We study the well-motivated problem of online distribution shift in which the data arrive in batches and the distribution of each batch can change arbitrarily over time. Since the shifts can be large or small, abrupt or gradual, the length of the relevant historical data to learn from may vary over time, which poses a major challenge in designing algorithms that can automatically adapt to the best ``attention span'' while remaining computationally efficient. We propose a meta-algorithm that takes any network architecture and any Online Learner (OL) algorithm as input and produces a new algorithm which provably enhances the performance of the given OL under non-stationarity. Our algorithm is efficient (it requires maintaining only $O(\log(T))$ OL instances) and adaptive (it automatically chooses OL instances with the ideal ``attention'' length at every timestamp). Experiments on various real-world datasets across text and image modalities show that our method consistently improves the accuracy of user specified OL algorithms for classification tasks. Key novel algorithmic ingredients include a \emph{multi-resolution instance} design inspired by wavelet theory and a cross-validation-through-time technique. Both could be of independent interest.

When deploying machine learning systems to the wild, it is highly desirable for them to effectively leverage prior knowledge to the unfamiliar domain while also firing alarms to anomalous inputs. In order to address these requirements, Universal Domain Adaptation (UniDA) has emerged as a novel research area in computer vision, focusing on achieving both adaptation ability and robustness (i.e., the ability to detect out-of-distribution samples). While UniDA has led significant progress in computer vision, its application on language input still needs to be explored despite its feasibility. In this paper, we propose a comprehensive benchmark for natural language that offers thorough viewpoints of the model's generalizability and robustness. Our benchmark encompasses multiple datasets with varying difficulty levels and characteristics, including temporal shifts and diverse domains. On top of our testbed, we validate existing UniDA methods from computer vision and state-of-the-art domain adaptation techniques from NLP literature, yielding valuable findings: We observe that UniDA methods originally designed for image input can be effectively transferred to the natural language domain while also underscoring the effect of adaptation difficulty in determining the model's performance.

Social media users hurled criticism at Microsoft this week for what many thought was an AI-generated obituary for NBA player Brandon Hunter on its website MSN. The controversy began after the obituary -- which had a headline that read "Brandon Hunter useless at 42" written by "Editor" -- appeared on the Microsoft-owned platform after Hunter's death on Tuesday. The obituary goes on to refer to the former Boston Celtics and Orlando Magic player having been "handed away on the age of 42" and claimed he "performed in 67 video games over two seasons and achieved a career-high of 17 factors in a recreation in opposition to the Milwaukee Bucks in 2004." The post appeared to follow a similar format to a story on TMZ Sports, Futurism noted, "albeit with altered punctuation and a use of synonyms so liberal that the result is essentially incomprehensible." You can compare both the obituary containing the error and the TMZ Sports story here.

Nelson left HuffPost at the end of 2018 to work full time on the concept for the game. While at HuffPost, Nelson wrote a newsletter with a humorist spin on the day's political headlines. But Nelson says the news cycle can often feel like a "SportsCenter" highlight reel, continually cycling through provocative tweets and sound bites from politicians. And, because of that, Nelson said the average reader may think American politics are nothing more than a "Twitter-fueled boxing match, where occasionally there's a Supreme Court nominee or a major bill."

A website that uses machine-learning to quickly turn innocuous photos of famous and everyday women into realistic deepfake nudes is racking up howls of outrage--and millions of page views. The year-old site has garnered more than 38 million hits since the start of 2021, The Huffington Post reported, with five million in June alone, according to BBC News. HuffPo declined to name the website, but the BBC identified it as Deepsukebe, with both outlets referring to language on the site claiming its mission is to'make all men's dreams come true.' On its now-suspended Twitter page, Deepsukebe referred to itself as an'AI-leveraged nudifier.' It claims it doesn't save the fake photos it generates, but an'incentive program' rewards posters who share links of their deepfakes. Users who get enough people to click on them can'nudify' more pictures faster.

How Tesla and Waymo are tackling a major problem for self-driving cars: Data Sean O'Kane, The Verge In order for cars to drive themselves, the vehicles and their systems require loads of data. And gathering those details are one of the main goals for companies developing the autonomous cars that will eventually take us to the office. The Verge takes a look at how two of the main players in the self-driving space -- Tesla and Waymo -- are gathering gobs of data in very different ways. Facebook didn't seem to care I was being sexually harassed until I decided to write about it Jesselyn Cook, HuffPost There were a few stories about issues with Facebook's reporting tools this week, but this one from a reporter at HuffPost is by far the most eye-opening. You may recall the Neopets hype in the early 2000s, but what you might not know is that the company, as The Outline describes it, "employed business practices directly connected to the Church of Scientology."

Disney is all about the details. Sure, every Disney fan knows that the rides, snacks, and the parks themselves are meticulously designed and maintained so visitors can be delighted at every turn, but did you know that this very thoughtful mentality also extends to the park's staff as well? Take, for example, the fact that Disney park employees are banned from ever pointing with just their index finger. Ready to move this morning! Sure, it may seem like a minute detail, but as it turns out there are two distinct conspiracy theories on why park employees are mandated to do the "double Disney point," including Walt Disney's smoking habits and simply being polite.

Now's your chance to prove it with The Huffington Post's new weekly audio news quiz, powered by the Google Assistant. The 10-question, multiple-choice quiz, which was built by AOL Alpha, is available on Google's voice-activated speaker, Google Home. You can access the quiz by saying, "Ok Google, talk to the HuffPost Headline Quiz." A new quiz will be available every Friday morning and will challenge you on a range of different topics, from politics to pop culture to world news. For the serious HuffPost fans, we've also added a bonus question at the end of each quiz that tests how closely you've been following our lead headlines each week.

Still, the fact remains that the police deployed a robot with the intent to kill a suspect, which some say sets a worrying precedent about lethal force that's completely separate from the ethical considerations of shooting a gun. "The legal framework for police use of force assumes human decision-making about immediate human threats," Elizabeth Joh, a professor of law specializing in policing and technology at the University of California Davis, told HuffPost. "What does that mean when the police are far away from a suspect posing a threat? What does'objectively reasonable' lethal robotic force look like?" Joh recognizes that this wasn't a complex killing machine, but she argues its deployment indicates how easy it would be for police to launch more advanced weaponry without oversight.

I learned how to debug programs and also learned how to change x and y coordinates,