Artificial Intelligence in Digital Innovation Hubs

Artificial Intelligence in Digital Innovation Hubs

Victor NegrescuVice-Rector of the Romanian National University of Political Studies and Public AdministrationFormer MEP

AI DIH NetworkGiovanna GalassoDirector - PwC EU [email protected]

Objectives of the AI DIH NetworkDevelop a structured cooperation approach to create a European Network of Digital Innovation Hubs

with focus on AI

Launch of the call for EoI

September – December 2018

Final selection of 30 Hubs with focus

on AI

January – February 2019

2-days KO meeting of the programme

March 2019

Implementation of the programme (webinars,

collaborative workshops and mentoring activities)

April – November 2019

Signature of a Framework Cooperation

Agreement between DIHs

November 2019

Development of policy recommendations for

collaboration and networking

March 2020

Where do we stand…

Help the DIHs to unlock their collaboration and networking potential via mentoring, coaching and co-

creating activities, mainly dedicated to collaboration

Signature of a co-drafted multi-lateral framework cooperation agreement between (at least) 10 Digital

Innovation Hubs

Develop, based on evidence resulting from the demonstration activities, a blueprint for cross-border

collaboration as well as supporting measures and policy recommendations for enhancing

collaboration and networking potential

Overview of the Programme

Programme introduction/ Kick-off workshop

TP2

TP3

TP4

TP5

TP6

TP7

TP8

AS IS analysis –Mapping DIH business models and services

Cooperation strategies for the provision of AI services

Costs and benefits of cooperation scenarios

Funding and financing of collaboration schemes

Legal aspects of collaboration

AI services and ecosystem – AI Digital Assessment

Towards future cooperation among DIHs

March 2019

April 2019

May 2019

June 2019

October 2019

October 2019

Jun - Nov 2019

November 2019

2participatory workshops

2participatory workshops

3 peer learning webinars

30 on-site visits

Collaborative platform

TP1

Webinar

Follow-up webinar

Follow-up webinar

Individual online mentoring sessions

Webinar

Selected DIHs - Geographical coverage

30 DIHs from 20 different

countries selected out of

150 applications

SmartIC

Latvian IT Cluster DIH

LTroboticsDIH

• FCAI

• Super IoT

SINTEF

nZEB Smart Home

• HPC4Poland

• PIAP HUB

Czech Institute of Informatics, Robotics

and Cybernetics

CeADAR

DTI

Smart Connected

Supplier Network

IMEC• DIGIHALL

• WEST

• Teralab

• AIR4S

• ITI Data Cycle Hub

• esHPC

CROBOHUB

4P DIH

PRODUTECH

• RIF BioRobotics Institute

• IP4VG

• DIH Lombardia

Know-Center GmbH

• DFKI

• VDTC (Fraunhofer IFF)

• Munich Innovation Hub for

Applied AI

Global AI TrendsMassimo PellegrinoNew Ventures Leader and Global Ethical AI Leader - PwC [email protected]

AI definition

“Artificial intelligence (AI) systems are software (and possibly alsohardware) systems designed by humans that, given a complexgoal, act in the physical or digital dimension by perceiving theirenvironment through data acquisition, interpreting the collectedstructured or unstructured data, reasoning on the knowledge, orprocessing the information, derived from this data and decidingthe best action(s) to take to achieve the given goal. AI systems caneither use symbolic rules or learn a number.”

Source: A definition of Artificial Intelligence: main capabilities and disciplines, AI HLEG - European Commission

Impact of AI on global GDP

Source: PwC - What’s the real value of AI for your business and how can you capitalize?

Impact of AI on GDP per regionSome economies have the potential to gain more than others in both absolute and relative terms. Though all economies should benefit, North America and China are expected to witness the greatest GDP gains from AI increasing productivity.

Source: PwC - What’s the real value of AI for your business and how can you capitalize?

Impact of AI on GDP per sector

He

alth

care

Au

tom

oti

ve

Fin

anci

al

serv

ice

s

Tran

spo

rtat

ion

an

d lo

gist

ics

Source: PwC’s AI Impact Index

Impact of AI on GDP per sectorTe

chn

olo

gy,

com

mu

nic

atio

n a

nd

e

nte

rtai

nm

ent

Re

tail

Ene

rgy

Man

ufa

ctu

rin

gSource: PwC’s AI Impact Index

AI patent trends

Three main categories of patents can be identified, which are: • AI techniques• Functional applications• Application fieldsOver 68% of AI patents fall into more than one category.

