Fairness and Machine Learning

Curated on Posted on by Stefaan Verhulst

Book by Solon Barocas, Moritz Hardt and Arvind Narayanan: “…introduces advanced undergraduate and graduate students to the intellectual foundations of this recently emergent field, drawing on a diverse range of disciplinary perspectives to identify the opportunities and hazards of automated decision-making. It surveys the risks in many applications of machine learning and provides a review of an emerging set of proposed solutions, showing how even well-intentioned applications may give rise to objectionable results. It covers the statistical and causal measures used to evaluate the fairness of machine learning models as well as the procedural and substantive aspects of decision-making that are core to debates about fairness, including a review of legal and philosophical perspectives on discrimination. This incisive textbook prepares students of machine learning to do quantitative work on fairness while reflecting critically on its foundations and its practical utility.

• Introduces the technical and normative foundations of fairness in automated decision-making
• Covers the formal and computational methods for characterizing and addressing problems
• Provides a critical assessment of their intellectual foundations and practical utility
• Features rich pedagogy and extensive instructor resources…(More)”

Power to the standards

Curated on Posted on by Stefaan Verhulst

Report by Gergana Baeva, Michael Puntschuh and Matthieu Binder: “Standards and norms will be of central importance when it comes to the practical implementation of legal requirements for developed and deployed AI systems.

Using expert interviews, our study “Power to the standards” documents the existing obstacles on the way to the standardization of AI. In addition to practical and technological challenges, questions of democratic policy arise. After all, requirements such as fairness or transparency are often regarded as criteria to be determined by the legislator, meaning that they are only partially susceptible to standardization.

Our study concludes that the targeted and comprehensive participation of civil society actors is particularly necessary in order to compensate for existing participation deficits within the standardization process…(More)”.

Rebalancing AI

Curated on Posted on by Stefaan Verhulst

Article by Daron Acemoglu and Simon Johnson: “Optimistic forecasts regarding the growth implications of AI abound. AI adoption could boost productivity growth by 1.5 percentage points per year over a 10-year period and raise global GDP by 7 percent ($7 trillion in additional output), according to Goldman Sachs. Industry insiders offer even more excited estimates, including a supposed 10 percent chance of an “explosive growth” scenario, with global output rising more than 30 percent a year.

All this techno-optimism draws on the “productivity bandwagon”: a deep-rooted belief that technological change—including automation—drives higher productivity, which raises net wages and generates shared prosperity.

Such optimism is at odds with the historical record and seems particularly inappropriate for the current path of “just let AI happen,” which focuses primarily on automation (replacing people). We must recognize that there is no singular, inevitable path of development for new technology. And, assuming that the goal is to sustainably improve economic outcomes for more people, what policies would put AI development on the right path, with greater focus on enhancing what all workers can do?…(More)”

What Will AI Do to Elections?

Curated on Posted on by Stefaan Verhulst

Article by Rishi Iyengar: “…Requests to X’s press team on how the platform was preparing for elections in 2024 yielded an automated response: “Busy now, please check back later”—a slight improvement from the initial Musk-era change where the auto-reply was a poop emoji.

X isn’t the only major social media platform with fewer content moderators. Meta, which owns Facebook, Instagram, and WhatsApp, has laid off more than 20,000 employees since November 2022—several of whom worked on trust and safety—while many YouTube employees working on misinformation policy were impacted by layoffs at parent company Google.

There could scarcely be a worse time to skimp on combating harmful content online. More than 50 countries, including the world’s three biggest democracies and Taiwan, an increasingly precarious geopolitical hot spot, are expected to hold national elections in 2024. Seven of the world’s 10 most populous countries—Bangladesh, India, Indonesia, Mexico, Pakistan, Russia, and the United States—will collectively send a third of the world’s population to the polls.

Elections, with their emotionally charged and often tribal dynamics, are where misinformation missteps come home to roost. If social media misinformation is the equivalent of yelling “fire” in a crowded theater, election misinformation is like doing so when there’s a horror movie playing and everyone’s already on edge.

