Goto

Collaborating Authors

Results


Viewpoint: What Donald Trump gets wrong about Somalia

BBC News

In our series of letters from African journalists, Ismail Einashe considers how Somalia has become caught up in the US election campaign. President Donald Trump is making Somali-American congresswoman Ilhan Omar one of the bogeywomen of his campaign for re-election to the White House in November - and by proxy her country of birth Somalia. In his most recent attack, at a campaign rally in Tulsa, Oklahoma, he tore into the 37-year-old alleging that she wanted to bring the "anarchy" of Somalia to the US. "She would like to make the government of our country just like the country from where she came - Somalia. And now, she's telling us how to run our country. Ms Omar, who arrived in the US as a child refugee in 1995, is the congressional representative for Minnesota, which includes the city of Minneapolis where African-American George Floyd was killed by police in May, reigniting Black Lives Matter protests. But it was Ms Omar's Somali heritage the president chose to focus on in Tulsa, perhaps to distract from all the turmoil and unrest closer to home. In response Ms Omar said his remarks were "racist". She added that his anger came out of a recent poll that had shown him trailing his rival, Democratic presidential candidate Joe Biden, in her state, which is home to a large Somali-American community. The president described Ms Omar as a "hate-filled, American-bashing socialist", warning she would have a role in shaping the country if Mr Biden were to win. This is despite the fact that the pair are on the opposite ends of the Democratic Party - Ms Omar had been a prominent supporter of Bernie Sanders to win the Democratic ticket. But such rhetoric plays well to his base, so the electoral stage has been set, the cast chosen - and Ms Omar and Somalia have starring roles. In fact they both debuted last year at Mr Trump's rally in North Carolina where the crowd chanted about Ms Omar: "Send her back!


Hackers hijack more than 1,200 accounts in Roblox and flood it with Trump 2020 propaganda

Daily Mail - Science & tech

Hackers have taken over more than 1,200 Roblox accounts, flooding the popular children's video game with Trump 2020 campaign propaganda. Initially, a handful of players reported that they were receiving strange messages in game supporting Donald Trump's reelection campaign, and later other players began noticing avatars wearing Trump-inspired apparel. This included donning red pants, a white shirt with an American flag and bald eagle graphic, and a red baseball cap with white lettering. In the'About' field of each hacked player's main profile page, the message'Ask your parents to vote for Trump this year!#MAGA2020' has replaced the original biographical information. 'Why is my avatar this?' one hacked player posted, according to a report from BBC'I'm not even American.'


Unreal Engine 5′s demo is the next-gen graphics showcase you've been waiting for

Washington Post - Technology News

Travis Scott performed a virtual concert tour in "Fortnite" that was viewed by almost 28 million people over three days. Live concert publication Pollstar speculates that the Travis Scott performance might end up as the biggest live event of the year. The virtual event was so successful, it even caught the attention of Democratic political consultants. The Post asked if Democratic nominee Joe Biden or his campaign had reached out after comments made by former Pete Buttigieg spokeswoman Lis Smith on Monday. Sweeney said no, but added that they wouldn't stop Biden if he decided to create his own map in the game's Creative Mode.


The Joe Biden em Fortnite /em Grand Canyon Hologram Tweet, Explained

Slate

Earlier today, you might have seen this tweet (or some variation of it) make its way across your screen. "Travis Scott's takeover of Fortnite… if we could do that with Joe Biden [for the convention], Joe Biden projected over the Grand Canyon" -- Lis Smith, democratic strategist, on politico live There's a good chance you'll recognize some subset of these words. You'll almost certainly be familiar with Joe Biden and the Grand Canyon, for instance. Maybe you also play Fortnite and listen to Travis Scott. Or perhaps your soul has withered away to the point that you already had Lis Smith's Politico Live appearance circled on your calendar.


How Facebook Works for Trump

The Atlantic - Technology

At the start of the year, Andrew "Boz" Bosworth, who led Facebook's ad team during the 2016 election, wrote that Trump "ran the single best digital ad campaign I've ever seen from any advertiser." Trump's team agrees, of course. But that might not mean what you think it does. He didn't do it via microtargeting--the ability to send highly differentiated audiences just the right messages to change attitudes or inspire action--either, despite conventional understanding. His campaign did so via pure, blunt constancy, using Facebook in exactly the way the tech giant intended: pouring heaps of money and data into Facebook's automated advertising system.


