Ai business innovator v001
1 like183 views
The document discusses artificial intelligence (AI) and machine learning. It provides examples of how machine learning works including showing the training process and making inferences from new data. It then discusses applications of AI like computer vision, natural language processing, reinforcement learning and more. It also discusses the growth and capabilities of AI over time as well as implications and applications in different industries.
![Quanti dati ci vogliono ?
add-for.comenrico.busto@add-for.com
INPUT SOLUZIONE
[7.45, 18.74, 323.27, 32.43, 873.79, 2.83, 82.25, 18.36] 0.45
… …
Dipende dalla complessità del problema](https://image.slidesharecdn.com/aibusinessinnovatorv001-180607162418/85/Ai-business-innovator-v001-5-320.jpg)
Ai business innovator v001
- 2. [email protected] Che cos’è l’ A.I. ?Che cos’è l’ A.I. ?
- 3. Paradigma ML / DL [email protected] DATI PROGRAMMA SOLUZIONE DATI PROGRAMMATRAINING SOLUZIONE NUOVI DATI SOLUZIONE INFERENZA
- 5. Quanti dati ci vogliono ? [email protected] INPUT SOLUZIONE [7.45, 18.74, 323.27, 32.43, 873.79, 2.83, 82.25, 18.36] 0.45 … … Dipende dalla complessità del problema
- 6. [email protected] Artificial Intelligence (AGI) Artificial General Intelligence (Strong AI, Full AI) (Narrow AI) Narrow Artificial Intelligence (Weak AI,Applied AI) Machine Learning Clustering, Classification, Regression Deep Learning Image /Video Understanding Voice Recognition Reinforcement Learning Data Mining / Data Wrangling Selection Preprocessing / Cleaning Transformation Aggregation DataBases Structured Unstructured Data Streams Web IoT Machines
- 7. NEOCORTEX 6 layers in mammals 2-4 mm thick size: large dinner napkin ~30 B neurons - 100 k x mm2 Brain mass: 1.2 - 1.4 kg Total Neurons: ~100 B
- 10. “Based on its level seen in the match, I think I’ll win the game by a near landslide” 2015-10 Lee Sedol
- 11. “Based on its level seen in the match, I think I’ll win the game by a near landslide” 2015-10 “I’ve heard that DeepMind’s AI is really strong, but I’m confident I can win at least at this time” 2016-02 Lee Sedol
- 12. “Based on its level seen in the match, I think I’ll win the game by a near landslide” 2015-10 “I’ve heard that DeepMind’s AI is really strong, but I’m confident I can win at least at this time” 2016-02 THENTHERE WASTHE FIRST GAME… “I’ve been very surprised because I didn’t think I would lose” 2016-03-09 Lee Sedol
- 13. “Based on its level seen in the match, I think I’ll win the game by a near landslide” 2015-10 “I’ve heard that DeepMind’s AI is really strong, but I’m confident I can win at least at this time” 2016-02 THENTHERE WASTHE FIRST GAME… “I’ve been very surprised because I didn’t think I would lose” 2016-03-09 THENTHE SECOND GAME… “I’m quite speechless… I’m in shock. I can admit that the third game is not going to be easy for me” 2016-03-10 Lee Sedol
- 14. “Based on its level seen in the match, I think I’ll win the game by a near landslide” 2015-10 “I’ve heard that DeepMind’s AI is really strong, but I’m confident I can win at least at this time” 2016-02 THENTHERE WASTHE FIRST GAME… “I’ve been very surprised because I didn’t think I would lose” 2016-03-09 THENTHE SECOND GAME… “I’m quite speechless… I’m in shock. I can admit that the third game is not going to be easy for me” 2016-03-10 GAME OVER… “I kind of felt powerless”2016-03-12 Lee Sedol
- 16. AlphaGo Lee - 48TPUs2016-03 4:1 agains Lee Sedol AlphaGo Master - 4TPUs2017-01 60:1 agains Professional players AlphaGo Zero - 4TPUs2017-01 hour 00 - no prior knowledge AlphaGo Zero day 03 - surpasses AlphaGo Lee AlphaGo Zero day 21 - surpasses AlphaGo Master AlphaGo Zero day 40 - no available metrics hour 03 - plays as a human beginnerAlphaGo Zero SUPERINTELLIGENCE …
