This paper is generally well written and clear, albeit targeting readers with formal backgrounds. The quality of the paper seems high in terms of its formal claims. The proposed mechanism is remarkable simple, making this an attractive approach. I really like the idea behind not making learning explicit (as opposed to rule induction for example). I have three main concerns about this paper: - In general it is very close to Juba's 2012 work [1].

Recent research in natural language processing (NLP) has achieved impressive performance in tasks such as machine translation (MT), news classification, and question-answering in high-resource languages. However, the performance of MT leaves much to be desired for low-resource languages. This is due to the smaller size of available parallel corpora in these languages, if such corpora are available at all. NLP in Ethiopian languages suffers from the same issues due to the unavailability of publicly accessible datasets for NLP tasks, including MT. To help the research community and foster research for Ethiopian languages, we introduce EthioMT -- a new parallel corpus for 15 languages. We also create a new benchmark by collecting a dataset for better-researched languages in Ethiopia. We evaluate the newly collected corpus and the benchmark dataset for 23 Ethiopian languages using transformer and fine-tuning approaches.

Rare object detection is a fundamental task in applied geospatial machine learning, however is often challenging due to large amounts of high-resolution satellite or aerial imagery and few or no labeled positive samples to start with. This paper addresses the problem of bootstrapping such a rare object detection task assuming there is no labeled data and no spatial prior over the area of interest. We propose novel offline and online cluster-based approaches for sampling patches that are significantly more efficient, in terms of exposing positive samples to a human annotator, than random sampling. We apply our methods for identifying bomas, or small enclosures for herd animals, in the Serengeti Mara region of Kenya and Tanzania. We demonstrate a significant enhancement in detection efficiency, achieving a positive sampling rate increase from 2% (random) to 30%. This advancement enables effective machine learning mapping even with minimal labeling budgets, exemplified by an F1 score on the boma detection task of 0.51 with a budget of 300 total patches.

From South Juba to Entebbe, from Marrakesh to Accra, on the cusp of technology in Africa, the need for responsible AI development and ethical data practices has never been more pressing. As technology continues to advance and shape the global economy, Africa is taking steps toward positioning itself as a leader in Artificial Intelligence (AI). Investments and innovations in AI are on the rise across the continent, with a growing number of countries beginning to develop policies and strategies to harness the power of this transformative technology. Although only a few countries have officially adopted AI strategies and policies, many more are actively working towards defining their AI policies. As philosopher and economist Amartya Sen noted, 'Development requires the removal of major sources of unfreedom that leave people with little choice and little opportunity of exercising their reasoned agency.'

It was while fleeing the civil war in South Sudan that Lual Mayen's mother gave birth to him 28 years ago. She had four children in tow and was near to the border with Uganda, in a town called Aswa. The journey was difficult; Mayen's two sisters died on the way and he became sick. No one thought he would survive. "I can't imagine what she had to go through. There was no food, no water, nothing," says Mayen. "I remember she said she was not the only woman who gave birth on the way. Other women abandoned their children because they didn't want them to suffer. But my mother thought: "He is a gift for me, I have to keep him."' Mayen's mother made it to northern Uganda with her newborn son and reunited with her husband in a refugee camp that remained their home for the next 22 years. Mayen grew up there, and although life was a struggle, he was happy and grateful for what he had. There wasn't much to do but Mayen says he found creative ways to keep himself entertained. Then, one day he had the chance to play the video game Grand Theft Auto, which mostly revolves around driving and shooting. "While I was playing, this thought came into my mind," he remembers. "In South Sudan, most of the population is under 30.

Linear regression is a technique frequently used in statistical and data analysis. The task for standard linear regression is to fit a linear relationship among variables in a data set. Often, the goal is to find the most parsimonious model that can describe the majority of the data. In this work, we consider the situation where only a small portion of the data can be accurately modeled using linear regression. More generally, in many kinds of real-world data, portions of the data of significant size can be predicted significantly more accurately than by the best linear model for the overall data distribution: Rosenfeld et al. (2015) showed that there are attributes that are significant risk factors for gastrointestinal cancer in certain subpopulations, but not in the overall population. Hainline et al. (2019) demonstrated that a variety of standard (real-world) regression benchmarks have portions that are fit significantly better by a different linear model than the best model for the overall data set; Calderon et al. (2020) presented further, similar findings. We will consider cases where linear regression fits well when the data set is conditioned on a simple condition, which is unknown to us. We study the task of finding such a linear model, together with a formula on the data attributes describing the condition, i.e., the portion of the data for which the linear model is accurate. This problem was introduced by Juba (2017), who gave an algorithm for conditional sparse linear regression, using the maximum residual as the objective.

Devices and tools activated through speaking will soon be the primary way people interact with technology, yet none of the main voice assistants, including Amazon's Alexa, Apple's Siri and Google Assistant, support a single native African language. Mozilla has sought to address this problem through the Common Voice project, which is now working to expand voice technology to the 100 million people who speak Kiswahili across Kenya, Uganda, Tanzania, Rwanda, Burundi and South Sudan. The open source project makes it easy for anyone to donate their voice to a publicly available database that can then be used to train voice-enabled devices, and over the past two years, more than 840 Rwandans have donated over 1,700 hours of voice data in Kinyarwanda, a language with over 12 million speakers. That voice data is now being used to help train voice chatbots with speech-to-text and text-to-speech functionality that has important information about COVID-19, according to Chenai Chair, special advisor for Africa Innovation at the Mozilla Foundation. A handful of major tech companies control the voice data that is currently used to train machine learning algorithms, posing a challenge for companies seeking to develop high-quality speech recognition technologies while also exacerbating the voice recognition divide between English speakers and the rest of the world.

REFUNITE, based in California, is trialing the app in Uganda where it has launched a pilot project involving 5,000 refugees – mainly form South Sudan and Democratic Republic of Congo. It hopes to scale up to 25,000 refugees within two years.

At this week's United Nations General Assembly, the World Bank, the United Nations, and the Red Cross teamed up with tech giants Amazon, Microsoft, and Google to announce an unlikely new tool to stop famine before it starts: artificial intelligence. The Famine Action Mechanism (FAM), as they're calling it, is the first global tool dedicated to preventing future famines -- no small news in a world where one in nine people don't have enough food. Building off of previous famine-prediction strategies, the tool will combine satellite data of things like rainfall and crop health with social media and news reports of more human factors, like violence or changing food prices. It will also establish a fund that will be automatically dispersed to a food crisis as soon as it meets certain criteria, speeding up the often-lengthy process for funding famine relief. For a famine to be declared in a country or region, three criteria have to be met: At least one in five households has an extreme lack of food; over 30 percent of children under five have acute malnutrition; and two out of 10,000 people die each day.

WASHINGTON – Tech giants Microsoft, Amazon and Google are joining forces with international organizations to help identify and head off famines in developing nations using data analysis and artificial intelligence, a new initiative unveiled Sunday. Rather than waiting to respond to a famine after many lives already have been lost, the tech firms "will use the predictive power of data to trigger funding" to take action before it becomes a crisis, the World Bank and United Nations announced in a joint statement. "The fact that millions of people -- many of them children -- still suffer from severe malnutrition and famine in the 21st century is a global tragedy," World Bank Group President Jim Yong Kim said in a statement. "We are forming an unprecedented global coalition to say, 'no more.' " Last year more than 20 million people faced famine conditions in Nigeria, Somalia, South Sudan and Yemen, while 124 million people currently live in crisis levels of food insecurity, requiring urgent humanitarian assistance for their survival, the agencies said. Over half of them live in areas affected by conflict.