In the rapidly advancing arena of large language models (LLMs), a key challenge is to enhance their capabilities amid a looming shortage of high-quality training data. Our study starts from an empirical strategy for the light continual training of LLMs using their original pre-training data sets, with a specific focus on selective retention of samples that incur moderately high losses. These samples are deemed informative and beneficial for model refinement, contrasting with the highest-loss samples, which would be discarded due to their correlation with data noise and complexity. We then formalize this strategy into a principled framework of Instance-Reweighted Distributionally Robust Optimization (IR-DRO). IR-DRO is designed to dynamically prioritize the training focus on informative samples through an instance reweighting mechanism, streamlined by a closed-form solution for straightforward integration into established training protocols. Through rigorous experimentation with various models and datasets, our findings indicate that our sample-targeted methods significantly improve LLM performance across multiple benchmarks, in both continual pre-training and instruction tuning scenarios. Our codes are available at https://github.com/VITA-Group/HardFocusTraining.

Sundar Pichai told '60 Minutes' that the state of the technology is still somewhat of a black box to researchers. Researchers are accusing Google and CBS News of overestimating the capabilities of artificial intelligence (AI) following an interview between the Alphabet CEO Sundar Pichai and "60 Minutes." During the recent interview, Pichai claimed that AI programs developed by Google had displayed "emergent properties," or the ability to learn unexpected skills they were not trained on, puzzling researchers. For example, Google tech executive James Manyika claimed the company's AI had learned the language of Bengali without significant implementation of the information beforehand. "We discovered that with very few amounts of prompting in Bengali," Manyika said, "it can now translate all of Bengali."

The future of west virginia politics is uncertain. The state has been trending Democratic for the last decade, but it's still a swing state. Democrats are hoping to keep that trend going with Hillary Clinton in 2016. But Republicans have their own hopes and dreams too. They're hoping to win back some seats in the House of Delegates, which they lost in 2012 when they didn't run enough candidates against Democratic incumbents.

Glossary of common statistical, machine learning, data science terms used commonly in industry. Explanation has been provided in plain and simple English.

Humans are good at figuring out the intricate physics of object-ice interactions that affect how giant stones slide across a frozen surface. Machines, however, can freeze up in the real world. Curly, a new curling-playing robot, has a better handle on those complexities, thanks to an artificially intelligent brain that can quickly assess and map the icy environment, the state of play and optimal strategies for winning, according to a paper published Wednesday in the journal Science Robotics by a team of roboticists at Korea University in Seoul. The white, turtle-shaped robot, recently beat out elite curling South Korean players in a series of four matches, losing only once, according to the study. Curly's triumph is the latest example of machines besting humans at their own games--but it marks an important step forward: Other big wins for the robots have been in digital environments, where the physics of the real world didn't get in the way.

A new study published in the journal Cell Systems on November 20, 2019, reports the use of machine learning to help form complex cell architectures from pluripotent stem cells, a sophisticated technology that could solve multiple issues that currently hampers the production of artificial tissues and organs. Medical scientists faced with irreparably damaged organs have long wanted to know how to stimulate their regeneration or to replace them with new ones, to prolong survival and to provide improved quality of life. Another equally important area of research involves creating artificial tissues which are identical to those in the body, in order to help understand how disease processes evolve and which drugs can be used to treat such disorders. This means that scientists must know how to direct the development of stem cells in the desired pattern to form multiple tissues in the right way. Pluripotent ('capable of multiple tasks') stem cells are cells that can divide indefinitely or can develop into any of the three germ layers found in the early embryo.

Interpretability of the underlying AI representations is a key raison d'\^{e}tre for Open Learner Modelling (OLM) -- a branch of Intelligent Tutoring Systems (ITS) research. OLMs provide tools for 'opening' up the AI models of learners' cognition and emotions for the purpose of supporting human learning and teaching. Over thirty years of research in ITS (also known as AI in Education) produced important work, which informs about how AI can be used in Education to best effects and, through the OLM research, what are the necessary considerations to make it interpretable and explainable for the benefit of learning. We argue that this work can provide a valuable starting point for a framework of interpretable AI, and as such is of relevance to the application of both knowledge-based and machine learning systems in other high-stakes contexts, beyond education.

Many of us have already decided that things will be different in 2018. We'll eat better, get more exercise, save more money or finally get around to decluttering those closets. But by the time February rolls around, most of us – perhaps as many as 80 percent of the Americans who make New Year's resolutions – will have already given up. Why does our self-control falter, so often leaving us to revert to our old ways? The answer to this question has consequences beyond our waistlines and bank balances.

In the video below, a robot called iCub demonstrates some impressive archery skills. What's remarkable about the robot isn't just its headdress, but how it learns over time to improve its aim until it's able to hit the bull's-eye. Researchers at the Italian Institute of Technology (IIT) in Genova taught the robot how to hold the bow and fire the arrow. A learning algorithm, dubbed "archer" (Augmented Reward Chained Regression) then used visual feedback to gradually improve the robot's aim. The robot, which has visual and physical sensors, is designed to resemble a 3-year-old-child and mimic methods of learning.