Many of language models' impressive capabilities originate from their in-context learning: based on instructions or examples, they can infer and perform new tasks without weight updates. In this work, we investigate when representations for new tasks are formed in language models, and how these representations change over the course of context. We study two different task representations: those that are ''transferrable'' -- vector representations that can transfer task contexts to another model instance, even without the full prompt -- and simpler representations of high-level task categories. We show that transferrable task representations evolve in non-monotonic and sporadic ways, while task identity representations persist throughout the context. Specifically, transferrable task representations exhibit a two-fold locality. They successfully condense evidence when more examples are provided in the context. But this evidence accrual process exhibits strong temporal locality along the sequence dimension, coming online only at certain tokens -- despite task identity being reliably decodable throughout the context. In some cases, transferrable task representations also show semantic locality, capturing a small task ''scope'' such as an independent subtask. Language models thus represent new tasks on the fly through both an inert, sustained sensitivity to the task and an active, just-in-time representation to support inference.

Biases in large-scale image datasets are known to influence the performance of computer vision models as a function of geographic context. To investigate the limitations of standard Internet data collection methods in low- and middle-income countries, we analyze human-centric image geo-diversity on a massive scale using geotagged Flickr images associated with each nation in Africa. We report the quantity and content of available data with comparisons to population-matched nations in Europe as well as the distribution of data according to fine-grained intra-national wealth estimates. Temporal analyses are performed at two-year intervals to expose emerging data trends. Furthermore, we present findings for an ``othering'' phenomenon as evidenced by a substantial number of images from Africa being taken by non-local photographers. The results of our study suggest that further work is required to capture image data representative of African people and their environments and, ultimately, to improve the applicability of computer vision models in a global context.

Five young inventors from Gambia are the latest students to be denied visas to enter the US for a prestigious international robotics contest in Washington. The teens found the rejection'very disheartening,' their coach, Mucktarr Darboe, said on Tuesday. Darboe, who is also a director in the largely Muslim West African nation's ministry of higher education, said the students were not given a reason for the visa denials in April, and he called the decision'disappointing and unfair.' The US Embassy in Banjul, Gambia's capital, could not immediately be reached for comment. Mucktarr Darboe is pictured with members of Gambia's student team that was denied visas to travel to Washington for a robotics contest Gambia has been through dramatic change in recent months, ousting via elections a longtime dictator, Yahya Jammeh, whose administration was accused of human rights abuses.

HERAT, AFGHANISTAN/DAKAR – Two Afghan girls refused visas to the United States for a robot-building competition said on Tuesday they were mystified by the decision, as the contest's organizers said teams from Iran and Sudan as well as a de facto Syrian team had gained visas. The unusual story of the Afghan all-girl team of robotics students emerged as the United States grapples with the legality of President Donald Trump's order to temporarily ban travel from six Muslim-majority countries. Afghanistan itself is not on the list and Team Afghanistan's robot, unlike its creators, has been allowed entry to the United States. Asked by Reuters on Tuesday why the girls were banned, a U.S. State Department spokesperson cited regulations prohibiting the agency from discussing individual visa cases. So the six team members will watch the ball-sorting machine compete in Washington D.C. via video link during the July 16-18 event from their hometown of Herat, in western Afghanistan, according to the FIRST Global contest organizers.

A school robotics team made up of five students from The Gambia has joined Afghanistan's all-girl robotics team in having their one-week visas rejected by the U.S. in order to attend the First Global Challenge robotics competition in Washington D.C. The five Gambian students on the robotics team will have their robot shipped from the West African country to the First Global Challenge competition in the U.S., which takes place from July 16-18. But the teens will not be able to attend the prestigious international event themselves, Al-Jazeera reports. Despite saying they worked "rigorous shifts" of seven hours during Ramadan to complete the robot, they will instead have to settle for presenting it via Skype. Director of The Gambia's ministry of higher education, research, science and technology, Moktar Darboe, told Al Jazeera that the students are "very disappointed" they won't be able to attend the competition.

Five teenage pupils from The Gambia who built a robot for a prestigious international competition in the United States will not be able to accompany their invention to the event after being denied a visa. The Gambian pupils become the second team of students to be refused entry to attend the FIRST Global robotics event in Washington, DC on July 16-18. On Saturday, it was reported that an all-girls team from Afghanistan were also denied a visa to travel to the US to showcase their creation at the same competition. We're not giving up, despite the challenges we face, we still continue to work hard. Moktar Darboe, director of The Gambia's ministry of higher education, research, science and technology, told Al Jazeera that the team, made up of high school pupils aged 17-18, were "very disappointed".

Billionaires in the world's tech capital Silicon Valley are reportedly preparing for the apocalypse by buying underground bunkers, guns, ammo and motorcycles. Fearful that artificial intelligence will displace so many jobs that there will be a revolt against those responsible for the technology, the are entrepreneurs readying themselves for doomsday like scenarios. Reid Hoffman, the co-founder of the professional social network, LinkedIn, told The New Yorker that he believes more than 50 per cent of billionaires in the Californian tech hub are preparing for the worst. "I own a couple of motorcycles. I have a bunch of guns and ammo. I figure that, with that, I can hole up in my house for some amount of time," he said.