ARTIFICIAL INTELLIGENCE AT WORK
CRIT February 1st 2017
See the relatedVideo
onYouTube:
bit.ly/AI-Applications
UPDATE YOUR TECHNOLOGY
Some companies lose their business because they keep on doing
always the same thing and don’t realise that the world around
them has changed.
PRESENTER NOTES
Sometimes is a matter of moving from one technology
to another and re-inventing an old business.
Yellow Pages for example, has been replaced by online
search engines like Google
PRESENTER NOTES
BLOCKBUSTER failed to invest its capitals in the new
web-based business model.
Now it has been replaced by NETFLIX.
PRESENTER NOTES
Internet and the availability of cheap sensors and
processors determined the end of KODAK and the rise
of INSTAGRAM.
PRESENTER NOTES
And remember that the Italian Yellow Pages,
BLOCKBUSTER and Kodak were prosperous companies
just then years ago.
PRESENTER NOTES
UNDERSTAND YOUR CUSTOMERS
Some other times the technology hasn’t been changed much,
sometimes it’s a matter of adapting the product to the new
customer demand or to pay attention to the market needs and to
target new needs.
PRESENTER NOTES
Red Bull was the first to sell the “energy drink” concept:
here the shift wasn’t related to the product’s technology, it was more about intuition and the discovery of a new business opportunity.
Red Bull was the first to target this market and Coca Cola lagged behind.
Just after realising it they tried to recover with burn.
In a while we’ll see an example on how Deep Learning and Artificial Intelligence can be used to discover new patterns in data to target
new business opportunities.
PRESENTER NOTES
OPTIMISE PRODUCTION
Another application in which the Artificial Intelligence will have a huge impact is the optimisation of the manufacturing processes to achieve better quality,
lower costs.

The manufacturing systems are becoming more agile to adapt faster to the market requests.

The same production line should be made to produce smaller products batches.

The requirements are no more just about quality and speed but also on production agility.

Here the new data analysis systems should be designed to rapidly adapt to the varying production environment.
PRESENTER NOTES
Amazon for example has been working on big numbers from it’s birth to be competitive on prices.
PRESENTER NOTES
Now the advanced automation helps Amazon
in keeping low the managing costs of its
departments.
At the same time the service is much faster
and adaptable to the markets requests.
PRESENTER NOTES
THE 4TH INDUSTRIAL REVOLUTION
We are facing the 4th Industrial Revolution that will
strongly modify our society.

Most companies are adopting those technologies and
the first to successfully implement it will be the next
business leaders.
PRESENTER NOTES
The FIRST INDUSTRIAL REVOLUTION began when humans start to use FOSSIL ENERGY instead of MUSCLE POWER.
The SECOND was about mass production and line assembly. It was about the standardisation or repetitive manual tasks.
The THIRD was the first automation wave. It was in the 70’. Simple repetitive tasks were automated.
Now we see the coming of the 4th and the biggest revolution.
PRESENTER NOTES
This is happening now: the
companies listed in this slide
provide the fundamental
software components to create
advanced Artificial Intelligence
applications.
PRESENTER NOTES
And those other companies
integrate that libraries to
create vertical products on
specific markets.
Artificial Intelligence can
perform tasks that since few
years ago could be just
performed by humans.
The FIRST industrial revolution
replaced the MUSCLE POWER,
the coming FOURTH industrial
revolution will replace the
BRAIN POWER.
As soon as the major
companies will adopt these
technologies, there will be an
enormous impact on the job
market.
Repetitive tasks of
progressively higher
complexity will be automated
as the technology evolves.
PRESENTER NOTES
This slide shows some results
from a research made by the
University of Oxford.
In this case the telephone
salespersons and the typists
are two professions that will
be impacted by machines that
can understand the human
spoken language and relative
concepts.
This will allow to automate call
centres, customer care desks,
automatic translation.
PRESENTER NOTES
GoogleTranslate
This is not science fiction: you have a working example in your
pocket: just check your Google Translate or your Apple Siri.
PRESENTER NOTES
Other jobs to be automated
are those involving vision and
handling.
This is made possible by the
Artificial Intelligence that can
understand highly complex
scenes in cluttered
environments.
Lets see two demo video that
we made in Addfor.
In the first one we will see how
Artificial Intelligence recognise
and classify objects in a
complex situation.
In the second video, the
Artificial Intelligence describes
a picture in plain english. Just
try to image how can archives
of million of images can be
scanned and queried using
plain written questions.
PRESENTER NOTES
See the relatedVideo
onYouTube:
bit.ly/AI-Applications
IT IS NOT AN OPTION
Adopting the new Artificial Intelligence technologies wouldn’t be an option, it is the same as it was with the steam engine or the assembly line invented
by Ford: those who do not adopt this technology will simply be excluded from the market.