- 17. AlphaGo Lee - 48TPUs2016-03 4:1 agains Lee Sedol Chinese government laid out a plan to become world leader in A.I. by 2030 2017-07 Chinese government blocked live video coverage of AlphaGo beating the world Go champion Ke Jie 2016-05 AlphaGo Master - 4TPUs2017-01 60:1 agains Professional players AlphaGo Zero - 4TPUs2017-01 hour 00 - no prior knowledge AlphaGo Zero day 03 - surpasses AlphaGo Lee AlphaGo Zero day 21 - surpasses AlphaGo Master AlphaGo Zero day 40 - no available metrics hour 03 - plays as a human beginnerAlphaGo Zero SUPERINTELLIGENCE …
- 20. MOLTIPLICATORE DI CAPACITA’MOLTIPLICATORE DI CAPACITA’ INNOVAZIONE DI PRODOTTO INNOVAZIONE AUTOMAZIONE INTELLIGENTEAUTOMAZIONE INTELLIGENTE1 2 3
- 22. 2 MOLTIPLICATORE DI CAPACITA’ UMANEMOLTIPLICATORE DI CAPACITA’ UMANE
- 23. 2 MOLTIPLICATORE DI CAPACITA’ UMANEMOLTIPLICATORE DI CAPACITA’ UMANE
- 24. 2 MOLTIPLICATORE DI CAPACITA’ UMANEMOLTIPLICATORE DI CAPACITA’ UMANE
- 25. INNOVAZIONEINNOVAZIONE3 DI PRODOTTODI PRODOTTO
- 27. I DATI SONO LA BASE DEL VALORE I DATI SONO LA BASE DEL VALORE
- 28. John Deere spent 305M$ on Lettuce Farming Robot John Deere spent 305M$ on Lettuce Farming Robot
- 29. GESTIRE IL CAMBIAMENTOGESTIRE IL CAMBIAMENTO
- 30. + ≠ www IMPARIAMO DAL PASSATOIMPARIAMO DAL PASSATO
- 31. + ≠ PER NON RIPETERE GLI STESSI ERRORIPER NON RIPETERE GLI STESSI ERRORI + ≠ www
- 34. Learning by DOINGLearning by DOING The design-thinking ideology asserts that a hands-on + user-centric approach to problem solving can lead to innovation, and innovation can lead to differentiation and a competitive advantage ** DATA AUGMENTATION
- 39. TRAININGTRAINING
- 40. DATA ENGINEER DATA SCIENTIST BUSINESS ANALYST APP DEVELOPER Organizza i dati e li rende disponibili Analizza i dati, trova e organizza informazioni e correlazioni Riporta le conoscenze acquisite dai dati sulle strategie aziendali Si interfaccia con le basi dati e crea le applicazioni software DEVOPS Gestisce le strage aziendali e Implementa nuovi modelli di Business Comprensione del Business e del Dominio Interfacciamento alle Basi Dati Esplorazione e comprensione dei Dati Trasformazione e pulizia Rimodellazione e Aggregazione Normalizzazioni e Statistiche Creazione del Modello Creazione dei Reports Messa in Produzione
- 41. Userai software allo stato dell’arte Risolverai in poche linee di codice problemi impossibili per Microsoft Excel Accederai a fonti eterogenei di dati Imparerai a fondere, completare e integrare dati di tipo diverso Imparerai a gestire serie storiche e fusi orari E poi, Pulizia dei dati, formattazioni multiple, creazione di reports Creare Grafici Significativi DATA ENGINEER TRAINING CAMP COSA IMPARERAI Artificial Intelligence Machine Learning Deep Learning Data Mining / Data Wrangling DataBases Data Streams
- 42. DATA SCIENTIST TRAINING CAMP COSA IMPARERAI Userai software allo stato dell’arte Scoprirai come superare i limiti della statistica tradizionale Preparazione dei dati, estrazione del contenuto informativo Metodi Previsionali: imparerai a “prevedere” il futuro osservando il passato Rilevamento di anomalie, tecniche di manutenzione predittiva Classificazione di eventi e comportamenti, recommendation systems Grafica e reportistica Artificial Intelligence Machine Learning Deep Learning Data Mining / Data Wrangling DataBases Data Streams
