When A.I. Fails the Language Test, Who Is Left Out of the Conversation?

Curated on Posted on by Stefaan Verhulst

Article by Sara Ruberg: “While the use of A.I. has exploded in the West, much of the rest of the world has been left out of the conversation since most of the technology is trained in English. A.I. experts worry that the language gap could exacerbate technological inequities, and that it could leave many regions and cultures behind.

A delay of access to good technology of even a few years, “can potentially lead to a few decades of economic delay,” said Sang Truong, a Ph.D. candidate at the Stanford Artificial Intelligence Laboratory at Stanford University on the team that built and tested a Vietnamese language model against others.

The tests his team ran found that A.I. tools across the board could get facts and diction wrong when working with Vietnamese, likely because it is a “low-resource” language by industry standards, which means that there aren’t sufficient data sets and content available online for the A.I. model to learn from.

Low-resource languages are spoken by tens and sometimes hundreds of millions of people around the world, but they yield less digital data because A.I. tech development and online engagement is centered in the United States and China. Other low-resource languages include Hindi, Bengali and Swahili, as well as lesser-known dialects spoken by smaller populations around the world.

An analysis of top websites by W3Techs, a tech survey company, found that English makes up over 60 percent of the internet’s language data. While English is widely spoken globally, native English speakers make up about 5 percent of the population, according to Ethnologue, a research organization that collects language data. Mandarin and Spanish are other examples of languages with a significant online presence and reliable digital data sets.

Academic institutions, grass-roots organizations and volunteer efforts are playing catch-up to build resources for speakers of languages who aren’t as well represented in the digital landscape.

Lelapa AI, based in Johannesburg, is one such company leading efforts on the African continent. The South African-based start-up is developing multilingual A.I. products for people and businesses in Africa…(More)”.

Feeding the Machine: The Hidden Human Labor Powering A.I.

Curated on Posted on by Stefaan Verhulst

Book by Mark Graham, Callum Cant, and James Muldoon: “Silicon Valley has sold us the illusion that artificial intelligence is a frictionless technology that will bring wealth and prosperity to humanity. But hidden beneath this smooth surface lies the grim reality of a precarious global workforce of millions laboring under often appalling conditions to make A.I. possible. This book presents an urgent, riveting investigation of the intricate network that maintains this exploitative system, revealing the untold truth of A.I.

Based on hundreds of interviews and thousands of hours of fieldwork over more than a decade, Feeding the Machine describes the lives of the workers deliberately concealed from view, and the power structures that determine their future. It gives voice to the people whom A.I. exploits, from accomplished writers and artists to the armies of data annotators, content moderators and warehouse workers, revealing how their dangerous, low-paid labor is connected to longer histories of gendered, racialized, and colonial exploitation.

A.I. is an extraction machine that feeds off humanity’s collective effort and intelligence, churning through ever-larger datasets to power its algorithms. This book is a call to arms that details what we need to do to fight for a more just digital future…(More)”.

AI firms will soon exhaust most of the internet’s data

Curated on Posted on by Stefaan Verhulst

Article by The Economist: “One approach is to focus on data quality rather than quantity. ai labs do not simply train their models on the entire internet. They filter and sequence data to maximise how much their models learn. Naveen Rao of Databricks, an ai firm, says that this is the “main differentiator” between ai models on the market. “True information” about the world obviously matters; so does lots of “reasoning”. That makes academic textbooks, for example, especially valuable. But setting the balance between data sources remains something of a dark art. What is more, the ordering in which the system encounters different types of data matters too. Lump all the data on one topic, like maths, at the end of the training process, and your model may become specialised at maths but forget some other concepts.

These considerations can get even more complex when the data are not just on different subjects but in different forms. In part because of the lack of new textual data, leading models like Openai’s gpt-4o and Google’s Gemini are now let loose on image, video and audio files as well as text during their self-supervised learning. Training on video is hardest given how dense with data points video files are. Current models typically look at a subset of frames to simplify things.

