Secular Technological Tailwinds

Trends vs. Predictions

1

“ A prediction or forecast is a statement about the way

things will happen in the future (…) ” Wikipedia

“Trend: To extend in a general direction : follow a general course ”

Webster Dictionary

http://en.wikipedia.org/wiki/Forecasting

http://en.wikipedia.org/wiki/Future

2

Predictions

3

Trends

4

• Concerned about learning theprocess, the interactions and theemerging laws.

Trends Predictions

• Concerned about being right.

5

Yes, we may use trends to make predictions.

Steve Ballmer, former chief executive of Microsoft, on the iPhone shortly

after Steve Jobs announced it. Ballmer went on to promote Microsoft's cheaper phones,

saying "right now we're selling millions and millions and

millions of phones a year. Apple is selling zero."

6

More information doesn’t imply more accuracy.

There is no free lunch.


7



“Trend: To extend in a general direction : follow a general course ”

Webster Dictionary




8



“Trend: To extend in a general direction : follow a general course

” Webster Dictionary




9



“Trend: To extend in a general direction : follow a general course

” Webster Dictionary

This is what this presentation is about.



Secular Technological Tailwinds(or trends!)

Dionisio Chiuratto AgourakisFounder / CEO – J!Quant

11

Dionisio [email protected]

https://plus.google.com/+DionisioChiuratto/

Facebook: Dionisio Chiuratto

Twitter: josaum

YO!: DIONISIO

12

Business (BBA)FGV-EAESP

13

MSc Comp. Eng.ITA

14

FounderJ!Quant

15

Cloud

Computing

16

Cloud for what?!

Why should we let things go from our premises?

17

Cloud for what?!

IT Overall Hardware Costs are Decreasing...

but change in IT is very expensive.

And it takes lots of time.

18

Cloud for what?!

So came outsourcing.

19

Cloud for what?!


Outsourcing was the first attempt to have IT-as-a-service.Whenever there was change, the contractor adapted (or should have).

“If IT isn’t our core business, why have it in-house?”

So the servers, databases, developers and support team were gone.

20

Cloud for what?!


Outsourcing was the first attempt to have IT-as-a-service.Whenever there was change, the contractor adapted (or should have).

“If IT isn’t our core business, why have it in-house?”

So the servers, databases, developers and support team were gone.

And so there was a bug...... and the company was unable to invoice customers for a week.

21

Cloud for what?!

There was (and there is) a need for reliability in IT, but with tight budgets and control over the process.

There was (and there is) a need for quickly scaling storage, processing, networking and licensing.

22

Cloud for what?!

I don’t want someone to take care of my ERP. I have the right guys.

I want someone to take the burden of hosting the infrastructure for my ERP.

I don’t want someone to take care of my DB. I have my DBA.

I want someone to take the burden of hosting the infrastructure for my DB.

And updating them.And taking care of their maintanance, power shutdowns, backups...

23

Cloud for what?!

The cloud concept can be understood as a evolution from the IT outsourcing desires.

Outsourcing

Infrastructure-as-a-Service (IaaS)

Platform-as-a-Service (PaaS)

Software-as-a-Service (SaaS)

Monitoring-as-a-Service (MaaS)

Communication-as-a-Service (CaaS)

AWS (EC2, RDS, ...)

AWS – Elastic Beamstalk

Office 365

Datadog

Skype

24

Got it. Now why so much buzz about it?

Conceptually, both for users and enterprises, the cloud allowed a detachment from the physical IT.

Before After

My computer with:my hardware, my apps, mymusic, my documents, mydata, my contacts, my e-mails and my games.

There is this place called“Cloud” which holdseverything, no matter thedevice I’m using to accessit.

25


We just need a minute (or twenty) to getthrough the evolution of the...

Ok! Hold on about cloud computing!

26

Internet1.0

Read-Only WEB

The Shopping Cart

Company Page

Your own (html-coded) Personal Home Page!

27

Internet 1.0

28

Internet 1.0The internet of pages!

Site 1 Site 2LINK

29

Internet2.0

Read/Write WEB

Blogs

Social Networks

30

Internet 2.0

31

Internet 2.0Social Internet / Collaborative

Open-sourceLike

ShareRepositories

Crowd-sourcingCrowd-funding

32

Internet 2.0PROSUMER

It became easy for anyone not only consume, but also produce content over the internet.

33


Everybody started producing contents, usingPaaS and SaaS.

Social media emerged.

Ok! We’re back with the cloud!

