Demystifying Artificial Intelligence
45 likes53,752 views
The presentation aims to introduce artificial intelligence (AI), provide an overview of its evolution and current state, and guide organizations in implementing AI solutions. It covers key topics including the history of AI, various types of AI such as artificial narrow intelligence (ANI) and artificial general intelligence (AGI), and practical applications in businesses today. The document emphasizes the importance of understanding AI's potential impact on society and addresses the necessary steps for organizations to leverage AI technology.
Demystifying Artificial Intelligence
- 2. 3 Goals of this presentation #1 Introduce you to Artificial Intelligence #2 Provide a high-level overview of today & tomorrow #3 Help organisations get started with AI
- 3. Speaker on Bitcoin & Blockchain Consultant at Duval Union Consulting sam.wouters@duvalunion.com Sam Wouters @sdwouters Made by Speaker on Artificial Intelligence
- 4. ‣ We are a digital consulting company. ‣ Founded in 2009. ‣ We work for large clients across all sectors. ‣ We have a strong vision on the impact of digital and act accordingly. ‣ We are advisors, writers, coaches and entrepreneurs.
- 5. ‣ Slide 8-22: A brief history of Artificial Intelligence ‣ Slide 23-36: Demystification ‣ Slide 37-44: AI is all around us ‣ Slide 45-63: Going from ANI to AGI ‣ Slide 64-75: Going from AGI to ASI ‣ Slide 76-98: How to use ANI ‣ Slide 99-110: How to start building Table of contents
- 6. How to define AI intelligence exhibited by machines
- 7. How to define AI ARTIFICIAL INTELLIGENCE DEEP LEARNING MACHINE LEARNING A building block of AI A subset of Machine Learning Intelligence by machines
- 9. First AI: The Staffelwalze The first known calculator to perform all 4 operations: addition, subtraction, multiplication and division. 1672
- 10. In the 1940s the Church–Turing thesis was created. It suggested that digital computers can simulate any process of formal reasoning and led researchers to consider the possibility of building an electronic brain. 1940s
- 11. AI started getting developed by philosophers and mathematicians in the 19th century. Alan Turing is one of the best known founding fathers of modern AI. 1950s
- 12. INTERVIEWER PERSON AI The interviewer has a limited amount of time to ask questions to the other rooms. An interviewer, a person and an AI get put in 3 different rooms. They communicate through text. If the interviewer can’t figure out who is human before time is up, the AI passes the Turing Test.
- 13. AI research was founded at Dartmouth College in 1956. The founders and their students wrote astonishing programs that were winning at checkers, solving word problems in algebra, proving logical theorems and speaking English. They expected to have pure AI within twenty years, but underestimated the difficulty and in 1974 all funding was cut off from AI research. 1956-74 Founding of AI & practice
- 14. After a few more attempts at revival, AI began to be used more in the 90s in all kinds of areas. In 1997, IBM’s Deep Blue beat Garry Kasparov, the reigning world champion at chess. 1990s
- 15. In 2011 IBM’s Watson beat Ken Jennings and Brad Rutter in the TV Quiz Jeopardy, the first sign of AI beating people at non-math. 1990s
- 16. In 2016, Google’s Deepmind beat Lee Sedol 4-1 at Go, putting AI 10 years ahead of expectations. 2016
- 17. Today AI is booming
- 19. AI Funding since 2012: $14.9 BILLION 2250 DEALS
- 21. MORE DATA & RESEARCH Why now? ERROR RATES ARE FALLING CHEAP NEURAL NETWORKS “2015 was a landmark year for AI, with software projects using AI at Google increasing from ‘sporadic usage’ in 2012 to over 2700 projects.” ~Jack Clark, Strategy & Communications Director at OpenAI
- 22. Why does AI matter? OUR LIVES OUR BUSINESSES OUR SOCIETY How will AI impact our purpose? How will AI change our businesses? How will AI impact our workforce?
