The AI Element

Episode Guests

• 
  Adrienne Mayor
• 
  Ronjon Nag

Show Notes

In 2020, and for our first episode in the new decade, we thought it would be good to continue to dig deeper into how AI has developed over time. Learning about the roots of AI, we are reminded that the north star of this field has always been what we tend to call artificial general intelligence today, intelligence that reflects the full breadth of human intelligence. This puts in context why the recent breakthroughs have been so significant, and at the same time there is still so far to go. On this week’s episode of The AI Element we are joined by two guests who share two very different histories of AI, one of its ancient roots and the other of contemporary challenges in operationalizing it for mass use.

Adrienne Mayor is a historian and research scholar at Stanford University whose recent work focuses on the earliest imaginings of AI in ancient myths. She shares some insights from ancient Greek myths like Homer’s Iliad and writings by Aristotle that show that AI and AGI have long been part of the human imagination.

Ronjon Nag reflects on the history of AI through his own experience. He’s an inventor, a distinguished Careers Research Fellow at Stanford and has pioneered a number of neural net applications since the 80s. He’s developed some of the first speech and handwriting recognition software and talks about the development of AI applications over the past 4 decades, and how though we’ve come a long way, there is still a long way to go.

00:55: Jonnie Penn, AI Historian: What not to optimize

2:00: Adrienne Mayor - Stanford University

2:17: Gods and Robots: Myths, Machines and Ancient Dreams of Technology

13:40: Talos Missile

14:13: TALOS (uniform) - Wikipedia

17:53: Harvard Divinity School

19:57: Ronjon Nag - Stanford University

20:30: Computers That Learn by Doing - Fortune Magazine

21:59: How William Shatner Changed the World - Martin Cooper, mobile phone inventor - Youtube

28:37: Google DeepMind

32:24: SpiNNaker Project

34:15: Grammatik - Wikipedia

34:38: Grammarly

36:13: The Boundaries of Humanity Project

Episode Guests

• 
  Jonnie Penn

Show Notes

In this bonus interview, historian, researcher and former TV star Jonnie Penn is joined by our Head of Public Policy and Government Relations Marc-Etienne Ouimette to talk about the history of AI. In the interview, Jonnie looks to the past to help answer a number of important questions about the future of AI. For instance, what parts of our social system do we not want to optimize? Who does this technological progress actually benefit? And how can more young people get involved in decision making processes surrounding tech?

1:19 - Jonnie Penn

1:25 - The Buried Life

1:54 - Berkman Klein Center

2:05 - MIT Media Lab

4:22 - Machines Who Think - Pamela McCorduck

23:21 - ‘Don’t Join a Union, Pop a Pill’ - Katrina Forrester

26:03 - The troubling case of the young japanese reporter who worked herself to death - Washington Post

27:14 - Germinal (novel) - Wikipedia

32:08 - The Cybernetic Brain - Andrew Pickering

42:52 - Marvin Minsky - Wikipedia

44:40 - Jonnie Penn - Twitter

Other links:

Jonnie Penn Publications - Berkman Klein Center

What History Can Tell Us About the Future of Artificial Intelligence - TEDx Talks

Episode Guests

• 
  Yoshua Bengio Co-Founder of Element AI and Scientific Director of the Montreal Institute for Learning Algorithms (MILA)

Show Notes

Can AI be used to help solve Climate Change? If so, how?

In this bonus interview, world renowned machine learning researcher Yoshua Bengio joins host Alex Shee to talk about AI’s role in solving the climate crisis. Segments of this interview was featured on our episode about sustainability, in which he shared some examples of how AI is being used to mitigate and adapt to climate change. In the full-length interview he shares his personal motivations for getting involved in climate action and goes in depth about his cross-disciplinary work to solve what he thinks is one of the world’s largest existential risks.