Katie Harbath prefers a different analogy, one that illustrates how nebulous and thorny the issues are and the sheer uncertainty surrounding them. “The metaphor I keep using is a kaleidoscope because there’s so many different aspects to this but depending how you turn the kaleidoscope, the pattern changes of what it’s going to look like,” she said in an interview in October. “And that’s how I feel about life post-2024. … I don’t know where in the kaleidoscope it’s going to land.”

Harbath has become something of an election whisperer to the tech industry, having spent a decade at Facebook from 2011 building the company’s election integrity efforts from scratch. She left in 2021 and founded Anchor Change, a public policy consulting firm that helps other platforms combat misinformation and prepare for elections in particular.

Had she been in her old job, Harbath said, her team would have completed risk assessments of global elections by late 2022 or early 2023 and then spent the rest of the year tailoring Meta’s products to them as well as setting up election “war rooms” where necessary. “Right now, we would be starting to move into execution mode.” She cautions against treating the resources that companies are putting into election integrity as a numbers game—“once you build some of those tools, maintaining them doesn’t take as many people”—but acknowledges that the allocation of resources reveals a company leadership’s priorities.

The companies insist they remain committed to election integrity. YouTube has “heavily invested in the policies and systems that help us successfully support elections around the world,” spokesperson Ivy Choi said in a statement. TikTok said it has a total of 40,000 safety professionals and works with 16 fact-checking organizations across 50 global languages. Meta declined to comment for this story, but a company representative directed Foreign Policy to a recent blog post by Nick Clegg, a former U.K. deputy prime minister who now serves as Meta’s head of global affairs. “We have around 40,000 people working on safety and security, with more than $20 billion invested in teams and technology in this area since 2016,” Clegg wrote in the post.

But there are other troubling signs. YouTube announced last June that it would stop taking down content spreading false claims about the 2020 U.S. election or past elections, and Meta quietly made a similar policy change to its political ad rules in 2022. And as past precedent has shown, the platforms tend to have even less cover outside the West, with major blind spots in local languages and context making misinformation and hate speech not only more pervasive but also more dangerous…(More)”.

How can Mixed Reality and AI improve emergency medical care?

Curated on Posted on by Stefaan Verhulst

Springwise: “Mixed reality (MR) refers to technologies that create immersive computer-generated environments in which parts of the physical and virtual environment are combined. With potential applications that range from education and engineering to entertainment, the market for MR is forecast to record revenues of just under $25 billion by 2032. Now, in a ground-breaking partnership, Singapore-based company Mediwave is teaming up with Sri Lanka’s 1990 Suwa Seriya to deploy MR and artificial intelligence (AI) to create a fully connected ambulance.

1990 Suwa Seriya is Sri Lanka’s national pre-hospital emergency ambulance service, which boasts response times that surpass even some services in developed countries. The innovative ambulance it has deployed uses Mediwave’s integrated Emergency Response Suite, which combines the latest communications equipment with internet-of-things (IoT) and AR capabilities to enhance the efficiency of the emergency response eco-system.

The connected ambulance ensures swift response times and digitises critical processes, while specialised care can be provided remotely through a Microsoft HoloLens. The technology enables Emergency Medical Technicians (EMTs) – staff who man ambulances in Sri Lanka – to connect with physicians at the Emergency Command and Control Centre. These physicians help the EMTs provide care during the so-called ‘golden hour’ of medical emergencies – the concept that rapid clinical investigation and care within 60 minutes of a traumatic injury is essential for a positive patient outcome…

Other applications of extended reality in the Springwise library include holograms that are used to train doctors, virtual environments for treating phobias, and an augmented reality contact lens…(More)”.

The Algorithm: How AI Decides Who Gets Hired, Monitored, Promoted, and Fired and Why We Need to Fight Back Now

Curated on Posted on by Stefaan Verhulst

Book by Hilke Schellmann: “Based on exclusive information from whistleblowers, internal documents, and real world test results, Emmy‑award winning Wall Street Journal contributor Hilke Schellmann delivers a shocking and illuminating expose on the next civil rights issue of our time: how AI has already taken over the workplace and shapes our future.
 
