![[How] can we build a Human AI?
Alex ‘Sandy’ Pentland, MIT:
“The big question that I'm asking myself these days is how
can we make a human artificial intelligence? Something
that is not a machine, but rather a cyber culture that we
can all live in as humans, with a human feel to it. I don't
want to think small—people talk about robots and stuff—I
want this to be global. What would happen if you had a
network of people where you could reinforce the ones that
were helping and maybe discourage the ones that weren't?
That begins to sound like a society or a company”.
The Human Strategy. www.thehumanstrategy.mit.edu](https://image.slidesharecdn.com/2emmanuel-180510171209/85/EMBD2018-Humanos-y-maquinas-Un-futuro-con-inteligencia-artificial-20-320.jpg)
EMBD2018 | Humanos y máquinas: Un futuro con inteligencia artificial.
- 2. Humans and Machines: A Future with Artificial Intelligence? Emmanuel Letouzé, PhD Visiting Scholar, MIT Media Lab Connection Science Fellow, MIT Director & co-Founder, Data-Pop Alliance Program Director, OPAL Project 3rd Encuentro Mundial Big Data Bogotá, April 25th 2017
- 3. Outline 1. A short introduction and 2 big questions 2. Concepts and context: AI 101 in 2018 3. AI, creative destruction, and destructive creation 4. A BrAIgther vision: A human strategy for a “Human AI” 5. Short-term milestone: Open Algorithms
- 4. Outline 1. A short introduction and 2 big questions 2. Concepts and context: AI 101 in 2018 3. AI, creative destruction, and destructive creation 4. A BrAIgther vision: A human strategy for a “Human AI” 5. Short-term milestone: Open Algorithms
- 6. 1. Are we ever going to be enslaved by robots? Lose our jobs? Have a machine-driven nuclear war? Or get all very rich? Get paid to do no work? Hopefully and probably none of the above. But… 2. Can we envision and build a better world where humans and machines cooperate – a ”Human AI” or “human-machine ecology”? What would it feel like, look like, and take?
- 7. Outline 1. A short introduction and 2 big questions 2. Concepts and context: AI 101 in 2018 3. AI, creative destruction, and destructive creation 4. A BrAIgther vision: A human strategy for a “Human AI” 5. Short-term milestone: Open Algorithms
- 8. What is “Artificial Intelligence” ?Artificial intelligence—broadly-- is the simulation of human intelligence processes by computer systems, especially artificial neural networks (ANNs) inspired by the biological neural networks that constitute animal brains, which can "learn" (i.e. progressively improve performance on) through iterations and feedback. Basically it’s algorithms that learn to automate parts or all of tasks, and the machines they power. (It’s also what has not been invented yet) Input(s) Hidden layer(s) Output(s)
- 9. The basics of AI is learning through many feedbacks 1. Try to guess / recognize. Right or Wrong? 2. Correct: +1. Reward! 3. Incorrect: -1. Penalty. 4. Repeat and learn through a feedback loop. (The) machine (is) learning!
- 11. Is AI some new (black) magic? No…but… No… 1.It is at least 60+ years old. 2.The big difference is data. More on that later. 3.It still generalizes poorly. It has no sense of context. It is still pretty stupid. 4.We are far from general AI. 5.Humans are still in control (for better or worse) …but… 1. The (good) magic / core of the current AI is the credit assignment function to encourage and reinforce neurons / functions that help the most achieve the goal (and reverse if not)
- 12. Ian Morris, chart of all human history Reflecting on and learning from history
- 13. Outline 1. A short introduction and 2 big questions 2. Concept and context: AI 101 in 2018 3. AI, creative destruction, and destructive creation 4. A BrAIgther vision: A human strategy for a “Human AI” 5. Short-term milestone: Open Algorithms
- 16. “I’m not too worried about machines replacing cartoonists,” the artist R. Kikuo Johnson says, about his cover for the Money Issue. Johnson may have switched from drawing with ink, brushes, and paper to using a stylus and a digital tablet, but he isn’t worried that computers will take over the rest of his cartooning process. “When robots are advanced enough to be neurotic, then maybe I’ll be concerned,” he said, “though I don’t think too many of us choose this field for job security, anyway.”
