In the video below, Melanie Mitchell explores the barrier of artificial understanding as a key limitation of contemporary AI. As we saw, this can induce a level of fragility into the systems that outsource decision-making to these AI tools.
Following Melanie's talk a number of members have asked for more resources on the mechanics of machine learning generally, and deep neural networks in particular. One such resource is this tutorial on SFI's Complexity Explorer website.
Michael Kearns, "The Ethical Algorithm" - June 26, 2020
The impact of a technology depends on its efficacy, growth trajectory, and implementation. Much of this nuance is obscured by media hype cycles, which tend to oscillate between presenting new technologies as panaceas or abject failures. This discussion will take stock of the present state of machine learning techniques and explore the likely pitfalls involved in using these tools to assist human decision-making. Specifically, the meeting will use complexity science as a lens to assess the current capabilities of AI, and explore the logical, behavioral, and ethical problems inherent in integrating AI into human systems. The meeting will also explore a few areas of complexity science where development could significantly affect AI in the near future.
Speaker: Melanie Mitchell
Date: Wednesday, June 24, 2020
Title: Artificial Intelligence: A Guide for Thinking Humans
In her book Artificial Intelligence: A Guide for Thinking Humans, Mitchell turns to the most urgent questions concerning AI today: How intelligent—really—are the best AI programs? How do they work? What can they actually do, and when do they fail? How humanlike do we expect them to become, and how soon do we need to worry about them surpassing us? Along the way, she introduces the dominant models of modern AI and machine learning, describing cutting-edge AI programs, their human inventors, and the historical lines of thought underpinning recent achievements. She meets with fellow experts such as Douglas Hofstadter, the cognitive scientist and Pulitzer Prize–winning author of the modern classic Gödel, Escher, Bach, who explains why he is “terrified” about the future of AI. She explores the profound disconnect between the hype and the actual achievements in AI, providing a clear sense of what the field has accomplished and how much further it has to go.
Speaker: Mike Price
Date: Thursday, June 25, 2020
Title: Density Functions and Characterizing the Scaling of Human Organizations
The typical machine learning task is either (1) regression or (2) classification with a single outcome label. Many important problems do not fall into this category, and I will discuss one such set of problems: estimating joint or conditional probability density functions. I first provide a brief overview of existing machine learning techniques for this, including generative models such as Boltzmann machines, then describe new work I am undertaking with colleagues on a conditional variational autoencoder. The application is the scaling of human organizations. Notably, our algorithm can accommodate missing data in input variables.
Speaker: Michael Kearns
Date: Friday, June 26, 2020
Time: 10:00 am US Mountain Time (12:00 pm EDT, 4:00 pm GMT)
Title: The Ethical Algorithm
Understanding and improving the science behind the algorithms that run our lives is rapidly becoming one of the most pressing issues of this century. Traditional fixes, such as laws, regulations and watchdog groups, have proven woefully inadequate. Reporting from the cutting edge of scientific research, Kearns's book The Ethical Algorithm offers a new approach: a set of principled solutions based on the emerging and exciting science of socially aware algorithm design. Michael Kearns and Aaron Roth explain how we can better embed human principles into machine code - without halting the advance of data-driven scientific exploration. Weaving together innovative research with stories of citizens, scientists, and activists on the front lines, The Ethical Algorithm offers a compelling vision for a future, one in which we can better protect humans from the unintended impacts of algorithms while continuing to inspire wondrous advances in technology.
Mike Price, "Density Functions and Characterizing the Scaling of Human Organizations" - June 25, 2020
In the video below, Michael Kearns contrasts the state of algorithmic techniques to protect privacy with those intended to prevent bias. Particular attention is paid to the concept of differential privacy and conflicting fairness criteria.
Melanie Mitchell, "Artificial Intelligence: A Guide for Thinking Humans" - June 24, 2020
In the video below, Mike Price gives the most technical talk of the meeting, focusing on a new method using conditional, variational auto-encoders to produce machine learning tools that generate joint probability density functions. Among other applications, the approach Mike laid out is particularly relevant for practitioners because it suggests a qualitatively rigorous methodology for decomposing the organizational performance attributable to leadership, policy, and culture from the performance attributable to more exogenous organizational characteristics. This is an active research project at SFI, and if anyone here today has a group working on near adjacent problems or techniques, please reach out to us at Applied@santafe.edu.