- 26. Human progress Time Reason #1: Projection “This is how things went in the past, so that will probably continue but a bit faster”
- 27. Human progress Time 1. Experimentation 2. Exponential growth 3. Level off Reason #2: Innovation happens in phases 1 2 3 1 2 3 1995-2007 2008-2015
- 28. It’s not about robots It’s about the brain
- 29. 3 Types of Artificial Intelligence Artificial Narrow IntelligenceANI
- 30. Specialised in ONE area ANI Artificial Narrow Intelligence
- 31. 3 Types of Artificial Intelligence Artificial Narrow IntelligenceANI AGI Artificial General Intelligence
- 32. AGI Specialised in ALL areas Artificial General Intelligence Our best work in progress
- 33. AGI 7 ABILITIES: 1. Reason 2. Plan 3. Solve problems 4. Think abstractly 5. Comprehend complex ideas 6. Learn quickly 7. Learn from experience Artificial General Intelligence
- 34. 3 Types of Artificial Intelligence Artificial Narrow IntelligenceANI AGI Artificial General Intelligence ASI Artificial Super Intelligence
- 35. ASI Smarter than humans in EVERY WAY Artificial Super Intelligence
- 36. 3 Types of Artificial Intelligence Artificial Narrow IntelligenceANI AGI Artificial General Intelligence ASI Artificial Super Intelligence
- 38. ANI is all around us
- 40. Google says its AI catches 99.9% of Gmail spam
- 41. Netflix uses AI to go from recommendations based on what you’ve seen, to what you like
- 42. A 19-year-old made a free chatbot lawyer that has appealed $3m in parking tickets
- 43. Identifying diseases by comparing huge amounts of data Freenome, a startup focused on detecting cancer through AI, set up by a 28-year old, landed a $65 million investment by Andreessen Horowitz & Google Ventures
- 44. Uber is using AI for Route optimization
- 46. We build computers to process large amounts of data = Hard to humans
- 47. Next step: things that are easy to humans What is a cat? What is a human face? Which of these is a tree?
- 48. A partially lit-up 3D rock to humans Grey & black 2D shapes to a computer
- 49. Heavy investments to achieve AGI
- 50. “I truly believe the computing Singularity is coming, that’s why I’m in a hurry – to aggregate the cash, to invest.” ~Softbank CEO Masayoshi Son on their $100B tech fund
- 52. How to achieve AGI CHEAPER COMPUTER POWER BECOMING SMARTER
- 53. Goal: $1000 for human brainpowerCHEAPER COMPUTER POWER 1015 calcs/second 20 watts of power 0.00126M3
- 54. Our best computer: $390M Tianhe-2CHEAPER COMPUTER POWER 1034 calcs/second 24 MW of power 2160M3
- 55. We should have enough cheap computing power by 2025 CHEAPER COMPUTER POWER
- 56. How to achieve AGI CHEAPER COMPUTER POWER BECOMING SMARTER
- 57. 3 Strategies #1 Reverse Engineering #2 Stimulating Evolution #3 Autonomy BECOMING SMARTER
- 58. #1: Reverse-EngineeringBECOMING SMARTER Replicating human brains in a digital way
- 59. #2: Replicating EvolutionBECOMING SMARTER Networks of hardware & software learning from each others mistakes
- 60. #3: Autonomy Letting the computer write its own rules and learn from experience BECOMING SMARTER Wolfpack gameGathering game
- 61. Experts expect we’ll achieve AGI between 2030-2060 BECOMING SMARTER
- 62. We could be missing 1 pieceBECOMING SMARTER
- 63. Example: digital money David Chaum - 1983 - Digital Cash Satoshi Nakamoto - 2008 - Bitcoin
- 65. AGI doesn’t existLevel of AI Time Where are are right now Human intelligence AGI ANI
- 66. An AGI has max skills in everything 0 25 50 75 100 Programming Pricing products Catching objects Fraud detection Identifying diseases Human AI
- 67. And it can re-engineer its brain
- 68. AGI doesn’t existLevel of AI Time <1 day Human intelligence AGI ANI ASI
- 69. What will an AI trillions of times smarter than us do?
- 70. “AI is likely to be either the best or worst thing ever to happen to humanity.” ~Stephen Hawking “If I had to guess at what our biggest existential threat is, it's probably AI.” ~Elon Musk “When a few people control a platform with extreme intelligence, it creates dangers in terms of power and control.” ~Bill Gates
- 71. But if we get it right… Cure diseases? Solve energy problems?