- 43. DEVELOPMENT OPERATIONS AI for Middle-Top Management Che cos’è l’Intelligenza Artificiale ? Come posso usarla per trasformare il mio business e trarne vantaggio Analisi dei dati: quali sono le tecniche per acquisire informazioni di valore Visione Artificiale: dalla sicurezza alla robotica, che cosa può fare DigitalTwins: che cosa sono e come si impiegano i “gemelli” matematici Natural Language Processing e Chatbots: stato dell’arte e applicazioni Poi svilupperemo insieme un piano industriale completo per prototipare, verificare e portale sul mercato un prodotto completamente innovativo. In 1 giorno risponderemo a : DEVOPS
- 46. COMPETENZA DI DOMINIO COMPETENZAAICOMPETENZAAI COMPETENZA DI DOMINIO
- 48. MACHINE VISION / IMAGE-VIDEO UNDERSTANDING SIGNAL PROCESSING / VIRTUAL SENSORS ANOMALY DETECTION / FRAUD DETECTION PREDICTIVE MAINTENANCE FORECASTING REINFORCEMENT LEARNING APPLICAZIONI ML / DLAPPLICAZIONI ML / DL
- 50. MACHINE VISION / IMAGE-VIDEO UNDERSTANDING
- 51. MACHINE VISION / IMAGE-VIDEO UNDERSTANDINGMACHINE VISION / IMAGE-VIDEO UNDERSTANDING RICONOSCIMENTO NUMERI DI TARGARICONOSCIMENTO NUMERI DI TARGA
- 52. MISURA DELLAVELOCITÀ Misura dellaVelocità Media / Massima in Tempo Reale per singola corsia - Previsione Code
- 53. STATISTICHE Conteggio veicoli per classe in tempo reale - Medie e distribuzioni statistiche
- 54. RILEVAZIONE PERICOLI Veicoli Fermi - Carichi Dispersi - Manovre Pericolose
- 55. MACHINE VISION / IMAGE-VIDEO UNDERSTANDINGMACHINE VISION / IMAGE-VIDEO UNDERSTANDING
- 56. MACHINE VISION / IMAGE-VIDEO UNDERSTANDINGMACHINE VISION / IMAGE-VIDEO UNDERSTANDING RICONOSCIMENTO BAGAGLIRICONOSCIMENTO BAGAGLI
- 57. FIND PEOPLE INVIDEO ARCHIVES www.add-for.com/portfolio-item/broadcasting-security Crowd Monitoring (Security Applications) https://youtu.be/4Z31NHkL1qw
- 58. FIND SIMILAR FACES Find celebrities and person of interest in Video Archives
- 59. UNDERSTAND EMOTIONSwww.add-for.com/portfolio-item/broadcasting-security Face Keypoints and Morphological Measures https://youtu.be/N5LwZE0VRnA
- 60. SCENE RETRIEVAL INVIDEO STREAMS ROSSI Valentino MARQUEZ Marc www.add-for.com/portfolio-item/broadcasting-security Specific actions in SportVideo Streams, available in Real-Time
- 61. MACHINE VISION / IMAGE-VIDEO UNDERSTANDINGMACHINE VISION / IMAGE-VIDEO UNDERSTANDING TRAIETTORIE E STATISTICHETRAIETTORIE E STATISTICHE
- 62. SIGNAL PROCESSING / VIRTUAL SENSORS
- 64. ADVANCED NONLINEAR CONTROLS WASHING MACHINES add-for.com
- 65. DRIFT ANGLE
- 67. ANOMALY DETECTION / FRAUD DETECTION
- 68. ANOMALY DETECTION / FRAUD DETECTIONANOMALY DETECTION / FRAUD DETECTION
- 69. LIMITI POCHI PARAMETRI SOLO SITUAZIONI SEMPLICI SOLO FRODI CONOSCIUTE SUBITO COMPLESSITÀ GESTIONE ONEROSA “Aggiungere regole è facile trovare regole significative è molto difficile”
- 70. MULTIMODAL VISUAL THERMAL VIBRATIONS NOISE ANOMALY DETECTION / FRAUD DETECTIONANOMALY DETECTION / FRAUD DETECTION
- 71. ADAPTIVE LIGHTING CONTROLS EARLY FAULT DETECTION add-for.com
- 74. Suggest the best experience to your Users RECOMMENDATION SYSTEMS
- 75. CHATBOTS
- 79. REGRESSION
- 80. CLASSIFICATION TRAINING Set - Solid Color TEST Set - SemiTransparent Lower Right - Classification accuracy onTest Set
- 81. CLUSTERING
- 82. Silhouette coefficients (as these values are referred to as) near +1 indicate that the sample is far away from the neighboring clusters. A value of 0 indicates that the sample is on or very close to the decision boundary between two neighboring clusters and negative values indicate that those samples might have been assigned to the wrong cluster. In this example the silhouette analysis is used to choose an optimal value for n_clusters. The silhouette plot shows that the n_clusters value of 3 is a bad pick for the given data due to the presence of clusters with below average silhouette scores and also due to wide fluctuations in the size of the silhouette plots. Silhouette analysis is more ambivalent in deciding between 2 and 4.