Historical analysis of the trends in AI patent applications and in scientific publications shows that between 2012 and 2017:• AI patent families grew on average by 28%

annually• Scientific publications grew on average by

5.6% annually

Source: WIPO Technology Trends 2019


AI patent trends

Machine learning represents 89% of patent families related to

an AI technique

Overall, the share of scientific publications is higher

than patent families for AI techniques


AI patent trends

Patent families and scientific publications related to machine

learning sub-categories as a share of the total for AI


AI patent trends


In terms of functional applications, computer vision represents 49% of all AI patentfamilies, while speech processing and natural language processing account for 14% and13%, respectively.

Main players in AI


Among the top 30 AI applicantsworldwide, there are:• 26 are companies,

of which 12 located in Japan and 6 in United States

• 4 are universities or research institutions, located in China and in the Republic of Korea

U.S.

U.S.

Japan

Republic of Korea

Japan

Japan

Japan

Japan

Japan

U.S.

Europe

Japan

Japan

Japan

China

Japan

China

Japan

Republic of Korea

Republic of Korea

Europe

Japan

U.S.

Europe

Europe

China

U.S.

U.S.

China

China

Main players in AI

In terms of patents applications, CAS (China) and ETRI (Republic of Korea) rank 1st and 2nd respectively. In Europe, top applicants are Fraunhofer (Germany) and CEA (France).


Ethics guidelines for trustworthy AI

1

2

3

4

Human agency and oversightIncluding fundamental rights, human agency and human oversight

Technical robustness and safetyIncluding resilience to attack and security, fall back plan and general safety, accuracy, reliability and reproducibility

Privacy and data governanceIncluding respect for privacy, quality and integrity of data, and access to data

TransparencyIncluding traceability, explainability and communication

5

6

7

Diversity, non-discrimination and fairnessIncluding the avoidance of unfair bias, accessibility and universal design, and stakeholder participation

Societal and environmental wellbeingIncluding sustainability and environmental friendliness, social impact, society and democracy

AccountabilityIncluding auditability, minimization and reporting of negative impact, trade-offs and redress

7 Requirements for trustworthy AI

Source: AI HLEG – European Commission

Innovation Services in Robotics and AI to SMEs Using the RIFs (Robotics Innovation Facilities) Concept

Paolo Dario - DIH RIF BioRobotics InstituteAI DIH Network

1998

2013

2018

2019

The DIH RIF BioRobotics Institute participates in the AI DIH Network Project, and creates a strong connection with the Industry 4.0 Competence Center ARTES 4.0, that signs the Framework Cooperation Agreement

Towards the creation of an AI DIH in Tuscany

2017

LINK Project (MIUR – Italian Ministry of University and Research) Funding: 26 M€Projects evaluated: 320Projects granted: 50

ECHORD++ Project 2013-2019

Funding: 20M€ Establishment of RIF (Robotics

Innovation Facility) Peccioli Pisa: an innovation driver to link discovery and

delivery, and to bridge the Valley of Death between innovation and market.

ARTES 4.0 Competence Center funded by the Italian Ministry of Economic Development (MiSE)Funding: 10.66M€ + 18M€Network of 127 partners (Universities, research centers, industries, foundations, etc..) in the areas of: Advanced Robotics and Enabling Digital Technologies (including AI)

TERRINet Project 2017-2021Funding: 5M€ Establishing a European Robotics Research Infrastructures Network

2012-2017

The lesson learned regarding the distances that SMEs are prepared to travel appears somewhat ‘Gaussian’ - most interactions are local within 50km with SME interaction rapidly dropping off with increasing distance. Prof. Chris Melhuish, Bristol RIF Coordinator, Echord++ Project

In 2015 TURF was invited to the launch of the RIF Peccioli Pisa under the ECHORD++ Project

Subsequently, it seemed natural to contact the RIF Peccioli Pisa when TURF conceived the idea of a portable, automatic and battery-driven weather station

