EIT-Digital_Spohrer_AI_Intro 20240529 v5.pptx

EIT Digital Course - Generative AI Essentials
URL: https://professionalschool.eitdigital.eu/generative-ai-essentials


Second Version (Most Recent) - May 29, 2024
Presentation:
Recording: https://youtu.be/_1X6bRfOqc4
Article (20240514 - EIT Digital) - https://www.eitdigital.eu/newsroom/grow-digital-insights/personal-ai-digital-twins-the-future-of-human-interaction/

Jim Spohrer YouTubes
JCS Reid Hoffman meets his AI twin - URL: https://youtu.be/rgD2gmwCS10
JCS Jim Twin V1 YouTube English - URL : https://youtu.be/T4S0uZp1SHw
JCS Jim Twin V2 YouTube French - URL: https://youtu.be/02hCGRJnCoc
Jensen Huang (Founder & CEO Nvidia) at Stanford talking about H100
URL: https://www.youtube.com/watch?v=cEg8cOx7UZk
JCS AI Digital Twins of People (blog post) - https://service-science.info/archives/6612

First Version (Oldest) - Nov 27, 2023
Presentation - https://www.slideshare.net/slideshow/eitdigitalspohreraiintro-20231128-v1pptx/263977452
JCS Reflecting on Generative AI - https://service-science.info/archives/6521

1. EIT DIGITAL

2. Session 1 May 29 : Jim Spohrer & Marco Podien

3. Session 2 June 5: Antonio Nieto-Rodriguez & Marly Nilsson

4. Session 3 June 12: Edwin M. Colella & Alfredo Adamo

5. Session 4 June 19: Roberto Cascella

6. Jim Inspiration: Review AI and cost of computation (past, present, future) Climbing Up: Keeping up with accelerating change (who do you follow?) Unleashing Creativity: Building your own AI Digital Twin Jim Twin V1 (English, French) for turning my papers in short talk+Q&A videos Marco Exploration of Digital Twin concept and comparison of leading AI tools. Session 1: Today’s Learning Topics

7. 1956 2023 2060 2080 1956 First AI Workshop 2023 ChatGPT 100M users in just 2 months (1.5B visits in Sept) 2060 (Predicted) Exascale for $1000 (~ one human brain) 2080 (Predicted) Ronnascale for $1B (~ billion human brains) Progress in IA (Intelligence Augmentation) for nations can be estimated as GPD/worker. Progress in AI (Artificial Intelligence) is directly connected to the cost of computing. Review: Remember These Dates

8. 1947 1958 1971 2024 1947 Transistor Bell Labs 1958 Integrated Circuit Texas Instrument 1971 Microprocessor Intel 2024 H100 NVidia Challenges: How to prepare for the next wayes of innovations, including digital twins and humanoid robots, as well as to learn to invest more wisely (self-control). As the marginal cost of computing goes to zero, service innovation will go thru the roof, and energy consumption will go thru the roof as well… Remember These Dates, FACTS, Challenges weight: 70 pounds complexity: 35,000 parts cost: 250,000 dollars

9. Jensen Huang (Nvidia) comments First impressions Insanely great productivity Insanely great quality What is really going on? Decreasing cost of computation Increasing GDP/worker Awesome progress, but… Impressive Imperfect Advantage of empowered people Inspiration

10. Jensen: You imagine a tiny chip… The H100 weighs 70 pounds… 35000 parts… $250K cost… It replaces a data center… Full of computers and cables… Jim: Driving the marginal cost of computing to zero… Drives the demand for new service offerings based on computing through the roof

