AI and Speed

By TAP Guest Blogger

Posted on November 28, 2018


Share

The Digital Life Initiative (DLI) at Cornell Tech explores the societal tensions arising from existing and emergent digital technologies. In September, the DLI team hosted the Speed Conference, a two-day event aimed at examining the societal and ethical challenges inherent in the radical processing speeds of artificial intelligence (AI). The DLI team shares the following highlights from this conference.

 

Speed Conference began with a provocation: How do we sustain meaningful human and societal oversight of AI given radically different processing speeds? Over the course of the two-day conference, twenty speakers from diverse fields presented suggestions for addressing these issues.

 

Along with scale and complexity, speed is one of the defining problems of algorithmic oversight. Intelligent systems and human actors operate at vastly different speeds, and these differences present challenges for managing and responding to algorithmic decisions. Despite speed’s importance, discussions about the topic have been isolated in areas such as robotics, finance, warfare, and online communication.

 

Speed Conference, held at Cornell Tech on September 28 and 29 2018, brought together nearly a hundred academics and practitioners to identify common themes and potential solutions in areas of speed, AI, and algorithmic oversight. Six panels of experts presented their research in areas of autonomous vehicles, warfare, information security, labor and manufacturing, content moderation, and finance. Panel discussions invited thought-provoking questions from the audience.

 

Panel 1 | Fast, Cheap, and Out of Control

 

The first panel served as an introduction to the conference. Its first speaker, Cornell tech professor and DLI director Helen Nissenbaum, began the discussion with an overview of the state of AI and the common concerns that have arisen from its use, such as accountability and oversight. Speakers Wendy Ju and Andrea Matwyshyn presented test cases and “fables” as provocations for thinking about the present and future of AI and speed.

 

The speakers also reflected on the importance of accessible AI research, the unpredictability of technology, and the value of alternate points of view in thinking about AI. The panel concluded with a discussion of the unmentioned point in the panel title: “cheap”. Panelists noted that AI may force changes in supply chain that can reduce the price of goods and services.

 

Panel 2 | Content Moderation

 

The second panel focused on speed in the context of content moderation, communication, and news. Speakers James Grimmelmann, Kate Klonick, Jason Farman, and Mike Ananny presented their ideas on how speed distorts the spread of ideas, and how speed is vital to preventing the spread of harmful information. But with the importance of speed comes the importance of lack of speed. The speakers also focused on the utility of slowness and pauses in the production, transmission, and consumption of news and other content.

 

The importance of context was an important theme throughout this panel. Context determines whether we consider something to be slow or fast, impacts how we interpret fast-moving memes, and helps us quickly decide whether a person or event is newsworthy. Context is also vital in determining how to regulate or govern these fast-moving systems.

 

Panel 3 | Warfare and Policing

 

The third panel of the conference explored issues of speed in warfare and autonomous weapons systems. Speakers Frank Pasquale, Peter Asaro, and Mary “Missy” Cummings brought both theoretical and practical perspectives to their analysis of the problems and potential of AI in warfare. They also present a variety of frameworks for automation in military contexts, and explore the points in these frameworks that are impacted by speed.

 

It is between the theoretical and practical considerations that one of Speed Conference’s most interesting debates arises. Legal and moral perspectives on the development of autonomous weapons are concerned with the potential for AI-driven conflict, the accountability and agency assigned to weapons systems, and the importance of considering human dignity when autonomous decisions are made. These perspectives consider the history of speed and its relation to military power, and promote the need for careful regulation of autonomous weapons systems.

 

On the other hand, consideration of autonomous systems as they are currently used in military contexts – rather than how they might theoretically be used – offers a different picture of the future. Instead of complicating moral judgments and causing errors, automation is used in a way that allows even more deliberation of weapon use and allows for better (and slower) decision-making. While theoretical perspectives concern themselves with the ethics of military applications of AI, practical frameworks point to places where autonomous technologies could have immediate utility.

