
TiVo
Suggestions
(2005)
Apple,
Google,
Micorsoft
(2011)
Machine
Learning,
Deep
Learning
(2013 ~)
KEY MOMENTS IN DEEP-
LEARNING HISTORY
KEY MOMENTS IN DEEP-
LEARNING HISTORY
擊碎玻璃天花板的華裔女科學家
http://deeplearning.net/2012/12/13/googles-large-scale-deep-learning-experiments/
KEY MOMENTS IN DEEP-
LEARNING HISTORY
Deep Dream
Deep Dream
Deep Dream
Deep Dream
Deep Dream
https://deepdreamgenerator.com/
Deep Dream
Deep Dream
Artistic Style
Artistic Style
Artistic Style
LipNet
LipNet
LipNet
Protect Communication
Protect Communication
Let’s Enhance!
Google RAISR
「Rapid and Accurate Image Super-Resolution」
Upsampling
(aliasing artifacts)
Google RAISR
Google RAISR
Google RAISR
Google Brain's super-
resolution technique
Plug & Play
Generative
Networks
Generative Adversarial Text to
Image Synthesis
Image-to-Image Translation with
Conditional Adversarial Nets
Pix2Pix
Pix2Pix
https://affinelayer.com/pixsrv/
https://ml4a.github.io/guides/Pix2Pix/
Ratbot and maze
Ratbot and maze
Bird-eye camera over the maze
Digital reward map updating
Superior learning performance of Ratbots
over unenhanced rats
Embed The Word
AI can Tell the Difference between Sports
CaptionBot
Microsoft Seeing AI
"I think it's a young girl throwing an
orange Frisbee in the park," Microsoft's
AI will tell you.
"I think it's a man jumping through the air doing a
trick on a skateboard," Microsoft's AI says.
AlphaGo 9–15 March 2016
AlphaGo 9–15 March 2016
OpenAI Gym
StarCraft II DeepMind feature
layer API
AI Coopetetion
Microsoft's AI Tay offends and goes
offline
Open Letter on Artificial
Intelligence 08/19/15
To Benefit People and Society
Microsoft + OpenAI
ASILOMAR AI PRINCIPLES
https://futureoflife.org/ai-principles/
ASILOMAR AI PRINCIPLES
https://futureoflife.org/ai-principles/
Auto-Pilot in China
Auto-Pilot in China
第一届中国智能车未来挑战赛(2009年6月4日—5日,西安浐灞生态区)
Auto-Pilot in China
Auto-Pilot in China
第六届中国智能车未来挑战赛(2014年11月15日—16日,常熟)
Auto-Pilot in China
第七届中国智能车未来挑战赛(2015年11月15日,常熟)
Auto-Pilot in China
Auto-Pilot in China
Auto-Pilot in China
Auto-Pilot in China
假人前方,需要車輛進行剎車,等待假人
「穿過」馬路,方可通過。
形象逼真的假人,走過人行橫道。同樣要求車輛對其進
行識別並且剎車。
Auto-Pilot in China
隧道內部光線微弱,考察車輛在暗環境下的應對策略。 隧道內設置了「白板」障礙物。對於白板的設計,考察
攝像頭和激光雷達的識別能力。
Auto-Pilot in China
Auto-Pilot in China
百度計劃於2018年進軍接送服務業,奇瑞EQ汽車車型小,無噪音,滿電後可運行數趟,因而為最佳選擇。
原文網址:https://read01.com/L4Moe6.html
Auto-Pilot in China
Auto-Pilot in China
chaocraig@gmail.com

Ai plus-ai intro 02-20170605

Editor's Notes

  • #73 1. AlphaGo v13 的網絡層數實際是不夠的。如果按照 AlphaGo v13 的架構,5x5往上面長11層3x3,相當於27x27,看上去夠大了吧?錯,這樣的半徑只有14。 因此,如果大龍的長或寬超出14? 2. 由於網絡的結構是往上一層層生長,如果隻長幾層,一般不會丟失重要信息,但如果一直長上去,就會越來越容易出現問題。所以,大龍甚至都不用長到14,電腦就已經不一定“知道”自己的大龍是一條聯通的大龍了。 按照AlphaGo v13 的架構,如果大龍只在一端有兩個真眼,另一端就甚至不一定知道自己已經活了(它只會知道自己有兩口氣,而這是網絡輸入告訴它的) 一個誤區:它會傾向於認為氣很多的棋塊就是活的。對於局部死活,這沒有問題,但對於大龍死活,這是不足夠的。 棋塊小的時候,可能出現的形狀不多,而且許多形狀經常在對局中出現,因此容易被神經網絡學會。而大龍越大,其可能出現的形狀就越多,網絡不一定能學會。