34


http://articles.economictimes.indiatimes.com/2012-05-27/news/31860969_1_instagram-largest-online-retailer-users

35


The cloud enabled us all to be creators in the internet.

The cloud enabled the mobile smart-devices widespread.

The cloud enabled companies to process and store HUGE amounts ofdata (more of that soon in the big data section)

The cloud enabled us all to connect with each other and share.

Without buying more hardware or dealing with any IT-specificproblems.

RESPECT THE BUZZ!

36

Semantic Web

37

Internet 2.0 – 3.0

38

It’s all about context!

The cloud got us covered with the infrastructure, platform andsoftwares. Now other questions arise:

“Who’s in the picture?”Twitter Trending Topics

#hashtagApple Siri

Microsoft CortanaGoogle Now

39

It’s all about context!

People have been producing and interactingwith contents over the internet. We’veevolved so much!

...

Meh.

40

The problem. Yes there is a problem.

The internet was made for people…Limited Time

Limited KnowledgeLimited Accuracy

Ambiguous Concepts…

41

How do we search for and fetch information?

42

How do we search for and fetch information?

How do we make a potato salad?

43

44

45

46

47

Internet < 3.0

A person browses, searches, filters and reports the results.

48

Semantic WebLinked DataOntologies

Internet > 3.0

49

Internet > 3.0A thing browses, searches, filters and reports the results.

50

Internet > 3.0URL = http://www.receitas.com.br/Palmirinha/Batata.html

Page with Recipes

Why only define addresses for pages?Pages are for people…

51

Internet > 3.0URI

A universal address for…anything.

52

Internet > 3.0

URI http://www.productontology.org/id/Potato_salad

URI http://purl.org/goodrelations/v1#ProductOrService

URI http://purl.org/goodrelations/v1#closes

http://www.productontology.org/id/Potato_salad

http://purl.org/goodrelations/v1#ProductOrService

http://purl.org/goodrelations/v1#closes

53

54

Internet > 3.0The Semantic Web has the purpose of conecting and relating data.

The Semantic Web is based on LINKED DATA.

The Semantic Web is MACHINE READABLE.

55

RDF triples?

Sorocaba , belongs to , SP stateSubject Predicate Object

SP state, has climate , Subtropical

RDF – Resource Description Framework

Subject Predicate Object

56

SPARQL

SELECT ?cityWHERE {

?city relationship:belongs ?state?state relationship:hasClimate “Subtropical”

}

List all cities in the state of SP that has the Subtropical climate.

How to make queries in a knowledge graph.

57http://commons.wikimedia.org/wiki/File:Rdf_graph_for_Eric_Miller.png

58

Wikipedia

59

Ontologies

Explicit representation of knowledge –definitions and relationships

Formal 1st order logic

XML-based

Object-Oriented(?!)

OWL – Ontology Web Language

http://www.inf.unibz.it/~franconi/dl/course/slides/kbs/kbs-modelling.pdf

60

61

62

63

Internet of Things

64

Ok, so far we’ve seen that:

• Cloud computing gave us a smart(cheap?) andscalable way of using platforms, softwares andinfrastructure;

• We are producing A LOT of content through;

• Semantic web is giving context, meaning andrelationship to data.

• Semantic web makes data machine-readable.

65

Now...The machines! (and other things)

As the cloud concept is interesting because wedetached from physical devices...

...the semantic web is interesting because it willdetach us from the graphical user interfaces. (andfrom browsing the internet)

66


• Hardware costs are falling. (and are more powerfull)

• M2M (machine to machine) communication is growing fast

• Advanced Software

• Cloud Services

http://www.microsoft.com/en-us/server-cloud/internet-of-things.aspx

67


http://www.opensource.com

68

Dev

ices

, se

nso

rs, c

ame

ras,

....

KnowledgeBase

Processing Results

Dev

ices

, se

nso

rs, c

ame

ras,

....

Cloud Services (IaaS, PaaS, SaaS)

Internet...of things

69

A lot of things connected throughthe internet, talking to each otherand living within the world aroundthem, creates data...a lot of data...

70

Big Data

71http://www.binarylaw.co.uk/2010/11/08/the-hype-cycle/

The hype cycle! (or the buzz evolution chart)

Take my money!!!

Meh. Doens’t work.

Interesting. Shouldanalyse the pros andcons!

72

What is Big Data?

• Companies have been producing and storing data since the ERP’s, CRM’s, BI’s, WMS’s, etc. adoption waves.