The development wouldn't stop: there is simply to much money involved.

NVIDIA was the best-performing stock in 2016 S&P 500 delivering a 225% total return, it increased its market capitalisation by 11 BILLION dollar just
in one day, in November 2016, the day of its Q3 earnings call.
PRESENTER NOTES
ALREADY HAVE SOME INTERNAL SKILL ?
ASK FOR A CUSTOMTRAINING
ASK FOR ANALYSIS AND PROOF OF CONCEPT
NO INTERNAL SKILLS ?
The adoption of a disruptive new technology can be risky and requires investments. You should take a safe path.

There are many available software solutions on the market but just some of them will fit your needs.

We are completely agnostic on hardware and software solutions, for this reason we can help our customers in making the best choices.

Maybe the best start could be to share some ideas with us.

Because the biggest risk for your company is just not being informed and not being ready.
PRESENTER NOTES
ARTIFICIAL INTELLIGENCE: SOME EXAMPLES
Now we will see some examples of Artificial intelligence algorithms
that are readily available to be deployed on vertical solutions.

This is just a partial list because those technologies can have a
broad range of applications, nevertheless we think that those
examples can be quite informative.
PRESENTER NOTES
UNDERSTAND THE MARKET
The first example is about finding Anomalies in the data streams. This class of algorithms can be applied to a wide range of problems
to find glitches in production systems, unusual behaviours in supply chains, financial systems, energy and communication networks.
Finding an Anomaly is important to trigger proper reactions, for example, it can be used for predictive maintenance or for discover
some Latent Business Opportunities in market data.
PRESENTER NOTES
EXAMPLE: FIND ANOMALIES INTAXI CALLS
In this case we analyse the NYC Taxi Calls Database: it contains
the number of taxi calls hour by hour in the year 2014.
PRESENTER NOTES
ALGORITHM FINDS ANOMALIES
BY ANALYSING RAW DATA
The algorithm has been supplied with only the raw data. No other data (like holidays or special events has been provided).
Here we see the algorithms that start to analyse the data and after a while understands that the data presents a normal seasonality and normal
behaviours at peak hours.
When the behaviour is different from what the algorithms predict the point is defined as an anomaly and is plotted in yellow or red depending of the
anomaly strength.
PRESENTER NOTES
FIRST ANOMALY: XMAS + NEWYEAR
The first strongest anomalies detected have an easy
explanation: on new year’s day there is a peak of taxi
calls just after midnight and on December 25th there
are few calls.
Remember that the algorithms do not have any
information about holidays, it sees Christmas just as a
standard day, for this reasons it detects an abnormal
pattern in the data.
PRESENTER NOTES
SECOND ANOMALY:
SUNDAY AFTER
HALLOWEEN
The second anomaly is detected for halloween and the night after
when the people get back home.
PRESENTER NOTES
THIRD ANOMALY:
The third anomaly was about a
strange low volume of calls the
night after the thanksgiving day. It
should be due to a bombing
thread spread from the news
PRESENTER NOTES
LAST ANOMALY:WHAT’S APP ON DEC 6 2014 ?
What’s strange is this anomaly detected on December the
6th: it’s difficult to spot by eyes but the system detected
here a strong glitch in data.
Seemingly nothing special happened in NYC that day.
PRESENTER NOTES
Just after looking back in the NYC news we found this event that
seemingly brought to NYC many unexpected visitors.

