Building Trust in AI: A Landscape Analysis of Government AI Programs

Paper by Susan Ariel Aaronson: “As countries around the world expand their use of artificial intelligence (AI), the Organisation for Economic Co-operation and Development (OECD) has developed the most comprehensive website on AI policy, the OECD.AI Policy Observatory. Although the website covers public policies on AI, the author of this paper found that many governments failed to evaluate or report on their AI initiatives. This lack of reporting is a missed opportunity for policy makers to learn from their programs (the author found that less than one percent of the programs listed on the OECD.AI website had been evaluated). In addition, the author found discrepancies between what governments said they were doing on the OECD.AI website and what they reported on their own websites. In some cases, there was no evidence of government actions; in other cases, links to government sites did not work. Evaluations of AI policies are important because they help governments demonstrate how they are building trust in both AI and AI governance and that policy makers are accountable to their fellow citizens…(More)”.

The Law of AI for Good

Paper by Orly Lobel: “Legal policy and scholarship are increasingly focused on regulating technology to safeguard against risks and harms, neglecting the ways in which the law should direct the use of new technology, and in particular artificial intelligence (AI), for positive purposes. This article pivots the debates about automation, finding that the focus on AI wrongs is descriptively inaccurate, undermining a balanced analysis of the benefits, potential, and risks involved in digital technology. Further, the focus on AI wrongs is normatively and prescriptively flawed, narrowing and distorting the law reforms currently dominating tech policy debates. The law-of-AI-wrongs focuses on reactive and defensive solutions to potential problems while obscuring the need to proactively direct and govern increasingly automated and datafied markets and societies. Analyzing a new Federal Trade Commission (FTC) report, the Biden administration’s 2022 AI Bill of Rights and American and European legislative reform efforts, including the Algorithmic Accountability Act of 2022, the Data Privacy and Protection Act of 2022, the European General Data Protection Regulation (GDPR) and the new draft EU AI Act, the article finds that governments are developing regulatory strategies that almost exclusively address the risks of AI while paying short shrift to its benefits. The policy focus on risks of digital technology is pervaded by logical fallacies and faulty assumptions, failing to evaluate AI in comparison to human decision-making and the status quo. The article presents a shift from the prevailing absolutist approach to one of comparative cost-benefit. The role of public policy should be to oversee digital advancements, verify capabilities, and scale and build public trust in the most promising technologies.

A more balanced regulatory approach to AI also illuminates tensions between current AI policies. Because AI requires better, more representative data, the right to privacy can conflict with the right to fair, unbiased, and accurate algorithmic decision-making. This article argues that the dominant policy frameworks regulating AI risks—emphasizing the right to human decision-making (human-in-the-loop) and the right to privacy (data minimization)—must be complemented with new corollary rights and duties: a right to automated decision-making (human-out-of-the-loop) and a right to complete and connected datasets (data maximization). Moreover, a shift to proactive governance of AI reveals the necessity for behavioral research on how to establish not only trustworthy AI, but also human rationality and trust in AI. Ironically, many of the legal protections currently proposed conflict with existing behavioral insights on human-machine trust. The article presents a blueprint for policymakers to engage in the deliberate study of how irrational aversion to automation can be mitigated through education, private-public governance, and smart policy design…(More)”

Machine Learning as a Tool for Hypothesis Generation

Paper by Jens Ludwig & Sendhil Mullainathan: “While hypothesis testing is a highly formalized activity, hypothesis generation remains largely informal. We propose a systematic procedure to generate novel hypotheses about human behavior, which uses the capacity of machine learning algorithms to notice patterns people might not. We illustrate the procedure with a concrete application: judge decisions about who to jail. We begin with a striking fact: The defendant’s face alone matters greatly for the judge’s jailing decision. In fact, an algorithm given only the pixels in the defendant’s mugshot accounts for up to half of the predictable variation. We develop a procedure that allows human subjects to interact with this black-box algorithm to produce hypotheses about what in the face influences judge decisions. The procedure generates hypotheses that are both interpretable and novel: They are not explained by demographics (e.g. race) or existing psychology research; nor are they already known (even if tacitly) to people or even experts. Though these results are specific, our procedure is general. It provides a way to produce novel, interpretable hypotheses from any high-dimensional dataset (e.g. cell phones, satellites, online behavior, news headlines, corporate filings, and high-frequency time series). A central tenet of our paper is that hypothesis generation is in and of itself a valuable activity, and hope this encourages future work in this largely “pre-scientific” stage of science…(More)”.