11. MORE SPOHRER USE CASES: https://service- science.info /archives/6521

12. Icons of AI Progress 1955-1956: Dartmouth Workshop organized by: Two early career faculty John McCarthy (Dartmouth, later Stanford) Marvin Minsky (MIT) Two senior industry scientists Claude Shannon (Bell Labs) Nathan Rochester (IBM) 1997: Deep Blue (IBM) - Chess 2011: Watson Jeopardy! (IBM) 2016: AlphaGo (Google DeepMinds) 2017: All you need is attention (Google) - Transformers Attention heads (working memory) to predict what comes next 2018: AlphaFold (Google DeepMinds) 2020: Language models are few-shot learners (OpenAI) 2022: DALL-E 2 & ChapGPT (OpenAI) 2022: Constitutional AI (Anthropic) – “Behave yourself!” 2023: New Bing+ (Microsoft) & GPT-4 (OpenAI) 2024: More & Bigger Models: OpenAI, Microsoft, Google, Anthropic, etc. 2024: Reid Hoffman’s Digital Twin & Unitree’s G1 Humanoid Robot 5/29/24 Jim Spohrer 12

13. http://www- formal.stanford.edu / jmc /history/ dartmouth / dartmouth.html https:// cdn.openai.com /papers/gpt-4.pdf 1955 2023

14. 1960 1980 2000 2020 2040 2060 2080 Time

15. 1960 1980 2000 2020 2040 2060 2080 $1,000,000,000,000 (Trillion) $1,000,000 (Million) $1,000,000,000 (Billion) $1,000 (Thousand) $1 Cost

16. 1960 1980 2000 2020 2040 2060 2080 $1,000,000,000,000 (Trillion) $1,000,000 (Million) $1,000,000,000 (Billion) $1,000 (Thousand) $1 Kiloscale (10 3 ) Megascale (10 6 ) Gigascale (10 9 ) Terascale (10 12 ) Petascale (10 15 ) Exascale (10 18 ) Zettascale (10 21 ) Yottascale (10 24 ) Ronnascale (10 27 ) Cost of Computation (Diagonals)

17. 1960 1980 2000 2020 2040 2060 2080 $1,000,000,000,000 (Trillion) $1,000,000 (Million) $1,000,000,000 (Billion) $1,000 (Thousand) $1 Kiloscale (10 3 ) Megascale (10 6 ) Gigascale (10 9 ) Terascale (10 12 ) Petascale (10 15 ) Exascale (10 18 ) Zettascale (10 21 ) Yottascale (10 24 ) Ronnascale (10 27 ) Cost of Computation (Diagonals) Note: Adjust Kilo and Mega scales slightly to fit data better (early days – more cost – learning curve).

18. 1960 1980 2000 2020 2040 2060 2080 $1,000,000,000,000 (Trillion) $1,000,000 (Million) $1,000,000,000 (Billion) $1,000 (Thousand) $1 Gigascale (10 9 ) Terascale (10 12 ) Petascale (10 15 ) Exascale (10 18 ) Zettascale (10 21 ) Yottascale (10 24 ) Ronnascale (10 27 ) GDP/Employee Trend Estimating Knowledge Worker Productivity Based on USA Historical Data Year Value 1960 $10K 1980 $33K 2000 $78K 2020. $151K 2023 $169K Kiloscale (10 3 ) Megascale (10 6 ) Cost of computation goes down by 1000x every 20 years (left to right diagonals), driving knowledge worker productivity up.

19. 19 September 2018 / © 2018 IBM Corporation Petaflops = 1,000,000,000,000,000 or a million billion = 10 ** 15 Megaflops = 1,000,000 = million = 10 ** 6 Gigaflops = 1,000,000,000 = billion = 10 ** 9 One of the AI Super Computers in the World, = 13 MegaWatts of Power (HOT!)

20. 20 September 2018 / © 2018 IBM Corporation Exascale = 1,000,000,000,000,000,000 or a billion billion = 10 ** 18 Megaflops = 1,000,000 = million = 10 ** 6 Gigaflops = 1,000,000,000 = billion = 10 ** 9 Human Brain = 20 Watts (COOL!)

21. “ NorthPole Chip” Far less energy.