Summarising, Artificial Intelligence can watch data streams 24/7
finding irregular behaviours, faults, and useful information to
improve your systems and your businesses.
PRESENTER NOTES
OPTIMIZE PRODUCTION
The second example is about production optimisation. In particular we will see how
Artificial Intelligence can use the sensor’s data to predict system failures.
PRESENTER NOTES
EXAMPLE: ANALYSE MACHINERY DATA
The most common way to supervise machines is to create Dashboards for human operators. This approach has many limits: Time and
human attention for example. Moreover humans cannot supervise hundreds of data streams at a time.
PRESENTER NOTES
ARTIFICIAL INTELLIGENCE
CAN WATCH 100’S OFVARIABLES
AND FIND USEFUL INFORMATIONS
FAILUREORIGINAL
I this example our algorithm controls 100’s of variables together. Every single variable by itself is meaningless, nevertheless, we made
a system based on a Deep Neural Network that combine all the variables to find the information we need. In this case the left graph
shows the raw variables while the right graph shows the output our our system that predict a jet engine failure.
PRESENTER NOTES
ORIGINAL DATA
FAILURE MODES
Here the Deep Learning algorithm finds three specific failure modes indicated by the three red circles on the right graph
PRESENTER NOTES
OFFER NEW FEATURES
The last example is about using
Artificial Intelligence to create new
feature and set new market
standards.
In this image you can see a NEST
thermostat that set new standards
for home automation devices.
PRESENTER NOTES
EXAMPLE: ADVANCEDTRACTION CONTROLS
But today I would like to bring you an Italian example that explains
how Ferrari was the first carmaker to put an Artificial Intelligence
system in its production Electronic Control Units. Nowadays this
system is installed on every Ferrari model.

To explain the system in detail I could like to give the floor to Marco
Fainello
PRESENTER NOTES
See the relatedVideo
onYouTube:
bit.ly/AI-Applications
We don’t know exactly how Intelligent Machines will evolve
It will be an Exponential Growth
…
This mean that it will happen sooner than we expect
it.linkedin.com/in/ebusto
ThankYou