Urban AI Guide

Guide by Popelka, S., Narvaez Zertuche, L., Beroche, H.: “The idea for this guide arose from conversations with city leaders, who were confronted with new technologies, like artificial intelligence, as a means of solving complex urban problems, but who felt they lacked the background knowledge to properly engage with and evaluate the solutions. In some instances, this knowledge gap produced a barrier to project implementation or led to unintended project outcomes.

The guide begins with a literature review, presenting the state of the art in research on urban artificial intelligence. It then diagrams and describes an “urban AI anatomy,” outlining and explaining the components that make up an urban AI system. Insights from experts in the Urban AI community enrich this section, illuminating considerations involved in each component. Finally, the guide concludes with an in-depth examination of three case studies: water meter lifecycle in Winnipeg, Canada, curb digitization and planning in Los Angeles, USA, and air quality monitoring in Vilnius, Lithuania. Collectively, the case studies highlight the diversity of ways in which artificial intelligence can be operationalized in urban contexts, as well as the steps and requirements necessary to implement an urban AI project.

Since the field of urban AI is constantly evolving, we anticipate updating the guide annually. Please consider filling out the contribution form, if you have an urban AI use case that has been operationalized. We may contact you to include the use case as a case study in a future edition of the guide.

As a continuation of the guide, we offer customized workshops on urban AI, oriented toward municipalities and other urban stakeholders, who are interested in learning more about how artificial intelligence interacts in urban environments. Please contact us if you would like more information on this program…(More)”.

The False Promise of ChatGPT

Article by Noam Chomsky: “…OpenAI’s ChatGPT, Google’s Bard and Microsoft’s Sydney are marvels of machine learning. Roughly speaking, they take huge amounts of data, search for patterns in it and become increasingly proficient at generating statistically probable outputs — such as seemingly humanlike language and thought. These programs have been hailed as the first glimmers on the horizon of artificial general intelligence — that long-prophesied moment when mechanical minds surpass human brains not only quantitatively in terms of processing speed and memory size but also qualitatively in terms of intellectual insight, artistic creativity and every other distinctively human faculty.

That day may come, but its dawn is not yet breaking, contrary to what can be read in hyperbolic headlines and reckoned by injudicious investments. The Borgesian revelation of understanding has not and will not — and, we submit, cannot — occur if machine learning programs like ChatGPT continue to dominate the field of A.I. However useful these programs may be in some narrow domains (they can be helpful in computer programming, for example, or in suggesting rhymes for light verse), we know from the science of linguistics and the philosophy of knowledge that they differ profoundly from how humans reason and use language. These differences place significant limitations on what these programs can do, encoding them with ineradicable defects.

It is at once comic and tragic, as Borges might have noted, that so much money and attention should be concentrated on so little a thing — something so trivial when contrasted with the human mind, which by dint of language, in the words of Wilhelm von Humboldt, can make “infinite use of finite means,” creating ideas and theories with universal reach…(More)”.

I, Human: AI, Automation, and the Quest to Reclaim What Makes Us Unique

Book by Thomas Chamorro-Premuzic: “For readers of “Sapiens” and “Homo Deus” and viewers of “The Social Dilemma,” psychologist Tomas Chamorro-Premuzic tackles one of the biggest questions facing our species: Will we use artificial intelligence to improve the way we work and live, or will we allow it to alienate us? It’s no secret that AI is changing the way we live, work, love, and entertain ourselves. Dating apps are using AI to pick our potential partners. Retailers are using AI to predict our behavior and desires. Rogue actors are using AI to persuade us with bots and misinformation. Companies are using AI to hire us–or not. In “I, Human” psychologist Tomas Chamorro-Premuzic takes readers on an enthralling and eye-opening journey across the AI landscape. Though AI has the potential to change our lives for the better, he argues, AI is also worsening our bad tendencies, making us more distracted, selfish, biased, narcissistic, entitled, predictable, and impatient. It doesn’t have to be this way. Filled with fascinating insights about human behavior and our complicated relationship with technology, I, Human will help us stand out and thrive when many of our decisions are being made for us. To do so, we’ll need to double down on our curiosity, adaptability, and emotional intelligence while relying on the lost virtues of empathy, humility, and self-control. This is just the beginning. As AI becomes smarter and more humanlike, our societies, our economies, and our humanity will undergo the most dramatic changes we’ve seen since the Industrial Revolution. Some of these changes will enhance our species. Others may dehumanize us and make us more machinelike in our interactions with people. It’s up to us to adapt and determine how we want to live and work. The choice is ours. What will we decide?…(More)”.

How ChatGPT Hijacks Democracy

Article by Nathan E. Sanders and Bruce Schneier:”…But for all the consternation over the potential for humans to be replaced by machines in formats like poetry and sitcom scripts, a far greater threat looms: artificial intelligence replacing humans in the democratic processes — not through voting, but through lobbying.