2:03 - Mila - Yoshua Bengio’s Lab

2:04 - IVADO - The Institute for Data Valorization

3:11 - Montreal Declaration for Responsible AI

7:55 - Tackling Climate Change with Machine Learning - Rolnick et al., arXiv

11:19 - Mila - AI for Humanity

11:30 - Mila - Climate Change

13:31 - Visualizing the Consequences of Climate Change Using Cycle-Consistent Adversarial Networks - Schmidt et al., arXiv

16:35 - GANS (Generative Adversarial Network) - Wikipedia

16:58 - Help the planet by uploading your pictures of flooded houses or buildings - Mila

19:10 - Use machine learning to find energy materials - Nature

23:20 - French-language federal leaders debate 2019 - Maclean’s

24:23 - MIT CSAIL Alliances Podcast

27: 16 - Greta Thunberg - Wikipedia

31:48 - AI Commons

Additional Links

Yoshua Bengio - Google Scholar

Climate Change: How Can AI Help? - Alexandre Lacoste

Climate Change AI

EAI Orkestrator - Optimize the use of your compute and storage resources

Episode Guests

• 
  Urvashi Vaneja Co-founder of Tandem Research
• 
  Sherif Elsayed-Ali Director of Partnerships
• 
  Yoshua Bengio Co-Founder of Element AI and Scientific Director of the Montreal Institute for Learning Algorithms (MILA)

Show Notes

This week we’re exploring if and how AI can help build a sustainable future. From solving climate change to improving health care, AI is being seen as a technology that can solve some of the world’s biggest problems. But can AI really save us? How can we be sure that AI for Good initiatives are actually helping the people they’re trying to reach? What role can, or should, AI practitioners play in finding solutions?

Urvashi Vaneja explains why we should be skeptical of AI for Good initiatives that claim to be a cure-all and she shows how to start thinking constructively about how to do better. Sherif Elsayed-Ali shows how AI for Good can help scale the positive impact of human rights organizations and also why we need to expand our current understanding of human rights. Yoshua Bengio reflects on his recent research into different ways AI can be used to mitigate against and adapt to a changing climate. Yoshua also shares why he decided to use his machine learning expertise to try and solve climate change — and how others in the machine learning community can help, too.

• AI for Good Summit
• UN Sustainable Development Goals
• Tandem Research
• AI for All: 10 Social Conundrums for India: Working Paper - Tandem Research
• Artificial Intelligence apps risk entrenching India’s socio-economic inequities
• India’s healthcare: Private vs public sector - Aljazeera
• Sherif Elsayed-Ali - Twitter
• Amnesty Tech - Twitter
• Universal Declaration of Human Rights
• Training a single AI model can emit as much carbon as five cars in their lifetime - MIT Technology Review
• Tackling Climate Change with Machine Learning - Bengio et al., arXiv
• Visualizing the Consequences of Climate Change Using Cycle-Consistent Adversarial Networks - Schmidt et al., arXiv
• Use machine learning to find energy materials - Nature

Additional Links

• AI-enabled human rights monitoring - Sherif Elsayed-Ali & Tanya O'Carroll
• Climate Change: How Can AI Help? - Alexandre Lacoste

Episode Guests

• 
  Daron Acemoglu Institute Professor at MIT
• 
  Karthik Ramakrishnan Head of Advisory at Element AI

Show Notes

Will AI take our jobs? AI’s main application is in the workplace, along all levels of skill and pay. Critics are worried that this could lead to job loss, but like any new technology that impacts work, it depends on how we implement it. How, then, can we create AI products that will enhance our capacity for work — not replace it?

MIT Institute Professor Daron Acemoglu sheds light on AI’s impact on the job market and how it could help both low-skilled and high-skilled workers alike. He also breaks down how, if we implement AI properly, it could help expand the labour market and reorganize the way we work. Karthik Ramakrishnan, Global Head of Advisory and Enablement at Element AI, talks about how we can successfully implement AI in organizations. The trick: bring workers into the process.

1:07 - Daron Acemoglu - MIT

1:20 - Why Nations Fail by Daron Acemoglu

1:20 - Automation and New Tasks: How Technology Displaces and Reinstates Labor

1:39 - Computer and Dynamo: The Modern Productivity Paradox in a Not-Too Distant Mirror

7:34 - Karthik Ramakrishnan - Twitter

10:37 - The four pillars of intelligent AI adoption - Karthik Ramakrishnan

13:25 - Building a strategic AI roadmap for your business - Karthik Ramakrishnan

18:44 - The Twenty Year History Of AI At Amazon - Forbes

20:34 - The Wrong Kind of AI? Artificial Intelligence and the Future of Labor Demand

21:49 - It’s good jobs, stupid - Daron Acemoglu

22:55 - The Future of Work? Work of the Future! - European Commission Report

Further Readings:

Artificial Intelligence, Automation and Work - Daron Acemoglu

The Four Pillars of Intelligent AI Adoption - Karthik Ramakrishnan

The Revolution Need Not Be Automated - Daron Acemoglu, Pascual Restrepo

The Wrong Kind of AI? Artificial Intelligence and the Future of Labor Demand - Daron Acemoglu, Pascual Restrepo

Next-Generation Digital Platforms: Toward Human–AI Hybrids - MIS Quarterly (PDF)

How To Become A Centaur - MIT Press

Know Your Customers’ “Jobs to Be Done” - HBR

Episode Guests

• 
  Neil Lawrence Professor of Machine Learning at the University of Sheffield

Show Notes

What is data feudalism? Should machines adapt to us or should we adapt to machines? How can we reinstate agency and control when it comes to our personal data?

In this bonus episode, Neil Lawrence, Professor of Machine Learning at the University of Cambridge, joins Element AI’s Head of Government and Public Policy Marc Etienne Ouimette to answer these questions and many more. Neil was featured in a previous episode of The AI Element, “In Data We Trust?”, in which he spoke about data trusts and data protection. In this extended interview he shares more of his thoughts on the future of AI and the growing data divide.

1:04 - The Alan Turing Institute - Professor Neil Lawrence

1:34 - Cambridge appoints first Deepmind professor of machine learning

2:07 - Jonnie Penn

2:25 - AI for social good workshop

3:02 - Isaac Asimov’s Foundation - Wikipedia

12:24 - Data Trusts could allay our privacy fears - The Guardian

23:05 - Sylvie Delacroix - Twitter

23:09 - Bottom-Up Data Trusts: Distributing the ‘One Size Fits All Approach to Data Governance - Sylvie Delacroix and Neil Lawrence

Other Readings

Data trusts: reinforced data governance that empowers the public - Element AI

Data Trusts - Inverse Probability

Inverse Probability - Neil Lawrence Blog

Talking Machines Podcast - Neil Lawrence Podcast

Episode Guests

• 
  Tanya O’Carroll Director of Amnesty Tech at Amnesty International
• 
  Alix Dunn Founder and Director of Computer Says Maybe
• 
  Jesse McWaters Financial Innovation Lead at World Economic Forum
• 
  Richard Zuroff Director of AI Advisory and Enablement at Element AI

Show Notes

AI is a powerful tool and with that power comes a great deal of responsibility. How can we be sure that we’re in control of AI systems? And what should the governance look like?

Data governance is an existing practice that covers a lot of good ground because of how integral data is to AI’s functioning. However, AI’s ability to learn and evolve over time means it will adapt to changes in its environment based on its given objective. That dynamic relationship between environment and model makes things like the design of the system and its objectives just as integral as the data the model runs on. Managing the risks of these new, dynamic systems has been widely branded as “AI Governance”.

Richard Zuroff breaks down the concept of AI governance and how it differs from data governance. Tanya O’Caroll and Alix Dunn tell us about the importance of governance in protecting human rights when building AI systems. Jesse McWaters shares his insights on AI’s impact on the financial sector and why a new form of governance will soon be necessary.

00:48 - How AI risk management is different and what to do about it - Element AI

05:07 - All the Ways Hiring Algorithms Can Introduce Bias - HBR

06:45 - The Why of Explainable AI - Element AI

07:37 - Amnesty Tech - Twitter

07:39 - Computer Says Maybe

07:52 - The Engine Room

10:18 - UN Guiding Principles on Business and Human Rights

14:18 - The Matthew Effect - Wikipedia

15:00 - Agile Ethics - Medium

17:48 - Human Rights Due Diligence

20:07 - The New Physics of Financial Services - World Economic Forum

24:30 - Consumer Financial Protection Bureau

29:00 - GDPR

Other Reading:

Putting AI Ethics Guidelines to Work - Element AI

AI-Enabled Human Rights Monitoring - Element AI

New Power Means New Responsibility: A Framework for AI Governance - JF Gagne

Podcast: Opening the AI Black Box - Element AI

Episode Guests

• 
  Ed Santow Australia’s Human Rights Commissioner
• 
  Christina Colclough Director of Platform and Agency Workers, Digitalisation and Trade at UNI Global Union
• 
  Neil Lawrence Professor of Machine Learning at the University of Sheffield

Show Notes

We don’t have enough control over our data—how it is collected, by whom, what it’s used for. We’re used to hitting “accept” to whatever agreement we need to use the online platforms, mobile apps and other digital services that run our daily lives. Yet public awareness is growing about the importance of privacy and data control. Major data breaches and scandals about the misuse of data have shown the failures of the private sector when it comes to self-regulation.