In 2016, through the interaction with RIF Peccioli Pisa, TURF came into contact with researchers at Scuola Superiore Sant’Anna who provided TURF with: technical solutions development, specifically the development of software and electronics; with proof of

concept services tailored at assessing the technical feasibility; and with support to streamline the selection of AI components suppliers

The first sales of their product were in 2017 and TURF now works with customers as football clubs (Premiere Leagues) in several countries, including Spain, Italy and England

TURF Europe and the collaboration with the RIF Peccioli Pisa, today the DIH RIF BioRobotics Institute, under the ECHORD++ Project

TURF Europe is a spin-off from Pisa University, qualifying as a small consultancy and R&D firm specializing in building and maintaining turfed areas of all types. TURF focuses on precision farming, which is a business growth area in Europe and which is of particular

value for farming communities in Tuscany.

Green-GO systemBringing precision farming to urban green areasFilippo Lulli, Alessio Forconi - TURF Europe

“Green-GO”: bringing precision farming to urban green areas

“GREEN GO”: bringing PA to urban green areas – DEI Madrid (ES) – 14th November 2019

The starting point

• A niche of agronomy: sports turf (stadia, golfs, etc.)• Difficult and varying microclimates• Very exacting customers• Clients scattered across Europe• Difficult to monitor microclimate• “Precision farming” is the future…

IDEA: DEVELOP OUR OWN PORTABLE WEATHER STATION

WHERE TO START ? ?WE ARE AGRONOMISTS…

Outcome 2018

WHAT INFO DOES IT GIVE YOU ?

ParametersSoil MEASURED: Temperature, volumetric water content, EC

CALCULATED: available water content

Air Temperature, humidity, pressure, vapour pressure

Wind Speed, direction.

Light MEASURED: photosynthetically active radiation (μmol/m2/s)CALCULATED: Daily Light Integral (mol/m2/d)

Canopy Evapotranspiration (mm/d)


Outcome

What’s the situation on my pitch ?


Outcome

What’s my historical data ?


2019-2020: keep pushing forward

GREEN-i


1. RGB HD + multispectral + Thermal IR2. Neural networks (pest diagnosis)3. Georeferenced grid (bot guidance)

2019-2020: keep pushing forward

GREEN-i


Ultralocalized automated low-dose treatments. Increase sustainability

AI Applied in Quality ControlDIGIHALL, Paris Region DIH, FranceValentina Ivanova, PhD

Service: test before you buy

• A customer is looking to explore the added value of the artificial intelligence to perform quality control in production line

• Customer Requirements:• Low cost < 10k€

• Flexibility

• Reliability

• Performance

• Testbed used: FFLOR (Future Factory @ LORraine)

Project principles• Objective: to realise an automated quality control demonstrator for

similar product type on our FFLOR test bed

• Proposed solution: Deep Learning technique• Approach – using open source Deep Learning Platforms (TensorFlow + Keras)

permitting to be used from non-specialists • Using neural network for image analysis, advantages:

• Flexibility

• Robustness

• Excellent performance

• Installation of an Embedded camera in the collaborative robot hand

• Collecting pictures from the inspected part from all angles

• Image analysis (classification)

From concept to demonstrator

How small companies use AI to maintain and improve their expertise:

A real example from the wooden furniture manufacturing sector, Florian Mohr

Our AI DIH

Berlin / Brandenburg

Rhineland-PalatinateKaiserslautern

North Rhine-Westphalia Dortmund

Lower SaxonyHannover

BavariaAugsburgBaden-Württemberg

Stuttgart

ThuringiaIlmenau

SaxonyChemnitz

Hamburg

Rhineland-PalatinateCoblenz

Saxony-AnhaltMagdeburg

HesseDarmstadt

SaarlandSaarbruck

The company and its use case

Carpentry Kasper

• staircases, tables and individual furniture made of solid wood

• 30 years of expertise in wood selection and handling

• two old carpenters masters combine most of the knowledge of the company

The pain of the small company

How to transfer experiential knowledge and skills of a handcraft process?

?