• Users have been producing content, giving away personal informations, and (lettingbig cloud tech companies) storing it since the PROSUMER feature of the web >= 2.0.

• With the widespread of IoT, billions of new devices are starting to produce and storedata.

73

What is Big Data?

That’s Big Data.

74

What is Big Data?

That’s Big Data. Really.

75

What is Big Data?

That’s Big Data. Really.

Just tons and tons of data that are now being created and stored.

76

So...?

Big Data by itself is nearly useless.There were important developments on processing and storing peta/hexa/yota-scale

data.But as for business and individuals, there were very few outcomes.

Nice solutions were deployed: visualization of twitter hash-tags in real time and geo-located. Truck-tracking, health-care experiments on collecting patient’s data, and so on.

77

So...?

But the killer application for Big Data is yet to come.

And other buzz-words (Analytics? Deep learning?) might take the merit for themselves.

78

So...?

And the reason is that...

79

Big Data won’t do anything without “Big Algorithms”.

80

Insights do not emerge by themselves.They need algorithms for Optimization, AI, Data Mining, Graphs, ...

And these requires processing power.

81

Parallelism

82http://5gnews.org/critique-pure-speed/

83http://5gnews.org/critique-pure-speed/

3-4 GHz

84

Where is the 100GHz processor?

CPU’s have been advancing in speed since they were born.

There are a number of factors resulting in the overall speed of a CPU, but by far themost straightforward is the clock speed (the MHz, GHz !)

85


Now it is easy to aknowledge that companies such as Intel or AMD are no longerincreasig the clock speed.

A barrier was hit around 3-4GHz. Why is that?

86


HEAT!

87


We can’t afford any more frequency increases with silicon boards.

88

But...computing power continues to expand, right?

Yes. There are two major avenues of computing power growth.

New Materialsand Hardwares

Parallelism

89

New Materials andHardwares

Parallelism

Post-SI Computing

• Graphene• Silicene• Quantum Computing

materials (nanotechnology)

Coarse-Grain Parallelism

• Distributed Computing• Hadoop

Fine-Grain Parallelism

• Multi-Core Processing• Many-Core Processing




• Multi-Core Processing• Many-Core Processing

90

New Materials andHardwares

Post-SI Computing

• Graphene• Silicene• Quantum Computing

materials (nanotechnology)

Under Research! In Stock!

Parallelism

91

Parallelism



Master

Slave Slave ... Slave

TaskTask

Task

MAP

Master

REDUCEPartial Ans.Partial Ans.

Partial Ans.

92

Parallelism



How many letters in this sentence?

Slave Slave Slave

Counts: 7 Counts: 9 Counts: 12

Master

Sum Reduce: 7 + 9 + 12 = 28

93

Parallelism


• Multi-Core Processing

http://www.techpowerup.com/reviews/Intel/Core_i7-5960X_5930K_5820K_Comparison/2.html

94

Parallelism


• Many-Core Processing


Massively Parallel Processing Paradigm

95

Parallelism





96

Parallelism





97

Paralellism is the way to go on speeding up applications and data processing.

All major tech companies are using parallelism (GPU’s, Xeon Phi’s, Hadoop) toanalyse, process and store data.

Parallelism is the way to go with Big Data.

98

Paralellism is the way to go on speeding up applications and data processing.

All major tech companies are using parallelism (GPU’s, Xeon Phi’s, Hadoop) toanalyse, process and store data.

Parallelism is the way to go with Big Data.

http://www.networkworld.com/article/2167576/tech-primers/hadoop---gpu--boost-performance-of-your-big-data-project-by-50x-200x-.html

99

But even with lots of data and the correct processingapproach through parallelism,

one question remains:

100

Where does insights comes from?

101

Algorithms

102

An algorithm is a set of instructions to be performed.

To be honest, all codes are algorithms by definition.

103

There are, however, non-trivial sets of instructions (i.e. algorithms) that aims tosolve specific problems.

104

What is the shortest path between two nodes in a directed graph?

105


Graphs 101:

Vertices (nodes) – Circles

Edges – Lines connecting them

If the lines have arrows – Directed Graph (Digraph)

If the lines doesn’t have arrows – Undirected Graph

The numbers over the edges indicates the weight of each one.

106


107


http://www.googlemaps.com

108


Ok! There’s an app for that!

109


Ok! There’s an app for that! algorithm for that!

110

Dijkstra Algorithm (1956)

http://en.wikipedia.org/wiki/Dijkstra%27s_algorithm

111

Dijkstra Algorithm (1956)

Shortest path is considered solved by the academia.