Now, governments and policymakers are stepping in with efforts to address the power imbalance between consumers and big companies when it comes to data. It’s about time — the impact of artificial intelligence could exacerbate that power imbalance, and help the data-rich get richer. Element AI’s Marc-Etienne Ouimette spoke with some of those leading the charge around taking back control of our data and the notion of data trusts — think a union, but for your data.

03:54 - NSW police may be investigated for ‘secret blacklist’ used to target children - The Guardian

08:30 - 94% of Australians do not read all privacy policies that apply to them – and that’s rational behaviour - The Conversation

09:16 - Click to agree with what? No one reads terms of service, studies confirm - The Guardian

10:57 - Up for Parole? Better Hope You’re First on the Docket - The New York Times

15:09 - The GDPR Covers Employee/HR Data and It's Tricky - Dickson Wright

18:01 - Companies are trying to test if they can make employees wear fitness trackers - Business Insider

20:03 - Silicon Valley & the Netherlands: Drivers of the future of automation - Netherlands in the USA

22:14 - Data Trusts - Neil Lawrence, inverseprobability.com

23:50 - Data trusts could allay our privacy fears - The Guardian

31:14 - Data trusts: reinforced data governance that empowers the public - Element AI

Further Reading

What is a data trust? - Open Data Institute

Data trusts: reinforced data governance that empowers the public - Element AI

Uncertainty and the Governance Dilemma for Artificial Intelligence - Dan Munro

Governing AI: Navigating Risks, Rewards and Uncertainty - Public Policy Forum

Anticipatory regulation - Nesta

Disturbing the ‘One Size Fits All’ Approach to Data Governance: Bottom-Up Data Trusts - Sylvie Delacroix & Neil Lawrence

Human Rights and Technology Issues - Australian Human Rights Commission

The Civic Trust - Sean McDonald & Keith Porcaro

Facebook’s privacy policy is longer than the US Constitution - The Next Web

Episode Guests

• 
  Nicole Rigillo Berggruen Research Fellow at Element AI
• 
  Cynthia Rudin Professor of Computer Science, Electrical and Computer Engineering, and Statistical Science at Duke University
• 
  Benjamin Thelonious Fels Founder of AI healthcare startup Macro-Eyes

Show Notes

“Explainability” is a big buzzword in AI right now. AI decision-making is beginning to change the world, and explainability is about the ability of an AI model to explain the reasons behind its decisions. The challenge for AI is that unlike previous technologies, how and why the models work isn’t always obvious — and that has big implications for trust, engagement and adoption.

Nicole Rigillo breaks down the definition of explainability and other key ideas including interpretability and trust. Cynthia Rudin talks about her work on explainable models, improving the parole-calculating models used in some U.S. jurisdictions and assessing seizure risk in medical patients. Benjamin Thelonious Fels says humans learn by observation, and that any explainability techniques need to take human nature into account.

01:11 - Facebook Chief AI Scientist Yann LeCun says rigorous testing can provide explainability

01:58 - Berggruen Institute, Transformation of the Human Program

05:34 - Judging Machines. Philosophical Aspects of Deep Learning - Arno Schubbach

06:31 - Do People Trust Algorithms More Than Companies Realize? - Harvard Business Review

08:25 - Introducing Activation Atlases - OpenAI

10:52 - Learning certifiably optimal rule lists for categorical data (CORELS) - YouTube

11:00 - CORELS: Learning Certifiably Optimal RulE ListS

11:45 - Stop Gambling with Black Box and Explainable Models on High-Stakes Decisions

16:52 - Transparent Machine Learning Models for Predicting Seizures in ICU Patients - Informs Magazine Podcast

19:49 - The Last Mile: Challenges of deployment - StartupFest Talk

24:41 - Developing predictive supply-chains using machine learning for improved immunization coverage - macro-eyes with UNICEF and the Bill and Melinda Gates Foundation