Whatever models are used, ownership is increasingly recognised as an issue. The material used in training llms is often copyrighted and used without consent from, or payment to, the rights holders. Some ai models peep behind paywalls. Model creators claim this sort of thing falls under the “fair use” exemption in American copyright law. ai models should be allowed to read copyrighted material when they learn, just as humans can, they say. But as Benedict Evans, a technology analyst, has put it, “a difference in scale” can lead to “a difference in principle”…

It is clear that access to more data—whether culled from specialist sources, generated synthetically or provided by human experts—is key to maintaining rapid progress in ai. Like oilfields, the most accessible data reserves have been depleted. The challenge now is to find new ones—or sustainable alternatives…(More)”.

Anonymization: The imperfect science of using data while preserving privacy

Curated on Posted on by Stefaan Verhulst

Paper by Andrea Gadotti et al: “Information about us, our actions, and our preferences is created at scale through surveys or scientific studies or as a result of our interaction with digital devices such as smartphones and fitness trackers. The ability to safely share and analyze such data is key for scientific and societal progress. Anonymization is considered by scientists and policy-makers as one of the main ways to share data while minimizing privacy risks. In this review, we offer a pragmatic perspective on the modern literature on privacy attacks and anonymization techniques. We discuss traditional de-identification techniques and their strong limitations in the age of big data. We then turn our attention to modern approaches to share anonymous aggregate data, such as data query systems, synthetic data, and differential privacy. We find that, although no perfect solution exists, applying modern techniques while auditing their guarantees against attacks is the best approach to safely use and share data today…(More)”.

Training LLMs to Draft Replies to Parliamentary Questions

Curated on Posted on by Stefaan Verhulst

Blog by Watson Chua: “In Singapore, the government is answerable to Parliament and Members of Parliament (MPs) may raise queries to any Minister on any matter in his portfolio. These questions can be answered orally during the Parliament sitting or through a written reply. Regardless of the medium, public servants in the ministries must gather materials to answer the question and prepare a response.

Generative AI and Large Language Models (LLMs) have already been applied to help public servants do this more effectively and efficiently. For example, Pair Search (publicly accessible) and the Hansard Analysis Tool (only accessible to public servants) help public servants search for relevant information in past Parliamentary Sittings relevant to the question and synthesise a response to it.

The existing systems draft the responses using prompt engineering and Retrieval Augmented Generation (RAG). To recap, RAG consists of two main parts:

  • Retriever: A search engine that finds documents relevant to the question
  • Generator: A text generation model (LLM) that takes in the instruction, the question, and the search results from the retriever to respond to the question
A typical RAG system. Illustration by Hrishi Olickel, taken from here.

Using a pre-trained instruction-tuned LLM like GPT-4o, the generator can usually generate a good response. However, it might not be exactly what is desired in terms of verbosity, style and writing prose, and additional human post-processing might be needed. Extensive prompt engineering or few-shot learning can be done to mold the response at the expense of incurring higher costs from using additional tokens in the prompt…(More)”

The double-edged sword of AI in education

Curated on Posted on by Stefaan Verhulst

Article by Rose Luckin: “Artificial intelligence (AI) could revolutionize education as profoundly as the internet has already revolutionized our lives. However, our experience with commercial internet platforms gives us pause. Consider how social media algorithms, designed to maximize engagement and ad revenue, have inadvertently promoted divisive content and misinformation, a development at odds with educational goals.

Like the commercialization of the internet, the AI consumerization trend, driven by massive investments across sectors, prioritizes profit over societal and educational benefits. This focus on monetization risks overshadowing crucial considerations about AI’s integration into educational contexts.

The consumerization of AI in education is a double-edged sword. While increasing accessibility, it could also undermine fundamental educational principles and reshape students’ attitudes toward learning. We must advocate for a thoughtful, education-centric approach to AI development that enhances, rather than replaces, human intelligence and recognises the value of effort in learning.

As generative AI systems for education emerge, technical experts and policymakers have a unique opportunity to ensure their design supports the interests of learners and educators.

Risk 1: Overestimating AI’s intelligence

In essence, learning is not merely an individual cognitive process but a deeply social endeavor, intricately linked to cultural context, language development, and the dynamic relationship between practical experience and theoretical knowledge…(More)”.

AI mass surveillance at Paris Olympics

Curated on Posted on by Stefaan Verhulst

Article by Anne Toomey McKenna: “The 2024 Paris Olympics is drawing the eyes of the world as thousands of athletes and support personnel and hundreds of thousands of visitors from around the globe converge in France. It’s not just the eyes of the world that will be watching. Artificial intelligence systems will be watching, too.

