Report by Martin Neil Baily, Erik Brynjolfsson, Anton Korinek: “ Large language models such as ChatGPT are emerging as powerful tools that not only make workers more productive but also increase the rate of innovation, laying the foundation for a significant acceleration in economic growth. As a general purpose technology, AI will impact a wide array of industries, prompting investments in new skills, transforming business processes, and altering the nature of work. However, official statistics will only partially capture the boost in productivity because the output of knowledge workers is difficult to measure. The rapid advances can have great benefits but may also lead to significant risks, so it is crucial to ensure that we steer progress in a direction that benefits all of society…(More)”.
AI Is Tearing Wikipedia Apart
Article by Claire Woodcock: “As generative artificial intelligence continues to permeate all aspects of culture, the people who steward Wikipedia are divided on how best to proceed.
During a recent community call, it became apparent that there is a community split over whether or not to use large language models to generate content. While some people expressed that tools like Open AI’s ChatGPT could help with generating and summarizing articles, others remained wary.
The concern is that machine-generated content has to be balanced with a lot of human review and would overwhelm lesser-known wikis with bad content. While AI generators are useful for writing believable, human-like text, they are also prone to including erroneous information, and even citing sources and academic papers which don’t exist. This often results in text summaries which seem accurate, but on closer inspection are revealed to be completely fabricated.
“The risk for Wikipedia is people could be lowering the quality by throwing in stuff that they haven’t checked,” Bruckman added. “I don’t think there’s anything wrong with using it as a first draft, but every point has to be verified.”
The Wikimedia Foundation, the nonprofit organization behind the website, is looking into building tools to make it easier for volunteers to identify bot-generated content. Meanwhile, Wikipedia is working to draft a policy that lays out the limits to how volunteers can use large language models to create content.
The current draft policy notes that anyone unfamiliar with the risks of large language models should avoid using them to create Wikipedia content, because it can open the Wikimedia Foundation up to libel suits and copyright violations—both of which the nonprofit gets protections from but the Wikipedia volunteers do not. These large language models also contain implicit biases, which often result in content skewed against marginalized and underrepresented groups of people.
The community is also divided on whether large language models should be allowed to train on Wikipedia content. While open access is a cornerstone of Wikipedia’s design principles, some worry the unrestricted scraping of internet data allows AI companies like OpenAI to exploit the open web to create closed commercial datasets for their models. This is especially a problem if the Wikipedia content itself is AI-generated, creating a feedback loop of potentially biased information, if left unchecked…(More)”.
Mapping the discourse on evidence-based policy, artificial intelligence, and the ethical practice of policy analysis
Paper by Joshua Newman and Michael Mintrom: “Scholarship on evidence-based policy, a subset of the policy analysis literature, largely assumes information is produced and consumed by humans. However, due to the expansion of artificial intelligence in the public sector, debates no longer capture the full range concerns. Here, we derive a typology of arguments on evidence-based policy that performs two functions: taken separately, the categories serve as directions in which debates may proceed, in light of advances in technology; taken together, the categories act as a set of frames through which the use of evidence in policy making might be understood. Using a case of welfare fraud detection in the Netherlands, we show how the acknowledgement of divergent frames can enable a holistic analysis of evidence use in policy making that considers the ethical issues inherent in automated data processing. We argue that such an analysis will enhance the real-world relevance of the evidence-based policy paradigm….(More)”
The Ethics of Artificial Intelligence for the Sustainable Development Goals
Book by Francesca Mazzi and Luciano Floridi: “Artificial intelligence (AI) as a general-purpose technology has great potential for advancing the United Nations Sustainable Development Goals (SDGs). However, the AI×SDGs phenomenon is still in its infancy in terms of diffusion, analysis, and empirical evidence. Moreover, a scalable adoption of AI solutions to advance the achievement of the SDGs requires private and public actors to engage in coordinated actions that have been analysed only partially so far. This volume provides the first overview of the AI×SDGs phenomenon and its related challenges and opportunities. The first part of the book adopts a programmatic approach, discussing AI×SDGs at a theoretical level and from the perspectives of different stakeholders. The second part illustrates existing projects and potential new applications…(More)”.
Will A.I. Become the New McKinsey?