- 72. Elon Musk launches Neuralink, a venture to merge the human brain with AI
- 73. An Open Letter Research priorities for robust and beneficial Artificial Intelligence The potential benefits (of AI) are huge, since everything that civilization has to offer is a product of human intelligence; we cannot predict what we might achieve when this intelligence is magnified by the tools AI may provide, but the eradication of disease and poverty are not unfathomable. Because of the great potential of AI, it is important to research how to reap its benefits while avoiding potential pitfalls. futureoflife.org/ai-open-letter/
- 75. Open AI’s “Universe”, a platform to train AI's across games, websites and other apps
- 77. How to use ANI There are 10 building blocks in AI COGNITION SENSORY PERCEPTION MACHINE LEARNING DEEP LEARNING IMAGE ANALYSIS NATURAL LANGUAGE GENERATION NATURAL LANGUAGE PROCESSING SPEECH RECOGNITION ROBOTICS KNOWLEDGE ENGINEERING
- 78. 10 building blocks of AI COGNITIONSENSORY PERCEPTION MACHINE LEARNING DEEP LEARNING IMAGE ANALYSIS NATURAL LANGUAGE GENERATION NATURAL LANGUAGE PROCESSING SPEECH RECOGNITIONROBOTICSKNOWLEDGE ENGINEERING
- 79. 10 building blocks of AI A process to understand and represent human knowledge in data structures. Knowledge Engineering can be used in applications to solve complex problems that are generally associated with human expertise. IBM Watson Health uses engineered knowledge in combination with over 290 medical journals, textbooks, and drug databases to help oncologists choose the best treatment for their patients. KNOWLEDGE ENGINEERING
- 80. 10 building blocks of AI A physical manifestation of an AI, allowing it to interact with the physical world. Robots are mostly used to automate repetitive tasks in controlled manufacturing environments across all industries. Applications are things such as transporting goods, assembling products, quality checks, sorting objects,… Amazon employs over 45.000 logistic robots in its warehouses and Tesla has a fully automated factory. ROBOTICS
- 81. 10 building blocks of AI A process to convert speech to text, to allow AI to listen to the physical world. Speech recognition can be used to allow applications to take commands from humans, to transcribe conversations or even participate in them. Popular examples are Apple’s Siri, Google Now and Amazon’s Alexa. SPEECH RECOGNITION
- 82. 10 building blocks of AI A process to understand the meaning of words in text. Natural Language Processing can be used to analyse any text to extract topics, sentiment, meaning and ultimately to gain knowledge. It is used to read the sentiment in financial markets, to analyse product reviews, monitor social media,… NATURAL LANGUAGE PROCESSING
- 83. 10 building blocks of AI A process to express stored information in an understandable way for humans. Natural Language Generation is the opposite of NL Processing. It allows AI’s to communicate back to humans about the information they processed. It is mostly used in virtual personal assistants such as Siri, Google Now and Alexa, but also in customer service chatbots. NATURAL LANGUAGE GENERATION
- 84. 10 building blocks of AI Converts objects to text, to allow AI to see the physical world. A technology that identifies and understands what can be seen in images and video. It assigns labels to objects and situations. Best known applications of Image Analysis are facial recognition, quality controls and self-driving technology. IMAGE ANALYSIS
- 85. 10 building blocks of AI Machine Learning consists of tools and algorithms to analyze data. Machine learning can be used in a large variety of ways, such as predictions, identifying patterns, recommendations,… Today it is used to better detect diseases, recommend content, improve products and services and many other applications. MACHINE LEARNING
- 86. 10 building blocks of AI The creation of an artificial brain to handle large amounts of data. Deep learning is a branch of machine learning that focuses on algorithms to create artificial neural networks. These networks are more efficient at handling data at large scale and are used by large Internet companies to organise information, analyse and predict behaviour, improve search and products & services and more. DEEP LEARNING
- 87. 10 building blocks of AI A process to convert physical characteristics to text to provide context. Sensors measure and collect information about people, places and all kinds of objects to provide context. Examples of this are location, weather, sound, presence, volume and pressure. Sensory perception is used to predict equipment or material failures before they happen, so maintenance can be adjusted to it. SENSORY PERCEPTION
- 88. 10 building blocks of AI Turing-complete applications, where we can’t tell AI apart from other humans. Cognitive applications have a mind of their own and are able to perceive, interact, learn, act and evolve by themselves. Cognition is a missing piece that, in combination with the other building blocks, will create pure AI. COGNITION