- 84. NEURAL NETWORKS
- 88. GENERAL ANOMALY DETECTOR -TAXI CALLS In this case we analyse an Open Source DB: the NYC Taxi Calls: it contains the number of taxi calls hour by hour in the year 2014. PRESENTER NOTES
- 89. ALGORITHM FINDS ANOMALIES BY ANALYSING RAW DATA The algorithm has been supplied with only the raw data. No other data (for example - holidays or special events) has been provided. Here we see the algorithms that start to analyse the data and after a while are able to deduct normal daily behaviour and during the peak hours. When the behaviour differs from the algorithm prediction the point is defined as an anomaly and is plotted in yellow or red depending of the anomaly strength. PRESENTER NOTES
- 90. FIRST ANOMALY: XMAS + NEWYEAR The first 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 very few calls. Remember that the algorithms do not have any information about holidays, it sees Christmas just as a standard day, for this reason it detects an abnormal pattern in the data. PRESENTER NOTES
- 91. SECOND ANOMALY: SUNDAY AFTER HALLOWEEN The second anomaly is detected for halloween and the night after when the people return home. PRESENTER NOTES
- 92. THIRD ANOMALY: The third anomaly was regarding a low volume of calls the night after Thanksgiving. It was probably due to a bomb threat spread by the news. PRESENTER NOTES
- 93. 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 eye but here the system detected a strong glitch in data. Seemingly nothing special happened in NYC that day. PRESENTER NOTES
- 94. After having looked back in the NYC News we found this event that seemingly brought many unexpected visitors to NYC Summarising, Artificial Intelligence can watch data streams 24/7 finding irregular behaviour, faults, and useful information to improve your systems and your businesses. PRESENTER NOTES
- 95. ANOMALY DETECTION An e-commerce company sells gift cards and sees an unexpected increase in the number of cards purchased. While that sounds like a great thing, there is also a corresponding drop in the revenue expected for the gift cards. Something strange is going on, and it turns out to be a price glitch— something quite common for e-commerce companies. Without looking at these two metrics together, it is hard to understand that there is a business incident that could cost the company a lot of money if not caught and addressed quickly.
- 96. WHAT IS AN ANOMALY? for both machines and humans – is identifying an anomaly.Very often the problem is ill-posed, making it hard to tell what an anomaly is.
- 97. Looking at this example, we could say that the two days highlighted in the red box are anomalous because every other day is much higher. Here the anomaly is related to the rate of change Conciseness: the system takes multiple metrics into account at once for a holistic look at what is happening.
- 98. As an example of what is meant by conciseness: consider the human body as a type of system. It is possible to measure a person's vital signs, including blood pressure, body temperature, pulse rate and respiration rate. With univariate anomaly detection, the system looks at each metric by itself, learning its normal patterns and yielding a list of anomalies for each single metric. The advantage of univariate anomaly detection is that it is a lot easier to do than other methods. However, when something unexpected happens that affects a lot of metrics, the system yields a storm of anomalies. Now someone has to sift through them to understand what is happening. Multivariate anomaly detection techniques take input from all the signals together as one, without separating them out. All of the metrics are taken as input but the output simply says there is something strange—an anomaly, without identifying which metric(s) it is associated with. In the healthcare analogy, the doctor would put in the vital signs and receive “the patient is sick,” without any further explanation about why. Addfor learns what is normal for each one of the metrics by themselves, and after detecting anomalies the system checks if it can combine them at the single metric level into groups and then give an interpretation to that group.
- 100. ORIGINAL DATA FAILURE MODES add-for.com HERE THE ALGORITHM PERFORM A DIMENSIONAL REDUCTION R770 > R2 TO VISUALISE THREE DIFFERENT FAILURE MODES
- 104. PROVEN SUCCESSFUL IN MANY APPLICATIONS
- 105. Reinforcement Learning - Paradigm add-for.com AGENT ACTION REWARD ENVIRONMENT
- 106. A.I. PLAYS WITHTHE (VIRTUAL) ENVIRONMENT TO LEARN
- 107. WE “STOLE”THE IDEA FROM
- 108. OUR SYSTEM ACTIVE SINCE JUNE 2017 ON SELECTED OVS STORES
- 109. INVENTIAMOINVENTIAMO