DFKI & MKZ

AI

no AI

Scope of an AI project

Capturing the manufacturing Process

• creation of a process map• creation of different process diagram

AI =

Artificial Intelligence

Advanced Informaticsor

Using AI for knowledge transfer

• recording of videos of the different manufacturing steps

• transcription of the (speech-to-text)

• translation into optional languages

• tagging of keywords

• creation of a database of keywords

→ quick and easy to find contents

In-Line Diabetic Retinopathy Detection

Munich Innovation Hub for Applied AIHolger Pfeifer (fortiss) – Fintan Buckley (Ubotica)

Munich DIH for Applied AI

–hosted at fortiss

Centre for AI

Munich Innovation Hub for Applied AI

AI-related services for SMEs Jointly operated by:

In-Line DR Detection

• Diabetic Retinopathy (DR) is the leading cause of vision loss in adults

• Cost of DR in UK in 2010/2011 was estimated at £57,741,842

• National screening programs in place in Ireland and other geographies


• Pilot Program – PC used to host AI based analysis

Intel Movidius VPUbased breadboard(H2020 Eyes Of Things)

WiFi


• Pilot Program – PC used for AI based analysis


• Production Level – integrate into the camera

• Glaucoma• Hypertension• Alzheimer’s• …..

Intel Movidius VPU


• Verifiable AI is critical for AI adoption in biomedical applications• “why did the system classify a retina as DR/non-DR”

• Neural Network Dependability Kit (NNDK) from fortiss• Uses formal methods to validate AI decisions


• Partnership with fortiss through the Munich Digital Innovation Hub• Introduced the Neural Network Dependability Kit (NNDK)

• Provided on-site training with the NNDK

• Supported integration of the NNDK into the solution• Bug fixes

• Feature enhancements

• Interpretation of results


• Global fundus camera market estimated to be 80,000 per annum by 2024

• Ubotica achieving a 30% market share by 2024• is expected to yield €10M/annum in revenue

• will create 10 to 12 full time positions to service the market

• Positive impact in other market segments

• Enabling early detection of DR will have a positive impact for European citizens and a corresponding reduction in Health Care costs

AI-driven media monitoring Expanding senses

Adam Olszewski Poznan Supercomputing and Networking CenterHPC4Poland DIH

• Predictive maintenance

• Deep neural networks

• Sentiment analysis

• Image recognition & digital immersion

• Real-time big data analysis

• Process mining and machine learning for

sequence optimization in production lines

Fully

Operational

DIH

DIH name

• Problem consulting

• Testing methodology

• Methodology road

mapping

• Analytical modelling

• Data selection and

acquisition

• Automation of maps

creation

• Algorithm

optimization

• Technology selection

• Experimenting

• Proof of Concept

V I S I O N A N D M I S S I O N

W e i n n o v a t e t h e P o l i s h m a n u f a c t u r i n g i n d u s t r y b y

r a i s i n g t h e a v a i l a b i l i t y o f a d v a n c e d I T s e r v i c e s ! ! !

D I H C O U N T R Y

K E Y A I E X P E R T I S E K E Y A I S E R V I C E S

Poznan

Poland

C O N TA C T D E TA I L S

Adam Olszewski

[email protected]

DIH official website

AI service on the market – PSMM use case:

brand monitoringadvanced impact analysissponsorship efficiencysentiment analysiscompetitor watchingcustomer satisfaction…

Our AI customer website – psmm.pl

AI Service: automated media stream analysis…turning traditional media analysis into one man’s job!

Problems addressed:Low speed, narrow scope, low quality of analysesSignificant workforce needed for each analysisBiased results depending on analysts’ beliefs

AI Service: automated media stream analysis

Searching for predefined words, expressions, imagesSimultaneous analysis of numerous TV and radio channels (live and historical)Social media analysisMarket specific media analysisAudio into text transcriptionContextual sentiment analysisSpeaker’s emotions analysis

AI in action – psmm use case:A. Image recognitionB. Text acquisitionC. Speech recognition

A.B.

C.

Logo recognition

Text recognition

Speech into text - live

AI service – how we do it• Speech recognition and analysis (own engine):

Semantic analysis of spoken content

Determining the emotional state of speaker based on acoustic markers

Classification of speakers (gender, age, other)

• Image recognition (neural network):

Identification of logotype appearance (number, size, time)

Contextual assessment of impact

• Text recognition and analysis (own analytical models, OS OCRs):

Identification of text

Text recognition

Semantic analysis of text

• Sentiment analysis embracing all above, e.g. to assess the efficiency of sponsorship campaigns.