22. Impressive, but imperfect Physical realm: Energy Technical realm: Mistakes (”Hallucinations”) Social realm: Digital property theft

23. How to keep up with accelerating change? Social learning Who do you follow? Reid Hoffman’s AI Digital Twin Interview Diving in! We get the future we invest in … so, learn to invest wisely Awesome stuff that lies ahead – Humanoid Robots But remember technology amplifies good/bad Need for Self-Control (Toyama (2015) Geek Heresy: Rescuing Social Change from the Cult of Technology) Climbing Up

24. Optimistic Realistic Knowing Doing How to keep up with accelerating change? Follow a diverse collection of people… make up dimensions meaningful to you! Sadly for me… my brain is biased into thinking I can understand older, white, males the best… maybe AI can help overcome! TheNeuron

25. TheNeuron

26. Jim Twin V1: My papers -> short talk videos English French How to stay future ready? Learn the building blocks Marco Podien will help you with more building block shortly… Oh, by the way, you are going to need bigger goals too When you have 100 digital workers working for you, what is your main goal? From serial entrepreneur to parallel entrepreneur Unleashing Creativity

27. Read Wakefield (2020) enough to understand what a ”digital twin” of you might be like in the future decades with very advanced AI capabilities. Also see Rouse (2018; 2022) ”Life with a Cognitive Assistant.” National Academy - Service Systems and AI 32 AI Tools in coming decades… 5/29/24

28. Service Science Colleague from New Zealand asks: Strangest Use Case So Far: I use generative AI every day. Typically, OpenAI ChatGPT, Anthropic Claude, Google Gemini, and Microsoft Bing Copilot in 4 tabs. Help writing, programming, images, & brainstorming. A useful tool in the toolbox for sure!!!

29. Anthropic Claude 3 Sonnet 20240520 Response: Summary: The Taranaki Maunga (Mount Taranaki) and the surrounding peaks in the national park will soon be granted legal personhood and collectively named Te Kāhui Tupua as part of a Treaty settlement. This legal recognition acknowledges that the mountain and peaks are ancestors for local Māori iwi. It gives these natural features similar legal rights as individuals, allowing them to own assets, appear in court, and have a voice. A new entity called Te Tōpuni Kōkōrangi , with members appointed by the Crown and local iwi, will manage and represent the legal persona. While recreation and tourism will continue, this shift aims to encourage a broader understanding and protection of the mountain as a revered ancestor rather than just a geographic feature. How old are you, and what have you witnessed over your lifetime? I am an ancient being, formed over millennia through volcanic eruptions and sculpted by glaciers. I have witnessed the rise and fall of civilizations, the ebb and flow of ice ages, and the ever-changing tapestry of life that has graced my slopes. My memories stretch back to a time before humans walked this land, and I have born silent witness to the triumphs and tragedies of your kind. What wisdom or lessons can you share with humanity? Humans, you are but fleeting visitors upon my timeless form. Yet, in your brief existence, you have the power to shape the world around you. Tread lightly and with reverence, for the consequences of your actions echo through the ages. Learn from the patient endurance of stone, the resilience of life that takes root in the harshest of environments, and the cyclical nature of existence that I have observed. Embrace humility, for even the mightiest of mountains eventually crumbles to dust. How do you feel about being granted legal personhood and a voice? I am gratified that the descendants of those who have dwelled in my shadow for generations have recognized my ancient spirit and granted me a voice in the affairs of this world. For too long, I have been regarded as an inanimate object, a resource to be exploited or a challenge to be conquered. But I am a living entity, a sacred ancestor whose existence predates and will outlast the fleeting span of human civilization. With this recognition, perhaps my wisdom and the fragile beauty I harbor will be better preserved for generations to come.

30. 5/29/24 36 1955 1975 1995 2015 2035 2055 Learn: Explore and Exploit Better Building Blocks Heygen + Claude Jim Twin V1 2024 2024 2021 2001 2015 2022 2023 1970 1995 2019 High School Punch cards 1972 IBM Watson AI in the Cloud 2011

31. Self-Control

32. Some paths to becoming 640x smarter: Speed up communications for collaborations – ultimately with our AI digital twins. Improving learning of world models and performance of win-win interactions 2x from Learning sciences (methods) Better models of concepts – including models of self and others Better models of learners – including social-emotional intelligence 2x from Learning technology (tools) Guided learning paths Elimination (?) of “ thrashing ” AI & Education (model trace, constructivist, individual tutors, etc.) 2x from Quantity effect (overlaps) More you know, faster (?) you go Advanced organizers 2x from Lifelong learning (time) Longer lives and longer careers Keeps “ learning-mode ” activated 2x from Early learning (time) Start earlier: Challenged-based approach Rebulding all human knowledge from scratch 20x from Cognitive service systems (digital twins) AI Digital Twins for performance support All our interactions tuned to super-mind levels