Hilke Schellmann, is an Emmy‑award winning investigative reporter, Wall Street Journal and Guardian contributor and Journalism Professor at NYU. In The Algorithm, she investigates the rise of artificial intelligence (AI) in the world of work. AI is now being used to decide who has access to an education, who gets hired, who gets fired, and who receives a promotion. Drawing on exclusive information from whistleblowers, internal documents and real‑world tests, Schellmann discovers that many of the algorithms making high‑stakes decisions are biased, racist, and do more harm than good. Algorithms are on the brink of dominating our lives and threaten our human future—if we don’t fight back. 
 
Schellmann takes readers on a journalistic detective story testing algorithms that have secretly analyzed job candidates’ facial expressions and tone of voice. She investigates algorithms that scan our online activity including Twitter and LinkedIn to construct personality profiles à la Cambridge Analytica. Her reporting reveals how employers track the location of their employees, the keystrokes they make, access everything on their screens and, during meetings, analyze group discussions to diagnose problems in a team. Even universities are now using predictive analytics for admission offers and financial aid…(More)”

Charting the Emerging Geography of AI

Curated on Posted on by Stefaan Verhulst

Article by Bhaskar Chakravorti, Ajay Bhalla, and Ravi Shankar Chaturvedi: “Given the high stakes of this race, which countries are in the lead? Which are gaining on the leaders? How might this hierarchy shape the future of AI? Identifying AI-leading countries is not straightforward, as data, knowledge, algorithms, and models can, in principle, cross borders. Even the U.S.–China rivalry is complicated by the fact that AI researchers from the two countries cooperate — and more so than researchers from any other pair of countries. Open-source models are out there for everyone to use, with licensing accessible even for cutting-edge models. Nonetheless, AI development benefits from scale economies and, as a result, is geographically clustered as many significant inputs are concentrated and don’t cross borders that easily….

Rapidly accumulating pools of data in digital economies around the world are clearly one of the critical drivers of AI development. In 2019, we introduced the idea of “gross data product” of countries determined by the volume, complexity, and accessibility of data consumed alongside the number of active internet users in the country. For this analysis, we recognized that gross data product is an essential asset for AI development — especially for generative AI, which requires massive, diverse datasets — and updated the 2019 analyses as a foundation, adding drivers that are critical for AI development overall. That essential data layer makes the index introduced here distinct from other indicators of AI “vibrancy” or measures of global investments, innovations, and implementation of AI…(More)”.

New group aims to professionalize AI auditing

Curated on Posted on by Stefaan Verhulst

Article by Louise Matsakis: “The newly formed International Association of Algorithmic Auditors (IAAA) is hoping to professionalize the sector by creating a code of conduct for AI auditors, training curriculums, and eventually, a certification program.

Over the last few years, lawmakers and researchers have repeatedly proposed the same solution for regulating artificial intelligence: require independent audits. But the industry remains a wild west; there are only a handful of reputable AI auditing firms and no established guardrails for how they should conduct their work.

Yet several jurisdictions have passed laws mandating tech firms to commission independent audits, including New York City. The idea is that AI firms should have to demonstrate their algorithms work as advertised, the same way companies need to prove they haven’t fudged their finances.

Since ChatGPT was released last year, a troubling norm has been established in the AI industry, which is that it’s perfectly acceptable to evaluate your own models in-house.

Leading startups like OpenAI and Anthropic regularly publish research about the AI systems they’re developing, including the potential risks. But they rarely commission independent audits, let alone publish the results, making it difficult for anyone to know what’s really happening under the hood…(More)”..(More)”

Considerations for Governing Open Foundation Models

Curated on Posted on by Stefaan Verhulst

Brief by Rishi Bommasani et al: “Foundation models (e.g., GPT-4, Llama 2) are at the epicenter of AI, driving technological innovation and billions in investment. This paradigm shift has sparked widespread demands for regulation. Animated by factors as diverse as declining transparency and unsafe labor practices, limited protections for copyright and creative work, as well as market concentration and productivity gains, many have called for policymakers to take action.