AI Service application areas - examples

• Manufacturing firms – to verify the impact of promo campaigns, to assess customer satisfaction, to monitor strategic moves and brand

• Marketing agencies, e.g. Press Service Media Monitoring

• Consulting firms – strategy building, market impact assessment

• Radio, television, e.g. TVP (the Polish national TV channels) –migrating from manual to automatic analyses, subtitling, etc.

From idea to featureGeneration of a data-driven AI use case for the SW industry

Dr. Robert Ginthör, Know-Center, Austria

* COMET is an Austrian funding program that supports a sustainable knowledge transfer from science to industry.

Know-Center GmbH

• Austria's leading research center for Data-driven Business and Big Data Analytics

• Established in 2001• Located at Technical University Graz• > 110 Staff• > 600 COMET* & industrial & international

EU projects

Reval

• SW solutions for treasury and risk management

• 130 staff (Reval Austria)• Member of ION group (6500 employees)

AI Challenge

Reval wanted to enhance its SW product with AI-based functionalities, but initially lacked concrete and well-founded ideas

• Problem!• Start!

• Skills?• Organization?• Tools, technology?

• Skills?• Organization?• Tools, technology?

AI Service – Data Value Check

SUPERMARKET CHECKOUT OPTIMIZATION

DESCRIPTION: In order to improve customer experience in supermarkets, the checkout process needs to be optimized. Since self-

checkout is not accepted by all customer segments, the classic payment process should be improved. To that end, the aim is to shorten

overall queueing times during checkout by predicting when additional cash registers should be opened to efficiently handle the workload.

DATA INVENTORY

VALUE: Increase customer experience by shortening wait times in queues. Improve employee satisfaction by identifying potential

bottlenecks in advance and act accordingly. Develop a database of data tracked via shopping carts and baskets as a basis for future use

cases.

INPUT: Checkout location, current locations of shopping carts and baskets, number of open checkouts, maximum checkout capacity.

FUNCTIONAL APPROACH: 1.) Check current queue lengths at checkouts. 2.) Locate approaching shopping carts and baskets

3.) Predict ETA at checkouts by comparing current locations with previously recorded paths and arrival times at checkouts. 4.) Inform shop

employees about the upcoming checkout workload, potential bottlenecks and best course of action.

OUTPUT: Estimated upcoming checkout workload, bottleneck indicator, recommended action.

VALUE SCORES EFFORT SCORES

VALUE IMPACT AREAS VALUE IMPACT SCORE(0-100: low impact – high impact)

DATA SUITABILITY SCORE(0-100: low suitability – high suitability)

PROTOTYPING EFFORT SCORE(0-100: high effort to low effort)

SCORING SUMMARY RECOMMENDATIONS

IMPACT-EFFORT MATRIX STATUS PROGRESS FLOW Although this procedure is primarily designed to increase the efficiency of checkout

processes in supermarkets, further improvements could be achieved based on the

implementation of the described use case, such as:

- optimization of the number of available cash registers (i.e., there may be too many)

- optimization of shelf arrangement (i.e., smart shelf arrangement may encourage

customers to purchase more items)

- workload estimate based on prior days (i.e., the existing workforce can be used more

efficiently)

- benchmarking supermarket performance

© Know-Center GmbH, www.know-center.at

90

70 30

RANK 2 RANK 1 RANK 51

6

6

7

9

22

24

35

Improve cooperation& partnerships

Strengthen brandreputation

Increase providedinformation quality

Improve performance& competitiveness

High Innovation level

Increase employeesatisfaction

Broaden & strengthencustomer base

Improve customersatisfaction

Evaluate Impact

Evaluate implementation effort

Evaluate data suitability

Top 5

X

X

X

0

50

100

0 50 100

EFFORT SCORE

IMP

AC

T S

CO

RE

Results

• 3 use cases finally selected for development

• 1 use case already implemented as new feature for Reval‘s SW solution

• First time use of Deep Learning at Reval

3 – 8 weeks

Potential schemes for cooperation among DIHsGiovanna GalassoDirector - PwC EU [email protected]