ARTIFICIAL INTELLIGENCE AT WORK

  • 1.
    ARTIFICIAL INTELLIGENCE ATWORK CRIT February 1st 2017
  • 2.
  • 3.
    UPDATE YOUR TECHNOLOGY Somecompanies lose their business because they keep on doing always the same thing and don’t realise that the world around them has changed. PRESENTER NOTES
  • 4.
    Sometimes is amatter of moving from one technology to another and re-inventing an old business. Yellow Pages for example, has been replaced by online search engines like Google PRESENTER NOTES
  • 5.
    BLOCKBUSTER failed toinvest its capitals in the new web-based business model. Now it has been replaced by NETFLIX. PRESENTER NOTES
  • 6.
    Internet and theavailability of cheap sensors and processors determined the end of KODAK and the rise of INSTAGRAM. PRESENTER NOTES
  • 7.
    And remember thatthe Italian Yellow Pages, BLOCKBUSTER and Kodak were prosperous companies just then years ago. PRESENTER NOTES
  • 8.
    UNDERSTAND YOUR CUSTOMERS Someother times the technology hasn’t been changed much, sometimes it’s a matter of adapting the product to the new customer demand or to pay attention to the market needs and to target new needs. PRESENTER NOTES
  • 9.
    Red Bull wasthe first to sell the “energy drink” concept: here the shift wasn’t related to the product’s technology, it was more about intuition and the discovery of a new business opportunity. Red Bull was the first to target this market and Coca Cola lagged behind. Just after realising it they tried to recover with burn. In a while we’ll see an example on how Deep Learning and Artificial Intelligence can be used to discover new patterns in data to target new business opportunities. PRESENTER NOTES
  • 10.
    OPTIMISE PRODUCTION Another applicationin which the Artificial Intelligence will have a huge impact is the optimisation of the manufacturing processes to achieve better quality, lower costs. The manufacturing systems are becoming more agile to adapt faster to the market requests. The same production line should be made to produce smaller products batches. The requirements are no more just about quality and speed but also on production agility. Here the new data analysis systems should be designed to rapidly adapt to the varying production environment. PRESENTER NOTES
  • 11.
    Amazon for examplehas been working on big numbers from it’s birth to be competitive on prices. PRESENTER NOTES
  • 12.
    Now the advancedautomation helps Amazon in keeping low the managing costs of its departments. At the same time the service is much faster and adaptable to the markets requests. PRESENTER NOTES
  • 13.
    THE 4TH INDUSTRIALREVOLUTION We are facing the 4th Industrial Revolution that will strongly modify our society. Most companies are adopting those technologies and the first to successfully implement it will be the next business leaders. PRESENTER NOTES
  • 14.
    The FIRST INDUSTRIALREVOLUTION began when humans start to use FOSSIL ENERGY instead of MUSCLE POWER. The SECOND was about mass production and line assembly. It was about the standardisation or repetitive manual tasks. The THIRD was the first automation wave. It was in the 70’. Simple repetitive tasks were automated. Now we see the coming of the 4th and the biggest revolution. PRESENTER NOTES
  • 15.
    This is happeningnow: the companies listed in this slide provide the fundamental software components to create advanced Artificial Intelligence applications. PRESENTER NOTES
  • 16.
    And those othercompanies integrate that libraries to create vertical products on specific markets. Artificial Intelligence can perform tasks that since few years ago could be just performed by humans. The FIRST industrial revolution replaced the MUSCLE POWER, the coming FOURTH industrial revolution will replace the BRAIN POWER. As soon as the major companies will adopt these technologies, there will be an enormous impact on the job market. Repetitive tasks of progressively higher complexity will be automated as the technology evolves. PRESENTER NOTES
  • 17.
    This slide showssome results from a research made by the University of Oxford. In this case the telephone salespersons and the typists are two professions that will be impacted by machines that can understand the human spoken language and relative concepts. This will allow to automate call centres, customer care desks, automatic translation. PRESENTER NOTES
  • 18.
    GoogleTranslate This is notscience fiction: you have a working example in your pocket: just check your Google Translate or your Apple Siri. PRESENTER NOTES
  • 19.
    Other jobs tobe automated are those involving vision and handling. This is made possible by the Artificial Intelligence that can understand highly complex scenes in cluttered environments. Lets see two demo video that we made in Addfor. In the first one we will see how Artificial Intelligence recognise and classify objects in a complex situation. In the second video, the Artificial Intelligence describes a picture in plain english. Just try to image how can archives of million of images can be scanned and queried using plain written questions. PRESENTER NOTES
  • 20.
  • 21.
    IT IS NOTAN OPTION Adopting the new Artificial Intelligence technologies wouldn’t be an option, it is the same as it was with the steam engine or the assembly line invented by Ford: those who do not adopt this technology will simply be excluded from the market. The development wouldn't stop: there is simply to much money involved. NVIDIA was the best-performing stock in 2016 S&P 500 delivering a 225% total return, it increased its market capitalisation by 11 BILLION dollar just in one day, in November 2016, the day of its Q3 earnings call. PRESENTER NOTES
