AI4All, Etica en IA - AI Saturdays Tarragona

Transcript

1. INTELIGENCIA ARTIFICIAL junto con COOPPEL por la PAZ 22/03/2019, Tarragona, INS Martí i Franqués tarragona@saturdays.ai
2. Institut Antoni de Martí i Franquès http://www.institutmartifranques.com/ Causa Cooppeel por la paz Josep Maria Ferran Jordi, Meritxell, Nuria y Lucía
3. Jordi Guix Big Data Consultant
4. Saturdays.AI is a Non-Profit Organization in a mission to deliver AI4Good brought to you with LOVE by diverse individuals like you, why not join us? CO ORGANIZER CO FOUNDER https://www.saturdays.ai CO ORGANIZER
5. Our course during the Saturdays….
6. Artificial Intelligence is happening
7. The 4th Industrial Revolution
8. But, what is AI? How it works?
9. The 4th Industrial Revolution “85% of jobs that will exist by 2030 have not been yet invented.“ - What is the root cause of this trend? - What will be the implications for our society?
10. Previous revolutions: 2nd
11. Previous revolutions: 3rd
12. Revolution NOW - Act & Learn now NO YES If it is written in Python, it’s probably machine learning If it is written in PowerPoint, it’s probably Artificial Intelligence sanity check
13. AI Saturdays….
14. Learn by doing 1st Code2Learn (~8 saturdays) code exercises on notebooks 2nd Build2Learn (~6 saturdays) build an end2end AI prototype + Demo Day credit: Fast.ai Jeremy Howard
15. AI Saturdays Community Began on January 2018! +15 +50 +1000 +100 Cities Ambassadors Members AI Projects + Alicante, Bogota, Lima, Monterrey, Santiago, Sevilla...
16. Community building Intensive focus In a world of distractions, we provide the right environment and incentives so our fellows can focus on learning and coding AI during intense work sessions. Together with our activities and support, we transform such focus into an active community. Pair-Programming The key to our courses are our alumni, their motivation and their ability to help each other. We match them according to their goals and ensure that they work together to push their projects beyond their comfort zone. Learn By Doing The AI scientist mindset is forged by combining the theoretical lessons and building open-source projects, which enables our students to implement and test end-to-end AI apps with real customers.
17. Enroll to AI Saturdays Tarragona en bit.ly/aifellows Tarragona, 12 Sábados de 09 a 14h desde 13.4.2019 a 29.6.2019 + DemoDay gratuito, sostenido por donaciones
19. Intelligent Assistants….
20. Robots kill the Human...
21. Artificial Intelligence? *Watson IBM (Cognitive System) -Entiende y procesa el lenguaje -Responde a preguntas complejas -Aprende de cada experiencia
22. Machine Learning: it’s a subset of AI which uses statistical methods to enable machines to improve with experience, without being explicitly programmed. Deep Learning is a subfield of machine learning concerned with algorihms inspired by the structure and function of the brain called artificial neural networks” (the computer can learn complicated tasks as image recognition) Machine Learning and Deep Learning?
23. Machine Learning
25. ...AI application! recurrent neural network (RNN) -sensor IoT -tensorflow - Deep Learning
26. In 2017 the estimated new cases of cancer in Spain have been 228,482 and by 2035 it is estimated that there will be 315,413 new cases of cancer The most diagnosed tumors remain colorectal, prostate, lung, breast, bladder and stomach. source: SEOM, Jan 2018 In 2018, an estimated 1,735,350 new cases of cancer will be diagnosed in the United States and 609,640 people will die from the disease. source: https://www.cancer.gov/ Built successful AI systems for healthcare Improving Health Treatment with Doctors Review million of cases (Doctors only thousands)
27. Predict whether the cancer is benign or malignant ? Wisconsin (Diagnostic) Data Set source:https://www.kaggle.com/uciml/breast-cancer-wisconsin-data 1. The characteristics are calculated from a digitalized image of a fine needle aspiration (FNA) of a breast mass 2. Class distribution: 357 benign, 212 malignant 3. Y=Diagnosis 4. X=radius, texture, perimeter, smoothness, compactness, concave points, symmetry, fractal dimension
28. Predict the onset of diabetes based on diagnostic measures https://www.kaggle.com/uciml/pima-indians-diabetes-database -National Institute of Diabetes and Digestive and Kidney Diseases (india) -All the patients here are women of at least 21 years of the Pima Indian heritage. -Predict the diagnostic form and the patient has diabetes or not x=Pregnancies, Glucose, Bood PBMI, insulin level, age,skin y= outcome 0 or 1
29. https://arxiv.org/abs/1805.00794 https://www.kaggle.com/shayanfazeli/heartbeat/home Identify a heart attack? -This dataset is composed of heartbeat signals -Deep neural network -make predictions with the average accuracies of 93.4% and 95.9% on arrhythmia classification and MI classification