Government and private companies will be using advanced AI tools and other surveillance tech to conduct pervasive and persistent surveillance before, during and after the Games. The Olympic world stage and international crowds pose increased security risks so significant that in recent years authorities and critics have described the Olympics as the “world’s largest security operations outside of war.”

The French government, hand in hand with the private tech sector, has harnessed that legitimate need for increased security as grounds to deploy technologically advanced surveillance and data gathering tools. Its surveillance plans to meet those risks, including controversial use of experimental AI video surveillance, are so extensive that the country had to change its laws to make the planned surveillance legal.

The plan goes beyond new AI video surveillance systems. According to news reports, the prime minister’s office has negotiated a provisional decree that is classified to permit the government to significantly ramp up traditional, surreptitious surveillance and information gathering tools for the duration of the Games. These include wiretapping; collecting geolocation, communications and computer data; and capturing greater amounts of visual and audio data…(More)”.

The impact of data portability on user empowerment, innovation, and competition

Curated on Posted on by Stefaan Verhulst

OECD Note: “Data portability enhances access to and sharing of data across digital services and platforms. It can empower users to play a more active role in the re-use of their data and can help stimulate competition and innovation by fostering interoperability while reducing switching costs and lock-in effects. However, the effectiveness of data portability in enhancing competition depends on the terms and conditions of data transfer and the extent to which competitors can make use of the data effectively. Additionally, there are potential downsides: data portability measures may unintentionally stifle competition in fast-evolving markets where interoperability requirements may disproportionately burden SMEs and start-ups. Data portability can also increase digital security and privacy risks by enabling data transfers to multiple destinations. This note presents the following five dimensions essential for designing and implementing data portability frameworks: sectoral scope; beneficiaries; type of data; legal obligations; and operational modality…(More)”.

Community consent: neither a ceiling nor a floor

Curated on Posted on by Stefaan Verhulst

Article by Jasmine McNealy: “The 23andMe breach and the Golden State Killer case are two of the more “flashy” cases, but questions of consent, especially the consent of all of those affected by biodata collection and analysis in more mundane or routine health and medical research projects, are just as important. The communities of people affected have expectations about their privacy and the possible impacts of inferences that could be made about them in data processing systems. Researchers must, then, acquire community consent when attempting to work with networked biodata. 

Several benefits of community consent exist, especially for marginalized and vulnerable populations. These benefits include:

  • Ensuring that information about the research project spreads throughout the community,
  • Removing potential barriers that might be created by resistance from community members,
  • Alleviating the possible concerns of individuals about the perspectives of community leaders, and 
  • Allowing the recruitment of participants using methods most salient to the community.

But community consent does not replace individual consent and limits exist for both community and individual consent. Therefore, within the context of a biorepository, understanding whether community consent might be a ceiling or a floor requires examining governance and autonomy…(More)”.

The Data That Powers A.I. Is Disappearing Fast

Curated on Posted on by Stefaan Verhulst

Article by Kevin Roose: “For years, the people building powerful artificial intelligence systems have used enormous troves of text, images and videos pulled from the internet to train their models.

Now, that data is drying up.

Over the past year, many of the most important web sources used for training A.I. models have restricted the use of their data, according to a study published this week by the Data Provenance Initiative, an M.I.T.-led research group.

The study, which looked at 14,000 web domains that are included in three commonly used A.I. training data sets, discovered an “emerging crisis in consent,” as publishers and online platforms have taken steps to prevent their data from being harvested.

The researchers estimate that in the three data sets — called C4, RefinedWeb and Dolma — 5 percent of all data, and 25 percent of data from the highest-quality sources, has been restricted. Those restrictions are set up through the Robots Exclusion Protocol, a decades-old method for website owners to prevent automated bots from crawling their pages using a file called robots.txt.

The study also found that as much as 45 percent of the data in one set, C4, had been restricted by websites’ terms of service.

“We’re seeing a rapid decline in consent to use data across the web that will have ramifications not just for A.I. companies, but for researchers, academics and noncommercial entities,” said Shayne Longpre, the study’s lead author, in an interview.

Data is the main ingredient in today’s generative A.I. systems, which are fed billions of examples of text, images and videos. Much of that data is scraped from public websites by researchers and compiled in large data sets, which can be downloaded and freely used, or supplemented with data from other sources…(More)”.