AI-based early warning system to monitor deepfakes, manipulated images

#artificialintelligence

Researchers are utilising artificial intelligence (AI) to develop an early warning system that can identify manipulated images, deepfake videos and disinformation online in 2020 US election. The project is an effort to combat the rise of coordinated social media campaigns to incite violence, sew discord and threaten the integrity of democratic elections. According to the study, published in the journal Bulletin of the Atomic Scientists, the scalable, automated system uses content-based image retrieval and applies computer vision-based techniques to root out political memes from multiple social networks. "Memes are easy to create and even easier to share. When it comes to political memes, these can be used to help get out the vote, but they can also be used to spread inaccurate information and cause harm," said study researcher Tim Weninger, Associate Professor at the University of Notre Dame in the US.


AI can predict your future behaviour with powerful new simulations

#artificialintelligence

The US presidential election campaign is in its final days. Donald Trump is behind in the polls and the pundits are predicting a win for his Democrat challenger, former vice president Joe Biden. He boasts that he will win again. With two weeks to go, his campaign unleashes an offensive in the crucial swing states: adverts, Facebook posts, WhatsApp groups and tweets. They warn of violent crime and civil unrest driven by immigrants and gangs, playing up Trump's endorsement by evangelicals and smearing Biden as a closet atheist. The initiative works and Trump snatches another unlikely victory.


Nonparametric Deconvolution Models

arXiv.org Machine Learning

We describe nonparametric deconvolution models (NDMs), a family of Bayesian nonparametric models for collections of data in which each observation is the average over the features from heterogeneous particles. For example, these types of data are found in elections, where we observe precinct-level vote tallies (observations) of individual citizens' votes (particles) across each of the candidates or ballot measures (features), where each voter is part of a specific voter cohort or demographic (factor). Like the hierarchical Dirichlet process, NDMs rely on two tiers of Dirichlet processes to explain the data with an unknown number of latent factors; each observation is modeled as a weighted average of these latent factors. Unlike existing models, NDMs recover how factor distributions vary locally for each observation. This uniquely allows NDMs both to deconvolve each observation into its constituent factors, and also to describe how the factor distributions specific to each observation vary across observations and deviate from the corresponding global factors. We present variational inference techniques for this family of models and study its performance on simulated data and voting data from California. We show that including local factors improves estimates of global factors and provides a novel scaffold for exploring data.


Outcome Correlation in Graph Neural Network Regression

arXiv.org Machine Learning

Graph neural networks aggregate features in vertex neighborhoods to learn vector representations of all vertices, using supervision from some labeled vertices during training. The predictor is then a function of the vector representation, and predictions are made independently on unlabeled nodes. This widely-adopted approach implicitly assumes that vertex labels are independent after conditioning on their neighborhoods. We show that this strong assumption is far from true on many real-world graph datasets and severely limits predictive power on a number of regression tasks. Given that traditional graph-based semi-supervised learning methods operate in the opposite manner by explicitly modeling the correlation in predicted outcomes, this limitation may not be all that surprising. Here, we address this issue with a simple and interpretable framework that can improve any graph neural network architecture by modeling correlation structure in regression outcome residuals. Specifically, we model the joint distribution of outcome residuals on vertices with a parameterized multivariate Gaussian, where the parameters are estimated by maximizing the marginal likelihood of the observed labels. Our model achieves substantially boosts the performance of graph neural networks, and the learned parameters can also be interpreted as the strength of correlation among connected vertices. To allow us to scale to large networks, we design linear time algorithms for low-variance, unbiased model parameter estimates based on stochastic trace estimation. We also provide a simplified version of our method that makes stronger assumptions on correlation structure but is extremely easy to implement and provides great practical performance in several cases.


Discriminative Metric Learning by Neighborhood Gerrymandering

Neural Information Processing Systems

We formulate the problem of metric learning for k nearest neighbor classification as a large margin structured prediction problem, with a latent variable representing the choice of neighbors and the task loss directly corresponding to classification error. We describe an efficient algorithm for exact loss augmented inference,and a fast gradient descent algorithm for learning in this model. The objective drives the metric to establish neighborhood boundaries that benefit the true class labels for the training points. Our approach, reminiscent of gerrymandering (redrawing of political boundaries to provide advantage to certain parties), is more direct in its handling of optimizing classification accuracy than those previously proposed. In experiments on a variety of data sets our method is shown to achieve excellent results compared to current state of the art in metric learning.