The computational time complexity is

This means that if you run the algorithm in your computer with say, 5 vertices, and it takes 10s to run.

When you run it with 10 vertices (2x), it is guaranteed that in the worst case it will take 40s (4x) to run.

112

Nice, but not every problem is under our control.

113

What is the shortest path in a directed graph, leaving from node A, passingthrough all vertices and returning to A? (TSP = Travelling Salesman Problem)

A A

http://mathworld.wolfram.com/TravelingSalesmanProblem.html

114

This is called NP-Complete. No one knows the time complexity of the problem, and there is no known analytical solution to it.

A A

http://mathworld.wolfram.com/TravelingSalesmanProblem.html

115

And there are the so-called “Learning Algorithms”

116

How do algorithms learn anyway?

117

How do algorithms learn anyway?

2 steps:

118

Step 1: Trainning

119

Number of Ed. Institutions in the city Ipads sold (x100)

1 1

2 4

3 4

… …

10 20

Trying to learn the law relating Ipad sales with educational institutions

120

121

Here, learn is the process in which we seek the best red-dotted line.

Because once we find it, we’ll know a mathematical formula relating the two variables.

Best can mean anything you like. In most cases, we are looking for the best fit, i.e. the line that minimizes the error across the training set.

122

Looking at past data, applying some black-box “learning algorithm” and inferring a mathematical relationship between variables.

The trainning phase is the most time and resource-consuming part of the process.

123

Step 2: Processing

124

Once we’ve learned the relationship, let’s say it is:

Ipads sold = 2*ed.Inst

If now we want to predict the number of ipads sold in a city with 20 educational institutions, we only need to do a few operations.

125

Running already-trained algorithms is lightweight (compared to trainning).

126

That’s how Siri (and Cortana, and Google Now) talks to you.

They’ve been trained for long hours on powerful hardwares, and they now can rely on the smartphone hardware to execute the algorithm and “understand” what you meant with “I am hungry”.

127

Trainning

• Heavyweight• Time and resource-

consuming• Seeks minimize

errors and maximize generality

• Take place in huge processing clouds

Processing (predicting, classifying, …)

• Lightweight (relative)• Real-time (or almost)• Only apply the pre-

determined operations• Might take place even

in mobile devices

128

Trainning

• Heavyweight• Time and resource-

consuming• Seeks minimize

errors and maximize generality

• Take place in huge processing clouds

Processing (predicting, classifying, …)

• Lightweight (relative)• Real-time (or almost)• Only apply the pre-

determined operations• Might take place even

in mobile devices

Both processes can be severely sped-up with parallel computing.

129http://www.gputechconf.com/

130

(really) nice

evidences

Drones, 3D-Printers, Self-Driving Cars, 4.0 Manufacturing, ...

131

Drones (UAV)

http://nypost.com/2015/04/12/war-against-isis-shows-limits-of-drones/

132

3D-Printing

http://www.3dprinters.nl/wat-is-een-3d-printer/

133http://www.telegraph.co.uk/motoring/road-safety/10570935/Autonomous-cars-is-this-the-end-of-driving.html

Self-driving cars

134

The 4th Industrial Revolution

135

One last thing…

136

One only prediction:

137

138

Algorithms will be the new standard.

139


For people’s jobs.

140http://www.forbes.com/sites/roberthof/2015/01/31/now-even-artificial-intelligence-gurus-fret-that-ai-will-steal-our-jobs/

“The U.S. took 200 years to get from 98% to 2% farming employment,”

“Over that span of 200 years we could retrain the descendants of farmers.”

“With this technology today, that transformation might happen much faster,”

Self-driving cars, he suggested could quickly put 5 million truck drivers out of work.

Andrew Ng

141



For companies' competition.

142

143



For companies' competition.

For countries’ sovereignty.

144

145

That’s why...

146

50 billionsNew devices will be connected by 2020

http://share.cisco.com/internet-of-things.html

147

14 trillionDollars added to the global economy by 2030

http://www.accenture.com/SiteCollectionDocuments/PDF/Accenture-Industrial-Internet-of-Things-Positioning-Paper-Report-2015.PDF

148

get ready!

Thanks!

149

Dionisio Chiuratto [email protected]

mailto:[email protected]

Secular Technological Tailwinds

Technology

general direction

way things

cloud concept

future wikipediatrend

trends predictions

steve ballmer

support team

core business