ChatGPT could automatically compose comments submitted in regulatory processes. It could write letters to the editor for publication in local newspapers. It could comment on news articles, blog entries and social media posts millions of times every day. It could mimic the work that the Russian Internet Research Agency did in its attempt to influence our 2016 elections, but without the agency’s reported multimillion-dollar budget and hundreds of employees.Automatically generated comments aren’t a new problem. For some time, we have struggled with bots, machines that automatically post content. Five years ago, at least a million automatically drafted comments were believed to have been submitted to the Federal Communications Commission regarding proposed regulations on net neutrality. In 2019, a Harvard undergraduate, as a test, used a text-generation program to submit 1,001 comments in response to a government request for public input on a Medicaid issue. Back then, submitting comments was just a game of overwhelming numbers…(More)”

ChatGPT reminds us why good questions matter

Article by Stefaan Verhulst and Anil Ananthaswamy: “Over 100 million people used ChatGPT in January alone, according to one estimate, making it the fastest-growing consumer application in history. By producing resumes, essays, jokes and even poetry in response to prompts, the software brings into focus not just language models’ arresting power, but the importance of framing our questions correctly.

To that end, a few years ago I initiated the 100 Questions Initiative, which seeks to catalyse a cultural shift in the way we leverage data and develop scientific insights. The project aims not only to generate new questions, but also reimagine the process of asking them…

As a species and a society, we tend to look for answers. Answers appear to provide a sense of clarity and certainty, and can help guide our actions and policy decisions. Yet any answer represents a provisional end-stage of a process that begins with questions – and often can generate more questions. Einstein drew attention to the critical importance of how questions are framed, which can often determine (or at least play a significant role in determining) the answers we ultimately reach. Frame a question differently and one might reach a different answer. Yet as a society we undervalue the act of questioning – who formulates questions, how they do so, the impact they have on what we investigate, and on the decisions we make. Nor do we pay sufficient attention to whether the answers are in fact addressing the questions initially posed…(More)”.

‘There is no standard’: investigation finds AI algorithms objectify women’s bodies

Article by Hilke Schellmann: “Images posted on social media are analyzed by artificial intelligence (AI) algorithms that decide what to amplify and what to suppress. Many of these algorithms, a Guardian investigation has found, have a gender bias, and may have been censoring and suppressing the reach of countless photos featuring women’s bodies.

These AI tools, developed by large technology companies, including Google and Microsoft, are meant to protect users by identifying violent or pornographic visuals so that social media companies can block it before anyone sees it. The companies claim that their AI tools can also detect “raciness” or how sexually suggestive an image is. With this classification, platforms – including Instagram and LinkedIn – may suppress contentious imagery.

Two Guardian journalists used the AI tools to analyze hundreds of photos of men and women in underwear, working out, using medical tests with partial nudity and found evidence that the AI tags photos of women in everyday situations as sexually suggestive. They also rate pictures of women as more “racy” or sexually suggestive than comparable pictures of men. As a result, the social media companies that leverage these or similar algorithms have suppressed the reach of countless images featuring women’s bodies, and hurt female-led businesses – further amplifying societal disparities.

Even medical pictures are affected by the issue. The AI algorithms were tested on images released by the US National Cancer Institute demonstrating how to do a clinical breast examination. Google’s AI gave this photo the highest score for raciness, Microsoft’s AI was 82% confident that the image was “explicitly sexual in nature”, and Amazon classified it as representing “explicit nudity”…(More)”.

Work and meaning in the age of AI

Report by Daniel Susskind: “It is often said that work is not only a source of income but also of meaning. In this paper, I explore the theoretical and empirical literature that addresses this relationship between work and meaning. I show that the relationship is far less clear than is commonly supposed: There is a great heterogeneity in its nature, both among today’s workers and workers over time. I explain why this relationship matters for policymakers and economists concerned about the impact of technology on work. In the short term, it is important for predicting labour market outcomes of interest. It also matters for understanding how artificial intelligence (AI) affects not only the quantity of work but its quality as well: These new technologies may erode the meaning that people get from their work. In the medium term, if jobs are lost, this relationship also matters for designing bold policy interventions like the ‘Universal Basic Income’ and ‘Job Guarantee Schemes’: Their design, and any choice between them, is heavily dependent on policymakers’—often tacit—assumptions about the nature of this underlying relationship between work and meaning. For instance, policymakers must decide whether to simply focus on replacing lost income alone (as with a Universal Basic Income) or, if they believe that work is an important and non-substitutable source of meaning, on protecting jobs for that additional role as well (as with a Job Guarantee Scheme). In closing, I explore the challenge that the age of AI presents for an important feature of liberal political theory: the idea of ‘neutrality.’..(More)”