- 89. How to use ANI There are 10 building blocks in AI COGNITION SENSORY PERCEPTION MACHINE LEARNING DEEP LEARNING IMAGE ANALYSIS NATURAL LANGUAGE GENERATION NATURAL LANGUAGE PROCESSING SPEECH RECOGNITION ROBOTICS KNOWLEDGE ENGINEERING 5 maturity phases Phase 1: In research phase but not in practical use yet Phase 2: Used in commercial applications but not accurate and consistent enough Phase 3: Accurate enough for applications, but still has technical challenges to overcome Phase 4: Has overcome the challenges in phase 3 but requires perfection Phase 5: Pure AI, indistinguishable from human intelligence
- 90. The maturity of AI in 5 phases Cognition Phase 1: In research phase but not in practical use yet Watch for promising future opportunities
- 91. The maturity of AI in 5 phases Cognition Phase 1: In research phase but not in practical use yet Phase 2: Used in commercial applications but not accurate and consistent enough Knowledge Engineering Deep Learning Image Analysis Natural Language Generation Pioneering a market before the value is created
- 92. The maturity of AI in 5 phases Cognition Phase 1: In research phase but not in practical use yet Phase 2: Used in commercial applications but not accurate and consistent enough Phase 3: Accurate enough for applications, but still has technical challenges to overcome Knowledge Engineering Deep Learning Image Analysis Natural Language Generation Speech Recognition Natural Language Processing Machine Learning Adds some value for your business today
- 93. Dealing with phase 3: AI+Humans
- 94. Dealing with phase 3: AI+Humans
- 95. The maturity of AI in 5 phases Cognition Knowledge Engineering Speech Recognition Sensory PerceptionRobotics Phase 1: In research phase but not in practical use yet Phase 2: Used in commercial applications but not accurate and consistent enough Phase 3: Accurate enough for applications, but still has technical challenges to overcome Phase 4: Has overcome the challenges in phase 3 but requires perfection Deep Learning Image Analysis Natural Language Generation Natural Language Processing Machine Learning Adds serious value for your business today
- 96. The maturity of AI in 5 phases Cognition Knowledge Engineering Speech Recognition Sensory PerceptionRobotics Phase 1: In research phase but not in practical use yet Phase 2: Used in commercial applications but not accurate and consistent enough Phase 3: Accurate enough for applications, but still has technical challenges to overcome Phase 4: Has overcome the challenges in phase 3 but requires perfection Phase 5: Pure AI, indistinguishable from human intelligence Deep Learning Image Analysis Natural Language Generation Natural Language Processing Machine Learning Doesn’t exist yet
- 97. How to use ANI There are 10 building blocks in AI COGNITION SENSORY PERCEPTION MACHINE LEARNING DEEP LEARNING IMAGE ANALYSIS NATURAL LANGUAGE GENERATION NATURAL LANGUAGE PROCESSING SPEECH RECOGNITION ROBOTICS KNOWLEDGE ENGINEERING 5 maturity phases Phase 1: In research phase but not in practical use yet Phase 2: Used in commercial applications but not accurate and consistent enough Phase 3: Accurate enough for applications, but still has technical challenges to overcome Phase 4: Has overcome the challenges in phase 3 but requires perfection Phase 5: Pure AI, indistinguishable from human intelligence And countless applications…
- 98. SENSORY PERCEPTION ROBOTICS SPEECH RECOGNITION SENSORY PERCEPTION NATURAL LANGUAGE PROCESSING NATURAL LANGUAGE GENERATION KNOWLEDGE ENGINEERING NATURAL LANGUAGE PROCESSING IMAGE ANALYSIS MACHINE LEARNING IMAGE ANALYSIS DEEP LEARNING A few ANI applications Do you repetitively deal with physical objects? e.g. moving objects, translating physical to digital, maintenance Are there interfaces you can improve? e.g. chatbots, voice commands, face detection Do you need data to be analysed? e.g. categorisation, predictions, recommendations
- 99. ARTIFICIAL INTELLIGENCE ARTIFICIAL INTELLIGENCE HOW TO START BUILDING
- 100. How to start building DEVELOPERS PLATFORMSCOMPUTING POWERDATA
- 101. Everyone needs AI talent, build a team!
- 102. How to start building DEVELOPERS PLATFORMSCOMPUTING POWERDATA
- 103. You need vast amounts of training data to perfect your algorithms
- 104. How to start building PLATFORMSCOMPUTING POWERDEVELOPERS DATA
- 105. You can start smaller than this :-)
- 106. How to start building PLATFORMSCOMPUTING POWERDEVELOPERS DATA
- 107. Platforms to get started tensorflow.org ibm.com/watson deepmind.com deeplearning.net/software/theano/
- 108. Facebook Messenger has 34.000 chatbots
- 109. Or leverage the crowd through Kaggle
- 110. Like other technology, AI is an open-source focused community
- 111. Get started on the next level of business!
- 112. ARTIFICIAL INTELLIGENCE ARTIFICIAL INTELLIGENCE WANT TO LEARN MORE? consulting@duvalunion.com
- 113. ARTIFICIAL INTELLIGENCE CHECK OUR BOOK ON DIGITAL TRANSFORMATION digitaltransformationbook.com Sold in 50 countries