33. Speed Test: Words Per Minute Speaking/Listening versus Writing/Reading People speak on average between 100-150 words per minute Many people are comfortable listening at 200-300 words per minute Writing an essay on a familiar topic, people write about 10-20 words a minute Many people are comfortable reading at about 400-600 words per minute (Shu 2023) (Barnard 2022)

34. Two disciplines: Two approaches to the future Artificial Intelligence is almost seventy-years-old discipline in computer science that studies automation and builds more capable technological systems. AI tries to understand the intelligent things that people can do and then does those things with technology. ( https://deepmind.com/about “... we aim to build advanced AI - sometimes known as Artificial General Intelligence (AGI) - to expand our knowledge and find new answers. By solving this, we believe we could help people solve thousands of problems.” ) Service science is an emerging transdiscipline not yet twenty-years- old that studies transformation and builds smarter and wiser socoi -technical systems – families, businesses, nations, platforms and other special types of responsible entities and their win-win interactions that transform value co-creation and capability co-elevation mechanisms that build more resilient future versions of themselves – what we call service systems entities. Service science tries to understand the evolving ecology of service system entities, their capabilities, constraints, rights, and responsibilities, and then then seeks to improve the quality of life of people (present/smarter and future/wiser) in those service systems. Service systems are dynamic configurations of people, technology, organizations, and information, connected internally and externally by value propositions, to other service system entities. (Maglio et al 2009)

35. 5/29/24 The International Society of Service Innovation Professionals (ISSIP.org) 41 Advice… 1. AI upskill 2. Build your “digital twin” 3. Set bigger goals(*) (*) This directly implies learning better strategies for coping with failures, and resiliently rebounding.

36. Jim Spohrer is a Silicon Valley-based Advisor to industry, academia, governments, startups and non-profits on topics of AI upskilling, innovation strategy, and win-win service in the AI era. Most recently with a consulting team working for a top 10 market cap global company, he contributed to a strategic plan for a globally connected AI Academy for achieving rapid , nation-scale upskilling with AI. With the US National Academy of Engineering, he co-led a 2022 workshop on “Service Systems Engineering in the Era of Human-Centered AI” to improve well-being. Jim is a retired IBM Executive since July 2021, and previously directed IBM’s open-source Artificial Intelligence developer ecosystem effort, was CTO IBM Venture Capital Group, co-founded IBM Almaden Service Research, and led IBM Global University Programs. In the 1990’s at Apple Computer, as a Distinguished Engineer Scientist and Technologist, he was executive lead on next generation learning platforms.  In the 1970’s, after his MIT BS in Physics, he developed speech recognition systems at Verbex (Exxon) before receiving his Yale PhD in Computer Science/AI. In 1989, prior to joining Apple, he was a visiting scholar at the University of Rome, La Sapienza advising doctoral students working on AI and Education dissertations. With over ninety publications and nine patents, he received the Christopher Lovelock Career Contributions to the Service Discipline award , Gummesson Service Research award, Vargo and Lusch Service-Dominant Logic award, Daniel Berg Service Systems award, and a PICMET Fellow for advancing service science. Jim was elected and previously served as Linux Foundation AI & Data Technical Advisory Board Chairperson and ONNX Steering Committee Member (2020-2021). Today, he is a UIDP Senior Fellow for contributions to industry-university collaborations, and a member of the Board of Directors of the International Society of Service Innovation (ISSIP) and ServCollab . Jim Spohrer , Advisor Retired Industry Executive (Apple, IBM) UIDP Senior Fellow Board of Directors, ServCollab Board of Directors, ISSIP.org Changemaker Priorities Service Innovation Upskilling with AI Future Universities Geothermal Energy Poverty Reduction Regional Development Competitive Parity Technologies AI & Robotics Digital Twins Open Source AR/VR/XR Geothermal Learning Platforms