Our journey towards cooperationAdopting a customer-

centric mindset

• Identification of

ecosystem needs

• Characterisation of the

personas

• Definition of the

customer journeys Cooperation strategies

• Review of the

customers’ journeys


cooperation

opportunities

• Co-development of

the cooperation

scenarios

Analysis of the co-created

scenarios

• Review of the scenarios and

analysis of the processes

involved

Costs and benefits

of cooperation

• Analysis of costs

benefits and risks

of cooperation

scenarios


solutions to

tackle

collaboration

hurdles

Onsite visits

One-to-one meetings to

assess the feasibility of

the scenarios for each

DIH

Magdeburg & Prague

Valencia

&

Brussels

The final scenarios and horizontal tools

Partnership to provide services jointly

Matchmaking

Development of a new service

To be able to respond to a client’s request leveraging on the

capabilities and infrastructure available in the network

To widen the DIH offer to the ecosystem by cooperating with

other DIHs in Europe facing similar challenges and needs

To provide new opportunities to the ecosystem, creating

connections with players operating in other regions

ObjectiveWillingness to cooperate

Link with the DEP

Exporting / Importing EDIH excellence

Connecting ecosystem

Cooperation scenarios

Horizontal tools To facilitate the process of connecting for DIHs

To ease networking and matchmaking

To enhance knowledge sharing

• Platform to share information on services/ skills/ AI testing facilities

• Regular face to face to face meetings/ working groups/ speed dates among DIHs

• AI-empowered digital platform/ mechanisms to facilitate matchmaking activities

• Platform/Repository to share information on data sources and recommended providers

• Platform/Repository to share use cases, good practices and lessons learnt

Development of a new service (1/2)

This scenario is used in case a DIH decides to enlarge its offerings by developing a new service to respond to theecosystem needs and wishes to leverage skills and capabilities available within the network to design the servicetogether

1 Identification of a partner for the service delivery

2 Assess and discuss the cooperation opportunity

DIH1, intending to include a new

service in its service offerings,

looks for partners with

complementary or similar

competencies willing to take part in

the development of the service

Once interested DIHs respond to

DIH1 invitation, DIHs discuss

together the potential cooperation

opportunity and the features of

the services that should be

developed

The three DIHs co-create and sign a

cooperation agreement, reflecting

the strategy and covering related

legal and financial issues (e.g. IPR

of the service and contents

developed)

The three DIHs implement the

strategy and activities agreed.

Once the new service is developed,

each DIH starts to providing it to

its ecosystem

3 Co-define and sign cooperation agreement

4 Implement the strategy

DIH 1 DIH 1

DIH 2

DIH 3

DIH 1

DIH 2

DIH 3

DIH 1

Ecosyst.

DIH 2

Ecosyst.

DIH 3

Ecosyst.Cooperation opportunity

Strategy & coop. agreementDIH Network

Development of a new service (2/2)

Possible solutions

A deeper knowledge of the EU markets and ecosystems, enabling to identify common needs and challenges more easily

This would tackle the risk to spend resources in co-developing a service that is not in line with ecosystem needs and requests for both DIHs

Possible solutions

The introduction of standard legal templates for governing the development of new services, such as

• an agreement that sets the rules concerning the liabilities of each hub involved

• an agreement regulating IP issues, on the basis of the relationship between the hubs involved

Next steps for the future and..

Facilitate DIHs reciprocal knowledge by mapping their specific technical expertise and competencies and supporting their networking activity

Develop standard legal templates (e.g. Service Level Agreement) reducing bureaucracy and time required for establishing cooperation and mitigating perceived risk of cooperation

Provide funds to cover costs of cross-border cooperation and incentivise it, until it becomes a well-established and regular mechanism

Test the structured approach to cooperation in H2020/ Horizon Europe/ DEP and in other technological fields

…our first concrete step in collaboration

The Framework Cooperation Agreement is designed to foster

collaboration among DIHs.

By endorsing it, the parties commit to cooperate, in the spirit of good

faith, so to:

• foster their cross-border collaboration by implementing co-designed

cooperation schemes,

• while, boosting the development of technological innovation in the

EU.

AI DIH Network

Date post:	27-Jan-2022
Category:	Documents
Upload:	others
View:	4 times
Download:	0 times

Artificial Intelligence in Digital Innovation Hubs

Documents