Central to the debate about how to regulate foundation models is the process by which foundation models are released. Some foundation models like Google DeepMind’s Flamingo are fully closed, meaning they are available only to the model developer; others, such as OpenAI’s GPT-4, are limited access, available to the public but only as a black box; and still others, such as Meta’s Llama 2, are more open, with widely available model weights enabling downstream modification and scrutiny. As of August 2023, the U.K.’s Competition and Markets Authority documents the most common release approach for publicly-disclosed models is open release based on data from Stanford’s Ecosystem Graphs. Developers like Meta, Stability AI, Hugging Face, Mistral, Together AI, and EleutherAI frequently release models openly.

Governments around the world are issuing policy related to foundation models. As part of these efforts, open foundation models have garnered significant attention: The recent U.S. Executive Order on Safe, Secure, and Trustworthy Artificial Intelligence tasks the National Telecommunications and Information Administration with preparing a report on open foundation models for the president. In the EU, open foundation models trained with fewer than 1025 floating point operations (a measure of the amount of compute expended) appear to be exempted under the recently negotiated AI Act. The U.K.’s AI Safety Institute will “consider open-source systems as well as those deployed with various forms of access controls” as part of its initial priorities. Beyond governments, the Partnership on AI has introduced guidelines for the safe deployment of foundation models, recommending against open release for the most capable foundation models.

Policy on foundation models should support the open foundation model ecosystem, while providing resources to monitor risks and create safeguards to address harms. Open foundation models provide significant benefits to society by promoting competition, accelerating innovation, and distributing power. For example, small businesses hoping to build generative AI applications could choose among a variety of open foundation models that offer different capabilities and are often less expensive than closed alternatives. Further, open models are marked by greater transparency and, thereby, accountability. When a model is released with its training data, independent third parties can better assess the model’s capabilities and risks…(More)”.

How Moral Can A.I. Really Be?

Curated on Posted on by Stefaan Verhulst

Article by Paul Bloom: “…The problem isn’t just that people do terrible things. It’s that people do terrible things that they consider morally good. In their 2014 book “Virtuous Violence,” the anthropologist Alan Fiske and the psychologist Tage Rai argue that violence is often itself a warped expression of morality. “People are impelled to violence when they feel that to regulate certain social relationships, imposing suffering or death is necessary, natural, legitimate, desirable, condoned, admired, and ethically gratifying,” they write. Their examples include suicide bombings, honor killings, and war. The philosopher Kate Manne, in her book “Down Girl,” makes a similar point about misogynistic violence, arguing that it’s partially rooted in moralistic feelings about women’s “proper” role in society. Are we sure we want A.I.s to be guided by our idea of morality?

Schwitzgebel suspects that A.I. alignment is the wrong paradigm. “What we should want, probably, is not that superintelligent AI align with our mixed-up, messy, and sometimes crappy values but instead that superintelligent AI have ethically good values,” he writes. Perhaps an A.I. could help to teach us new values, rather than absorbing old ones. Stewart, the former graduate student, argued that if researchers treat L.L.M.s as minds and study them psychologically, future A.I. systems could help humans discover moral truths. He imagined some sort of A.I. God—a perfect combination of all the great moral minds, from Buddha to Jesus. A being that’s better than us.

Would humans ever live by values that are supposed to be superior to our own? Perhaps we’ll listen when a super-intelligent agent tells us that we’re wrong about the facts—“this plan will never work; this alternative has a better chance.” But who knows how we’ll respond if one tells us, “You think this plan is right, but it’s actually wrong.” How would you feel if your self-driving car tried to save animals by refusing to take you to a steakhouse? Would a government be happy with a military A.I. that refuses to wage wars it considers unjust? If an A.I. pushed us to prioritize the interests of others over our own, we might ignore it; if it forced us to do something that we consider plainly wrong, we would consider its morality arbitrary and cruel, to the point of being immoral. Perhaps we would accept such perverse demands from God, but we are unlikely to give this sort of deference to our own creations. We want alignment with our own values, then, not because they are the morally best ones, but because they are ours…(More)”