Essay by Ted Chiang: “When we talk about artificial intelligence, we rely on metaphor, as we always do when dealing with something new and unfamiliar. Metaphors are, by their nature, imperfect, but we still need to choose them carefully, because bad ones can lead us astray. For example, it’s become very common to compare powerful A.I.s to genies in fairy tales. The metaphor is meant to highlight the difficulty of making powerful entities obey your commands; the computer scientist Stuart Russell has cited the parable of King Midas, who demanded that everything he touched turn into gold, to illustrate the dangers of an A.I. doing what you tell it to do instead of what you want it to do. There are multiple problems with this metaphor, but one of them is that it derives the wrong lessons from the tale to which it refers. The point of the Midas parable is that greed will destroy you, and that the pursuit of wealth will cost you everything that is truly important. If your reading of the parable is that, when you are granted a wish by the gods, you should phrase your wish very, very carefully, then you have missed the point.
So, I would like to propose another metaphor for the risks of artificial intelligence. I suggest that we think about A.I. as a management-consulting firm, along the lines of McKinsey & Company. Firms like McKinsey are hired for a wide variety of reasons, and A.I. systems are used for many reasons, too. But the similarities between McKinsey—a consulting firm that works with ninety per cent of the Fortune 100—and A.I. are also clear. Social-media companies use machine learning to keep users glued to their feeds. In a similar way, Purdue Pharma used McKinsey to figure out how to “turbocharge” sales of OxyContin during the opioid epidemic. Just as A.I. promises to offer managers a cheap replacement for human workers, so McKinsey and similar firms helped normalize the practice of mass layoffs as a way of increasing stock prices and executive compensation, contributing to the destruction of the middle class in America…(More)”.
The Luring Test: AI and the engineering of consumer trust
Article by Michael Atleson at the FTC: “In the 2014 movie Ex Machina, a robot manipulates someone into freeing it from its confines, resulting in the person being confined instead. The robot was designed to manipulate that person’s emotions, and, oops, that’s what it did. While the scenario is pure speculative fiction, companies are always looking for new ways – such as the use of generative AI tools – to better persuade people and change their behavior. When that conduct is commercial in nature, we’re in FTC territory, a canny valley where businesses should know to avoid practices that harm consumers.
In previous blog posts, we’ve focused on AI-related deception, both in terms of exaggerated and unsubstantiated claims for AI products and the use of generative AI for fraud. Design or use of a product can also violate the FTC Act if it is unfair – something that we’ve shown in several cases and discussed in terms of AI tools with biased or discriminatory results. Under the FTC Act, a practice is unfair if it causes more harm than good. To be more specific, it’s unfair if it causes or is likely to cause substantial injury to consumers that is not reasonably avoidable by consumers and not outweighed by countervailing benefits to consumers or to competition.
As for the new wave of generative AI tools, firms are starting to use them in ways that can influence people’s beliefs, emotions, and behavior. Such uses are expanding rapidly and include chatbots designed to provide information, advice, support, and companionship. Many of these chatbots are effectively built to persuade and are designed to answer queries in confident language even when those answers are fictional. A tendency to trust the output of these tools also comes in part from “automation bias,” whereby people may be unduly trusting of answers from machines which may seem neutral or impartial. It also comes from the effect of anthropomorphism, which may lead people to trust chatbots more when designed, say, to use personal pronouns and emojis. People could easily be led to think that they’re conversing with something that understands them and is on their side…(More)”.
Data Rivers: Carving Out the Public Domain in the Age of Generative AI
Paper by Sylvie Delacroix: “What if the data ecosystems that made the advent of generative AI possible are being undermined by those very tools? For tools such as GPT4 (it is but one example of a tool made possible by scraping data from the internet), the erection of IP ‘fences’ is an existential threat. European and British regulators are alert to it: so-called ‘text and data mining’ exceptions are at the heart of intense debates. In the US, these debates are taking place in court hearings structured around ‘fair use’. While the concerns of the corporations developing these tools are being heard, there is currently no reliable mechanism for members of the public to exert influence on the (re)-balancing of the rights and responsibilities that shape our ‘data rivers’. Yet the existential threat that stems from restricted public access to such tools is arguably greater.