 

Panel 4 | With All Deliberate Speed (security)

 

The conference’s fourth panel explored the broad issues of information security and Internet infrastructure. Panelists Steven Jackson, Fenwick McKelvey, Paul Ohm, and Jonathan Frankle discussed their work on internet infrastructure and security, and how speed impacts these areas. To a great extent, the speakers on this panel all discussed the contrast between how speed and delay are seen and the impact they actually have.

 

For example, while many stories about speed focus on the desirability of speedy technologies, and neglect the realities of vastly differing timescales, system failures, and ignored human effort. Similarly, providers of internet services and applications focus on speed as happiness (in contrast to delay as frustration), and promote differential speed as a form of fairness, both of which can mask the reality of net (non) neutrality and government censorship. Excitement about speed also overlooks instances in which delay or inefficiency might be desirable in the context of the Internet, security, and infrastructure.

 

Panel 5 | Labor and Manufacturing

 

The fifth panel focused on issues of labor and markets. Speakers Argyri Panezi and Rory Van Loo discussed speed in relation to the gig economy and online purchasing behavior. Their talks described the ways in which speed is a requirement of the modern economy: we expect goods and services to be delivered quickly and seamlessly, and for the best possible price. But what does this mean for the people who provide those services? And what happens to companies when the algorithms that deliver recommendations to consumers favor one firm over another?

 

One of this panel’s unifying themes was the extreme disconnect between law and current market or technological practice. Laws currently in place to protect workers are not always effective in the gig economy, and the regulations that protect consumers from unfair price discrimination may not be up to the task of dealing with AI-driven markets. Both speakers noted the need for careful thinking about the implications for speed for market turbulence, labor, and consumer protection.

 

Panel 6 | Finance

 

The sixth and last panel reported on issues relating to speed and financial markets. Simona Abis, Katie Brennan, Larry Tabb, and Michael O’Connor described their work in this area from both theoretical and practical perspectives. One theme of this panel was the similarities and differences between humans and machines as actors in financial markets. While funds controlled by automated systems and by human investors tend to perform differently, trading algorithms also exhibit surprisingly human-like characteristics, such as predatory behavior, raising questions of accountability and agency.

 

Another theme of the Finance panel was that, in many cases, increases in speed were not the most important factor influencing financial markets, even where algorithmic trading is used. As computational speed and capacity increases – often toward its theoretical limits – factors other than speed become more important. Data is one of those factors: increasing quantity and quality of financial data can help improve financial models, overcoming many of the problems introduced by high-frequency trading. Another important factor is the deliberate introduction of inefficiency. Whether introduced as blockchain proof-of-work or as intentional delays in the execution of stock trades, inefficiency can create fairer and more trustworthy markets.

 

Common Themes

 

Speakers throughout the two days of Speed Conference addressed several overarching themes. First was the issue of regulation: how should speedy technologies be governed, and how should people be protected from their effects? Some speakers noted that internal regulation by technology companies has been effective, but what happens when external laws are needed? And what should be done when the speed of regulation itself is fundamentally slower than that of technological development?

 

Another theme is that of accountability and agency. Perspectives from autonomous warfare suggest that technologies should not (or cannot) be thought of as having agency, and therefore cannot be considered accountable. However, perspectives from finance suggest that the complex behavior demonstrated by high-frequency trading algorithms serve as an argument for expanding definitions of agency to include non-human systems.

 

A final theme is the importance of context in considering the effects of speed. The human effort that allows for technologies to be (or appear to be) fast, the technical and legal infrastructure that support speed, the history and varying timescales that help us define speed, and the differential impact of speed on different groups of people all need to be considered. Careful consideration of these contexts is vital when deliberating the future of speed.

 

Read the full Speed Conference Report.

 

View the video recordings from the Speed Conference:

 

This conference report was written by Jessie G Taft, Research Initiative Coordinator with the Digital Life Initiative at Cornell Tech.

 

Note: The “Speed Conference” was supported by a gift from Microsoft Corporation. Additionally, the Technology | Academics | Policy (TAP) website is sponsored by Microsoft. Microsoft respects academic freedom, and works to enable dialogue on the most critical technology policy issues being debated. While Microsoft provides administrative and financial support for the site’s platform and content, there is no payment made to scholars for appearing or blogging on the site.

 


Share