30. CRUCIAL to UNDERSTAND AI! Wow!
31. ATTENTION!
32. ….IT IS NOT ALL GOLD WHAT SHINES ......
33. Cospiracy Theory “Orchestrated George W. Bush a plan to topple the Twin Towers and kill thousands of people in 2001?” “The world is flat? “
34. The AI increases human prejudice and discrimination Source: Latest data available from the companies, since 2017.
35. Chinese Street Surveillance • About 20 million CCTV cameras have been installed with AI technology in China • Such technology can be used to identify a person's age, gender and clothes • Police can track down criminals using facial recognition and their database https://www.dailymail.co.uk/news/article-4918342/China-installs-20-million-AI-equipped-street-cameras.html
36. Chinese Street Surveillance
38. European Ethic Code El Parlamento Europeo (PE), tras una propuesta de borrador (junio 2016) y un informe razonado (febrero 2017) ha aprobado un informe sobre Robótica en el que se establece un Código Ético de Conducta. Los principios sancionados por el Parlamento Europeo 1. Proteger a los humanos del daño causado por robots: la dignidad humana. 2. Respetar el rechazo a ser cuidado por un robot. 3. Proteger la libertad humana frente a los robots. 4. Proteger la privacidad y el uso de datos: especialmente cuando avancen los coches autónomos, los drones, los asistentes personales o los robots de seguridad. 5. Protección de la humanidad ante el riesgo de manipulación por parte de los robots: Especialmente en ciertos colectivos –ancianos, niños, dependientes- que puedan generar una empatía artificial. 6. Evitar la disolución de los lazos sociales haciendo que los robots monopolizan, en un cierto sentido, las relaciones de determinados grupos. 7. Igualdad de acceso al progreso en robótica: Al igual que la brecha digital, la brecha robótica puede ser esencial. 8. Restricción del acceso a tecnologías de mejora regulando la idea del transhumanismo y la búsqueda de mejoras físicas y/o mentales.
39. ...El hombre es feliz por sus actos, lo cual le da el placer que se encuentra en el alma como esencia, si se desarrolla (ser bueno y feliz) el hombre alcanza el placer… Philosophy of AI on Saturdays Nosotro somos felices democratizando el conocimiento del AI, nosotros queremos que puede ayudar el progreso científico, social y tecnológico.
40. Deep learning to Save Lives in the Sea https://github.com/sfrias/sat4survive a CNN using Satellites to detect adrift small boats in the coast of Libya and empower anyone with a smartphone to save lives Inspired by
41. Bias Corrector Aims to reduce cognitive biases in AI by creating a tool that helps to generate online content that is less offensive and that shows special to sexist and racist, proof of concept combining a RNN and a LSTM network. Authors: Irene Font Perajordi, Nora Lama and David Collado
42. Far away on implementation, right PoC https://www.blog.google/technology/ai/fight-against-illegal-deforestation-tensorflow/ Reducing gender bias in Google Translate https://www.blog.google/products/translate/reducing-gender-bias-google-translate/ AI Sensors to fight Illegal Deforestation
43. Charging Everywhere EVs with Machine learning https://github.com/sfrias/place2plug in partnership with
44. Thank You unete en www.saturdays.ai y TGN a bit.ly/aifellows tarragona@saturdays.ai