- 19. Outline 1. A short introduction and 2 big questions 2. Concept and context: AI 101 in 2018 3. AI, creative destruction, and destructive creation 4. A BrAIgther vision: A human strategy for a “Human AI” 5. Short-term milestone: Open Algorithms
- 20. [How] can we build a Human AI? Alex ‘Sandy’ Pentland, MIT: “The big question that I'm asking myself these days is how can we make a human artificial intelligence? Something that is not a machine, but rather a cyber culture that we can all live in as humans, with a human feel to it. I don't want to think small—people talk about robots and stuff—I want this to be global. What would happen if you had a network of people where you could reinforce the ones that were helping and maybe discourage the ones that weren't? That begins to sound like a society or a company”. The Human Strategy. www.thehumanstrategy.mit.edu
- 21. 1. Key principle 2. Key features 3. Key requirements Taking the key insights of AI especially •role of data •credit assignment function reinforcing “neurons” that work (teams, groups, policies) apply this general framework to entire societies Leveraging human-machine complementarities through cooperation: • humans do the strategy and oversight and machines do the tactics and bookkeeping. • Humans + Machines >> Humans or Machines. (E.g. chess competition) • New jobs will be created (e.g. machine prison guards • Good data on the system’s functioning and performance • Good feedback and response systems (i.e. “human or society in the loop”) • Some general agreement on inputs (facts) and outputs (goals) • Sufficient human skills to adapt, implement and oversee Vision of a “Human AI”
- 22. Key contemporary challenges for a Human AI1. Some powerful agents have strong incentives for this not to work (e.g. economic and political monopolies benefit from status quo) 2. Most societies / countries currently lack the appropriate data sources, capacities and culture for this 3. There is widespread (and growing?) digital and analog segregation that feed on and fuel distrust, disdain, echo chambers, alternative facts narratives, etc., hampering cooperation and consensus building 4. We know AI can and has been used to nurture 3. (cf Facebook newsfeed; Amazon Prime..)
- 23. Humans are and should remain meddling socio- political animals experimenting and working with machines “Automation may ultimately not take our jobs by being better than us at performing them, but by us being worse and worse at effecting even the most simple tasks, such a reading a document for ten minutes. Robots may also not replace our lovers and friends but not by being as good as them and giving warmth and companionship, but by technology making us worse at dealing with real people. Losing to automation may be a much more banal process that we conjecture it to be. Machines may not need to catch up, if we drop the ball as fast as we seem to be dropping it. ” Manuel Cebrian and Iyad Rahwan, MIT
- 24. Outline 1. A short introduction and 2 big questions 2. Concept and context: AI 101 in 2018 3. AI, creative destruction, and destructive creation 4. A BrAIgther vision: A human strategy for a “Human AI” 5. Short-term milestone: Open Algorithms
- 25. “….move from the feared tyranny of data and algorithms to a data-enabled model of democratic governance running against tyrants and autocrats, and for the people.” Pillars of positive data-enabled positive disruption
- 26. “Open Algorithms”: a bold new paradigm -and project
- 27. 2010 2011 2012 2013 2014 2015 Origin of OPAL: the (Big) Data (& AI) Revolution(s),experimentations, expectations, …and questions 2016 2017 How can societies ‘leverage’ human behavioural data collected by private companies for public good in a sustainable, safe, ethical ? 2018 Data Challenges arranged by private companies like Orange, Telefónica and BBVA that gave teams of researchers access to aggregated anonymized data, revealing and spurring a thirst for these kinds of insights about human societies
- 28. Building a 1.5 M Euros Minimum Viable Product with 2 tracks, 2 phases, in Senegal and ColombiaMinimum Viable Technology 1 Minimum Viable Technology 2 Minimum Viable Governance 1 Minimum Viable Governance 2 Q4 2017 Q1 2018 Q2 2018 Q3 2018 Q4 2018 A technology track with an open platform and open algorithms running on telecom operators’ data that never leave the companies’ servers A governance track, with a participatory design (FUTs), an orientation and ethics committee (CODE), and capacity building activities
- 29. OPAL Consortium Founding organizations Local partners Donors OPAL is a unique Public-Private-People- Partnership aiming to build the foundations of a Human AI
- 30. OPAL’s future human-AI ecosystem / ecology Certified open algorithms developed by developers are sent and run on the servers of partner private companies, behind their firewalls Partner private companies (here a telecom operator) allow OPAL to access its servers through a secured platform A governance system including a Council for the Orientations of Development and Ethics (CODE) ensures that the algorithms and use cases are ethically sound, context relevant, etc.; users benefit from capacity building activities Key indicators derived from private sector data such as population density, poverty levels, or mobility patterns, feed back into use cases in various public policy and economic domains
- 31. Thank you [email protected] 3rd Encuentro Mundial Big Data Bogotá, April 25th 2017