Further Reading

A missing ingredient for mass adoption of AI: trust - Element AI

Breaking down AI’s trustability challenges - Element AI

The Why of Explainable AI - Element AI

Episode Guests

• 
  Rediet Abebe Co-Founder and Co-Organizer of Black in AI, Co-Founder and Co-Organizer of Mechanism Design for Social Good
• 
  Charles C Onu Founder and AI Research Lead at Ubenwa

Show Notes

Charles C Onu is using AI to detect birth asphyxia in babies. His story is inspiring because of its impact on society and the field of healthcare (in 2016, 1,000,000 babies died from asphyxia), but also because of his humble beginnings. In this episode, Charles shows us that a passion for solving problems can help you overcome many obstacles.

Host Alex Shee also sits down with Rediet Abebe, Co-Founder and Co-Organizer of Black in AI, to expand on how others are using AI to change not just their industry, but the world.

Episode Guests

• 
  Daniel Gross Partner at Y Combinator and Head of the AI Track
• 
  Natacha Mainville Chief Innovation Officer at TandemLaunch
• 
  Jordan Fisher CEO at Standard Cognition
• 
  JF Gagné CEO of Element AI

Show Notes

Many have dystopian projections of what our future with AI will look like, but professionals working in AI see things differently. For some, our future with AI may simply mean more free time and cheaper access to quality services.

We check in with Jordan Fisher, Daniel Gross, Natacha Mainville and JF Gagné who together paint a picture of what a not-so-distant future might look, especially in retail and insurance. They may not know exactly what the year 2050 will look like, but they are hopeful.

Episode Guests

• 
  Frederic Michaud Former Director at Element AI
• 
  Oren Falkowitz CEO of Area 1 Security

Show Notes

Every touchpoint with a prospect is an opportunity to nurture that relationship, but also a potential entry point for hackers. Given that cybercriminals are more resourceful than ever, cybersecurity experts need to be just as sharp.

Oren Falkowitz is combining past experience at the NSA and US Cyber Command with AI to combat phishing attacks worldwide. In this episode, Alex Shee speaks to him and cybersecurity expert Frederic Michaud about how AI is currently being used to make businesses safer.

Episode Guests

• 
  Naomi Goldapple Former Program Director at Element AI
• 
  Chris Benson Chief Scientist for Artificial Intelligence & Machine Learning at Honeywell

Show Notes

Successful adopters of AI develop an AI-first strategy supporting all functional areas of the business: marketing, product development, customer support, sales, and beyond. What does this look like in practice? Naomi Goldapple consults with execs about AI strategy on the regular, provides some insight.

Alex also talks to Chris Benson who was hired at Honeywell to inject AI into the traditional-but-transforming manufacturing and logistics space. He shares some case studies of AI transformation and touches on the pervasive fear of job loss.

Episode Guests

• 
  JF Gagné CEO of Element AI
• 
  Natacha Mainville Chief Innovation Officer at TandemLaunch

Show Notes

As AI seeps into every industry, businesses are being forced to adapt. Old school industries may not be as lean or quick to pivot as startups, but they have access to a motherlode of funding and data. Still, red tape and outdated infrastructure may block them from the timely AI transformation they need to stay afloat tomorrow.

Alex Shee speaks with serial AI entrepreneur JF Gagné about this tension between startups and more corporate environments. Then, 15-year veteran of the insurance industry Natacha Mainville shares some real-world examples of how AI is flipping the industry on its head, forcing incumbents to keep up.

Episode Guests

• 
  Yoshua Bengio Co-Founder of Element AI and Scientific Director of the Montreal Institute for Learning Algorithms (MILA)
• 
  Daniel Gross Partner at Y Combinator and Head of the AI Track

Show Notes

Societal hype around AI is a byproduct of a few recent scientific breakthroughs — speech recognition, computer vision, natural language processing — in short, a computer’s ability to acquire human senses and mimic the human brain.

Element AI co-founder Yoshua Bengio, world-renowned professor and head of the Montreal Institute for Learning Algorithms, has been at the front lines of the Deep Learning Revolution that has enabled this kind of innovation. In this episode, he gives an overview of where the tech is actually at: how close is it to mirroring human senses?