Engage 2017 - Making the most of machine intelligence - Nell Watson - Open Eth & Quantacorp

© 2017 Singularity University

Making the Most of Machine Intelligence

December 7th, 2017

Andrej Karpathy

Fei-Fei Li, Stanford

1996 2017

Year Breakthrough in AI Datasets (first available) Algorithms (first proposed)

1994 Human-level spontaneous speech recognition

Spoken Wall Street Journal articles and other texts (1991)

Hidden Markov Model (1984

1997 IBM Deep Blue defeats Garry Kasparov

700,000 Grandmaster chess games, a.k.a. the Extended Book (1991)

Negascout planning algorithm (1983)

2005 Google’s Arabic and Chinese-to-English translations

1.8 trillion tokens from Google Web and news pages (2005)

Statistical machine translation algorithm (1988)

2011 IBM Watson becomes the world Jeopardy Champion

8.6m docs: Wikipedia, Wikiquote, Project Gutenberg (2010)

Mixture-of-Experts Algorithm (1991)

2014 Google’s GoogLeNet Object Classification at Human-level

ImageNet corpus of 1.5m labelled images,1k object categories (2010)

Convolutional Neural Networks algorithm (1989)

2015 Google’s Deepmind learns 29 Atari games by observation

Arcade Learning Environment dataset of 50+ Atari games (2013) Q-learning algorithm (1992)

Average No. of years to breakthrough: 3 Years 18 Years

Year Breakthrough in AI Datasets (first available) Algorithms (first proposed)

1994 Human-level spontaneous speech recognition

Spoken Wall Street Journal articles and other texts (1991)

Hidden Markov Model (1984

1997 IBM Deep Blue defeats Garry Kasparov

700,000 Grandmaster chess games, a.k.a. the Extended Book (1991)

Negascout planning algorithm (1983)

2005 Google’s Arabic and Chinese-to-English translations

1.8 trillion tokens from Google Web and news pages (2005)

Statistical machine translation algorithm (1988)

2011 IBM Watson becomes the world Jeopardy Champion

8.6m docs: Wikipedia, Wikiquote, Project Gutenberg (2010)

Mixture-of-Experts Algorithm (1991)

2014 Google’s GoogLeNet Object Classification at Human-level

ImageNet corpus of 1.5m labelled images,1k object categories (2010)

Convolutional Neural Networks algorithm (1989)

2015 Google’s Deepmind learns 29 Atari games by observation

Arcade Learning Environment dataset of 50+ Atari games (2013) Q-learning algorithm (1992)

Average No. of years to breakthrough: 3 Years 18 Years

HUMAN HISTORY UP TO YEAR 2000

DOUBLED EVERY 10 HOURS BY 2010

DOUBLES EVERY 2 MINUTES IN 2017

5 EXABYTES (5 Bn GB)

1 sec.

Intuitive Intelligence is now a Utility

C19th C 20th C 21st

21NEURAL STORYTELLER

24

pix2pix

FRA LUCA PACIOLI

DIFFUSED KNOWLEDGE ON BUSINESS PERFORMANCE

Distributed Risk

Illustration:MarkMontgomery

Trust+

Shared Epistemology,Decentralised

TOP 500

x 1000+SUPERCOMPUTERS

•Fungibility

•Property

•Reward•Utility

Reverse ICO

s o l i d i t y

SMART CONTRACTS

Technology Best Use-Case

BITCOIN STORE OF VALUE

ETHEREUM SMART CONTRACTS

IOTA SCALABILITY, IOT

RIPPLE TRANSACTIONS OF MONEY

HYPERLEDGER CORPORATE-FRIENDLY BLOCKCHAIN INFRASTRUCTURE

FILECOIN DISTRIBUTED STORAGE

GNOSIS PREDICTION MARKETS

IMPAK NOTARY, PROPERTY RIGHTS

DUBAI

Machine Payable Web

APPEAL TO BOTS MAY HAVE MORE LEVERAGE THAN APPEAL TO HUMANS

Web 1.0 Web 2.0 Web 3.0

Nodes Documents People / Brands Economic Agents

Edges Hyperlinks Likes, Follows, Friends Tokens

PrimaryActivity Linking Liking, Following

Buying, Selling, Holding

PrimaryRelationships

Semantic ‘Shared Keywords’

Social ‘Friends in Common’

Economic ‘Common Interests’

Primary Content

Discovery

Keyword-based search

Social-based feed ?DominantPlatform Google Facebook ?

TECHNOLOGY MAKES SCARCE THINGS

ABUNDANT

Sometimes I watch you sleep.

Imperva Incapsula Bot Traffic Report

Imperva Incapsula Bot Traffic Report

Fei-Fei Li

Non-fiction is for facts, fiction is for values.

NELL@NELLWATSON.COMNell Watson, AI & Robotics Faculty

Disclaimer:Any opinions expressed herein are the speaker’s, and not necessarily those of SU.

@NELLWATSON