  • 22.
    ALREADY HAVE SOMEINTERNAL SKILL ? ASK FOR A CUSTOMTRAINING ASK FOR ANALYSIS AND PROOF OF CONCEPT NO INTERNAL SKILLS ? The adoption of a disruptive new technology can be risky and requires investments. You should take a safe path. There are many available software solutions on the market but just some of them will fit your needs. We are completely agnostic on hardware and software solutions, for this reason we can help our customers in making the best choices. Maybe the best start could be to share some ideas with us. Because the biggest risk for your company is just not being informed and not being ready. PRESENTER NOTES
  • 23.
    ARTIFICIAL INTELLIGENCE: SOMEEXAMPLES Now we will see some examples of Artificial intelligence algorithms that are readily available to be deployed on vertical solutions. This is just a partial list because those technologies can have a broad range of applications, nevertheless we think that those examples can be quite informative. PRESENTER NOTES
  • 24.
    UNDERSTAND THE MARKET Thefirst example is about finding Anomalies in the data streams. This class of algorithms can be applied to a wide range of problems to find glitches in production systems, unusual behaviours in supply chains, financial systems, energy and communication networks. Finding an Anomaly is important to trigger proper reactions, for example, it can be used for predictive maintenance or for discover some Latent Business Opportunities in market data. PRESENTER NOTES
  • 25.
    EXAMPLE: FIND ANOMALIESINTAXI CALLS In this case we analyse the NYC Taxi Calls Database: it contains the number of taxi calls hour by hour in the year 2014. PRESENTER NOTES
  • 26.
    ALGORITHM FINDS ANOMALIES BYANALYSING RAW DATA The algorithm has been supplied with only the raw data. No other data (like holidays or special events has been provided). Here we see the algorithms that start to analyse the data and after a while understands that the data presents a normal seasonality and normal behaviours at peak hours. When the behaviour is different from what the algorithms predict the point is defined as an anomaly and is plotted in yellow or red depending of the anomaly strength. PRESENTER NOTES
  • 27.
    FIRST ANOMALY: XMAS+ NEWYEAR The first strongest anomalies detected have an easy explanation: on new year’s day there is a peak of taxi calls just after midnight and on December 25th there are few calls. Remember that the algorithms do not have any information about holidays, it sees Christmas just as a standard day, for this reasons it detects an abnormal pattern in the data. PRESENTER NOTES
  • 28.
    SECOND ANOMALY: SUNDAY AFTER HALLOWEEN Thesecond anomaly is detected for halloween and the night after when the people get back home. PRESENTER NOTES
  • 29.
    THIRD ANOMALY: The thirdanomaly was about a strange low volume of calls the night after the thanksgiving day. It should be due to a bombing thread spread from the news PRESENTER NOTES
  • 30.
    LAST ANOMALY:WHAT’S APPON DEC 6 2014 ? What’s strange is this anomaly detected on December the 6th: it’s difficult to spot by eyes but the system detected here a strong glitch in data. Seemingly nothing special happened in NYC that day. PRESENTER NOTES
  • 31.
    Just after lookingback in the NYC news we found this event that seemingly brought to NYC many unexpected visitors. Summarising, Artificial Intelligence can watch data streams 24/7 finding irregular behaviours, faults, and useful information to improve your systems and your businesses. PRESENTER NOTES
  • 32.
    OPTIMIZE PRODUCTION The secondexample is about production optimisation. In particular we will see how Artificial Intelligence can use the sensor’s data to predict system failures. PRESENTER NOTES
  • 33.
    EXAMPLE: ANALYSE MACHINERYDATA The most common way to supervise machines is to create Dashboards for human operators. This approach has many limits: Time and human attention for example. Moreover humans cannot supervise hundreds of data streams at a time. PRESENTER NOTES
  • 34.
    ARTIFICIAL INTELLIGENCE CAN WATCH100’S OFVARIABLES AND FIND USEFUL INFORMATIONS FAILUREORIGINAL I this example our algorithm controls 100’s of variables together. Every single variable by itself is meaningless, nevertheless, we made a system based on a Deep Neural Network that combine all the variables to find the information we need. In this case the left graph shows the raw variables while the right graph shows the output our our system that predict a jet engine failure. PRESENTER NOTES
  • 35.
    ORIGINAL DATA FAILURE MODES Herethe Deep Learning algorithm finds three specific failure modes indicated by the three red circles on the right graph PRESENTER NOTES
  • 36.
    OFFER NEW FEATURES Thelast example is about using Artificial Intelligence to create new feature and set new market standards. In this image you can see a NEST thermostat that set new standards for home automation devices. PRESENTER NOTES
  • 37.
    EXAMPLE: ADVANCEDTRACTION CONTROLS Buttoday I would like to bring you an Italian example that explains how Ferrari was the first carmaker to put an Artificial Intelligence system in its production Electronic Control Units. Nowadays this system is installed on every Ferrari model. To explain the system in detail I could like to give the floor to Marco Fainello PRESENTER NOTES
  • 38.
  • 39.
    We don’t knowexactly how Intelligent Machines will evolve It will be an Exponential Growth … This mean that it will happen sooner than we expect
  • 40.