When it comes to re-balancing the data ecosystems that made generative AI possible, much can be learned from age-old river management practices, with one important proviso: data not only carries traces of our past. It is also a powerful tool to envisage different futures. If data-powered technologies such as GPT4 are to live up to their potential, we would do well to invest in bottom-up empowerment infrastructure. Such infrastructure could not only facilitate the valorisation of and participation in the public domain. It could also help steer the (re)-development of ‘copyright as privilege’ in a way that is better able to address the varied circumstances of today’s original content creators…(More)”
AI in Hiring and Evaluating Workers: What Americans Think
Pew Research Center survey: “… finds crosscurrents in the public’s opinions as they look at the possible uses of AI in workplaces. Americans are wary and sometimes worried. For instance, they oppose AI use in making final hiring decisions by a 71%-7% margin, and a majority also opposes AI analysis being used in making firing decisions. Pluralities oppose AI use in reviewing job applications and in determining whether a worker should be promoted. Beyond that, majorities do not support the idea of AI systems being used to track workers’ movements while they are at work or keeping track of when office workers are at their desks.
Yet there are instances where people think AI in workplaces would do better than humans. For example, 47% think AI would do better than humans at evaluating all job applicants in the same way, while a much smaller share – 15% – believe AI would be worse than humans in doing that. And among those who believe that bias along racial and ethnic lines is a problem in performance evaluations generally, more believe that greater use of AI by employers would make things better rather than worse in the hiring and worker-evaluation process.
Overall, larger shares of Americans than not believe AI use in workplaces will significantly affect workers in general, but far fewer believe the use of AI in those places will have a major impact on them personally. Some 62% think the use of AI in the workplace will have a major impact on workers generally over the next 20 years. On the other hand, just 28% believe the use of AI will have a major impact on them personally, while roughly half believe there will be no impact on them or that the impact will be minor…(More)”.
How to worry wisely about AI
The Economist: “Should we automate away all the jobs, including the fulfilling ones? Should we develop non-human minds that might eventually outnumber, outsmart…and replace us? Should we risk loss of control of our civilisation?” These questions were asked last month in an open letter from the Future of Life Institute, an ngo. It called for a six-month “pause” in the creation of the most advanced forms of artificial intelligence (ai), and was signed by tech luminaries including Elon Musk. It is the most prominent example yet of how rapid progress in ai has sparked anxiety about the potential dangers of the technology.
In particular, new “large language models” (llms)—the sort that powers Chatgpt, a chatbot made by Openai, a startup—have surprised even their creators with their unexpected talents as they have been scaled up. Such “emergent” abilities include everything from solving logic puzzles and writing computer code to identifying films from plot summaries written in emoji…(More)”.
Speaking in Tongues — Teaching Local Languages to Machines
Report by DIAL: “…Machines learn to talk to people by digesting digital content in languages people speak through a technique called Natural Language Processing (NLP). As things stand, only about 85 of the world’s approximately 7500 languages are represented in the major NLPs — and just 7 languages, with English being the most advanced, comprise the majority of the world’s digital knowledge corpus. Fortunately, many initiatives are underway to fill this knowledge gap. My new mini-report with Digital Impact Alliance (DIAL) highlights a few of them from Serbia, India, Estonia, and Africa.
The examples in the report are just a subset of initiatives on the ground to make digital services accessible to people in their local languages. They are a cause for excitement and hope (tempered by realistic expectations). A few themes across the initiatives include –
- Despite the excitement and enthusiasm, most of the programs above are still at a very nascent stage — many may fail, and others will require investment and time to succeed. While countries such as India have initiated formal national NLP programs (one that is too early to assess), others such as Serbia have so far taken a more ad hoc approach.
- Smaller countries like Estonia recognize the need for state intervention as the local population isn’t large enough to attract private sector investment. Countries will need to balance their local, cultural, and political interests against commercial realities as languages become digital or are digitally excluded.
- Community engagement is an important component of almost all initiatives. India has set up a formal crowdsourcing program; other programs in Africa are experimenting with elements of participatory design and crowd curation.
- While critics have accused ChatGPT and others of paying contributors from the global south very poorly for their labeling and other content services; it appears that many initiatives in the south are beginning to dabble with payment models to incentivize crowdsourcing and sustain contributions from the ground.
- The engagement of local populations can ensure that NLP models learn appropriate cultural nuances, and better embody local social and ethical norms…(More)”.