In real-world classification tasks, each class often comprises multiple finer-grained subclasses. As the subclass labels are frequently unavailable, models trained using only the coarser-grained class labels often exhibit highly variable performance across different subclasses. This phenomenon, known as hidden stratification, has important consequences for models deployed in safety-critical applications such as medicine. We propose GEORGE, a method to both measure and mitigate hidden stratification even when subclass labels are unknown. We first observe that unlabeled subclasses are often separable in the feature space of deep models, and exploit this fact to estimate subclass labels for the training data via clustering techniques. We then use these approximate subclass labels as a form of noisy supervision in a distributionally robust optimization objective. We theoretically characterize the performance of GEORGE in terms of the worst-case generalization error across any subclass. We empirically validate GEORGE on a mix of real-world and benchmark image classification datasets, and show that our approach boosts worst-case subclass accuracy by up to 15 percentage points compared to standard training techniques, without requiring any information about the subclasses.

Our goal is to deploy a high-accuracy system starting with zero training examples. We consider an on-the-job setting, where as inputs arrive, we use real-time crowdsourcing to resolve uncertainty where needed and output our prediction when confident. As the model improves over time, the reliance on crowdsourcing queries decreases. We cast our setting as a stochastic game based on Bayesian decision theory, which allows us to balance latency, cost, and accuracy objectives in a principled way. Computing the optimal policy is intractable, so we develop an approximation based on Monte Carlo Tree Search. We tested our approach on three datasets--named-entity recognition, sentiment classification, and image classification. On the NER task we obtained more than an order of magnitude reduction in cost compared to full human annotation, while boosting performance relative to the expert provided labels.

On July 13, Boston's Merlin Labs announced that it would be working with the US Air Force to add autonomy to the C-130J Super Hercules cargo transport plane. Merlin's technology is a kind of advanced auto-copilot, designed to take over the responsibilities of one crew member in flight while being supervised by a human pilot. If the technology delivers as promised, it will allow planes that normally fly with two human pilots to operate with just one, and could even allow single-seater planes to fly fully autonomously. The same day that Merlin announced its partnership with the Air Force, it also announced a second round of $105 million in funding, which combined with a first round means the company has $130 million of runway to develop its technologies. This funding, says Merlin Labs CEO Matthew George, will help the company continue to develop "the world's most capable, safest and flexible pilot, that will eventually enable very large aircraft to fly with reduced crew and small aircraft to fly totally uncrewed."

Autonomous flight is a grand challenge in aviation -- and a gold mine. The first company to crack it at scale stands to reap handsome profits from transportation and logistics alone. In 2020, the size of the global cargo airline industry was $110.8 billion, according to Statista, and one source estimates that it'll generate hundreds of billions in revenue by 2027. Xwing is one of the startups chasing after self-flying planes, as is Reliable Robotics, Pyka and the unicorn Volocopter. Roughly a year ago, Boston-based Merlin Labs emerged from stealth with an autonomous flight system designed to be installed in existing aircraft.

The early '90s were a nostalgic and unique time in gaming. As a bookish 11-year-old in braces and a hot pink jogging set who didn't fit in on the playground, my only comforts were my best friend Denise, our after-school episodes of Star Trek: The Next Generation over Pringles, and the family desktop, in all of its beige glory. While most of my classmates in elementary school argued over whether the Super Nintendo or the Sega Genesis was the superior console, I turned my attention to PC gaming, then just an afterthought. Most kids in my grade only used their family computers to boot up Mavis Beacon Teaches Typing or Encarta to finish their book reports. Because my mom's longtime boyfriend was a computer repair person who often brought his work home, I grew up around piles of IDE cables and optical disc drives, excited about weekend trips to Fry's Electronics to check out of the large, glossy cardboard boxes housing the latest entertaining floppies and CDs.

The history of AI is often told as the story of machines getting smarter over time. What's lost is the human element in the narrative, how intelligent machines are designed, trained, and powered by human minds and bodies. In this six-part series, we explore that human history of AI--how innovators, thinkers, workers, and sometimes hucksters have created algorithms that can replicate human thought and behavior (or at least appear to). While it can be exciting to be swept up by the idea of super-intelligent computers that have no need for human input, the true history of smart machines shows that our AI is only as good as we are. In the 1970s, Dr. Geoffrey Franglen of St. George's Hospital Medical School in London began writing an algorithm to screen student applications for admission.

A well -funded start up growing the data science function implementing AI solutions based on Machine Learning algorithms with global clients. Apply for job To apply for this job email your details to george@mbnsolutions.com Company Details Ref: 7301-GD Apply for job To apply for this job email your details to george@mbnsolutions.com To apply for this job email your details to george@mbnsolutions.com

A well-funded Tech start up in Bristol looking for a Data Scientist to join a growing team. You will be able to implement high level machine learning techniques and big data technologies on a unique platform. To apply for this job email your details to george@mbnsolutions.com

After an introduction to LOGO thinking and language, the benefits of children writing simple AI programs using the proper tools were described. Finally, the ways in which an Al system designed for education can interact with children were discussed. These ideas should be implemented and tested with children. Only then will the effects on education be known.

He then solves the simpler problem and uses the information obtained in the solution of the simpler problem in the solution of the original complex problem.