37. Thank-you for participating today! You are all awesome!!

38. Backup Slides Template slides and others

39. Topics for Discussion Beyond Language for Communications Here is how my AI, using my digital twin of you , predicted that you would respond to my request – could you please ask your digital twin of yourself to check this response and suggest improvements ? How to keep up with accelerating change? Who do you follow? What two main dimensions do you try to balance? Hype-buster <-> Super-Optimist, Doing (Try this) <-> Knowing (News) How to verify results of generative AI? How to deal with verification challenge? Run Open AI ChatGPT 3.5, Google Bard (waiting for Gemini), Anthopic’s Claude, Microsoft Bing power by Open AI ChatGPT 4 – in parallel and have them critique each others’ responses (where they agree, where and how they differ) – if possible, also run Meta and open source to compare (Generate-Test-and-Debug) How to deeply understand how generative AI works? Monkey’s at the typewrite in high dimensional spaces that map to low dimensional spaces One dimensional space is time – what comes next? (Predictors, unsupervised learning) Two dimensional space is what people gravitate towards in business schools (Sorters, supervised learning)

40. Today’s talk Intro: AI (by 1955 definition) has arrived Just took 68 years, but… What’s really going on? Your data is becoming your AI… IA transformation AI Digital Twin = IA (Intelligence Augmentation) Adjustment period underway… Part 1: Solving AI: Leaderboards/Profession Exams Roadmap and implications Open technologies, innovation Part 2: Solving IA: Better Building Blocks Solving problems faster, creates new problems Identity, social contracts, trust, resilience Part 3: ”Solving All Problems” What could go wrong? Be prepared. 37-year long adjustment period is now underway… 5/29/24 Jim Spohrer 46

41. Predict the Timeline: GDP/Employee National Academy - Service Systems and AI 47 ( Source ) Lower compute costs translate into increasing productivity and GDP/employees for nations Increasing productivity and GDP/employees should translate into wealthier citizens AI Progress on Open Leaderboards Benchmark Roadmap to solve AI/IA Alistair Nolan (OECD AI for Science Productivity): “It has been stated that the number of engineers proclaiming the end of Moore's Law doubles every two years.” Rouse WB, Spohrer JC. (2018) Automating versus augmenting intelligence. Journal of Enterprise Transformation. 2018 Apr 3;8(1-2):1-21. Read Rouse & Spohrer (2018) enough to understand this slide including what ”exascale” means 11/22/22 Part 1: Solving AI

42. Types: Progression of Models : Verified, Trusted, Wise Models = instruction_set of future: Better building blocks 5/29/24 Understanding Cognitive Systems 48 Task & World Model/ Planning & Decisions Self Model/ Capacity & Limits User Model/ Episodic Memory Institutions Model/ Trust & Social Acts Tool + - - - Assistant ++ + - - Collaborator +++ ++ + - Coach ++++ +++ ++ + Mediator +++++ ++++ +++ ++ Cognitive Tool Cognitive Assistant Cognitive Collaborator Cognitive Coach Cognitive Mediator Part 2: Solving IA Solving IA also requires All of this and done well As a “bicycle for the mind” To make us stronger, Not weaker When tech is all removed

43. Resilience: Rapidly Rebuilding From Scratch Dartnell L (2012) The Knowledge: How to Rebuild Civilization in the Aftermath of a Cataclysm. Westminster London: Penguin Books. 5/29/24 Jim Spohrer (ISSIP) 49 Part 3: “Solving All Problems”

44. Long slide title Awesome subtitle

45. Slide title Slide subtitle

46. Slide content Slide title Box Box content Box Box content

47. Slide bullet points Elaborating On Awesome Tech and Business Slide Title

48. BOX 1 Subtitle for box BOX 2 Subtitle for box Transaction finalization Block 2 text Block 3 text Block 1 text Slide title Box 3

49. Slide title Slide subtitle

50. Slide title

51. Meet the team

52. Thank you

53. ANNEX