The shift of public debate to the digital sphere has been accompanied by a rise in online hate speech. While many promising approaches for hate speech classification have been proposed, studies often focus only on a single language, usually English, and do not address three key concerns: post-deployment performance, classifier maintenance and infrastructural limitations. In this paper, we introduce a new human-in-the-loop BERT-based hate speech classification pipeline and trace its development from initial data collection and annotation all the way to post-deployment. Our classifier, trained using data from our original corpus of over 422k examples, is specifically developed for the inherently multilingual setting of Switzerland and outperforms with its F1 score of 80.5 the currently best-performing BERT-based multilingual classifier by 5.8 F1 points in German and 3.6 F1 points in French. Our systematic evaluations over a 12-month period further highlight the vital importance of continuous, human-in-the-loop classifier maintenance to ensure robust hate speech classification post-deployment.

In the scenario of black-box adversarial attack, the target model's parameters are unknown, and the attacker aims to find a successful adversarial perturbation based on query feedback under a query budget. Due to the limited feedback information, existing query-based black-box attack methods often require many queries for attacking each benign example. To reduce query cost, we propose to utilize the feedback information across historical attacks, dubbed example-level adversarial transferability. Specifically, by treating the attack on each benign example as one task, we develop a meta-learning framework by training a meta-generator to produce perturbations conditioned on benign examples. When attacking a new benign example, the meta generator can be quickly fine-tuned based on the feedback information of the new task as well as a few historical attacks to produce effective perturbations. Moreover, since the meta-train procedure consumes many queries to learn a generalizable generator, we utilize model-level adversarial transferability to train the meta-generator on a white-box surrogate model, then transfer it to help the attack against the target model. The proposed framework with the two types of adversarial transferability can be naturally combined with any off-the-shelf query-based attack methods to boost their performance, which is verified by extensive experiments.

Artificial intelligence is all over the news. When ChatGPT, OpenAI's new chatbot, was released last month it seemed, finally, to match the hype that generative A.I. has been promising for years--an easy-to-use machine intelligence for the general public. Wild predictions soon followed: The death of search engines, the end of homework, the hollowing-out of creative professions. And, for the first time, such predictions didn't seem abstract. When an A.I. bot like ChatGPT can write a coherent story or essay in seconds, and visual applications like Midjourney, Stable Diffusion and DALL-E 2, produce similarly comprehensible images you have to wonder if human creativity--slow and often uncertain--might be superfluous.

NOAA is tracking Hurricane Fiona by drone, as the storm moves through the Caribbean. In partnership with NOAA, Saildrone Inc. is deploying seven ocean drones to collect data from hurricanes during the 2022 hurricane season. The National Oceanic and Atmospheric Administration has released the first image from its NOAA-21 Visible Infrared Imaging Radiometer Suite (VIIRS) instrument. The recently-launched satellite captured a stunning panoramic view of the Earth, created from swaths of data captured throughout the full globe over a period of 24 hours between Dec. 5 and Dec. 6. Polar-orbiting satellites observe the entire planet twice each day, unlike geostationary satellites.

He also fears that the advancement of machine learning and artificial intelligence to fight potential cybercrime is similarly being used by …

Developing cars that can drive without a human is a unique challenge. There are many kinds of events that fully autonomous vehicles have to be prepared to handle in milliseconds, and mistakes can have serious consequences. Solving these problems requires innovation across a number of fields, such as AI and machine learning, advanced sensors, simulation software that can mimic real-world driving, and computing frameworks to evaluate the system's performance. In 2007, I joined the Urban Challenge that was run by the US Defense Advanced Research Project Agency (Darpa) to test and develop autonomous vehicles (AVs). I vividly remember the first moment our car, Junior, drove by itself in the parking lot using software I was working on just hours before.

Rep. Rob Wittman, R-Va., joins'Fox News Live' to react to the United States Air Force's unveiling of its new B-21 raider stealth bomber, named'The Raider' for Jimmy Doolittle's famous bombing raid on Japan in WW2. With the launch of the Air Force's hypersonic missile off the coast of California earlier this month, the Navy's development of water-based drones over the summer and the recent unveiling of the B-21 Raiders, the U.S. military has made major technological advancements over the past year. The military unveiled the U.S. Air Force B-21 Raider in Palmdale, California. The B-21 Raider is the first new American bomber aircraft in more than three decades. In an email to Fox News Digital, a spokesperson confirmed the Air Force would transition its three-bomber fleet to a two-bomber fleet of B-21s and modernized B-52s.

Artificial intelligence is used in almost all segments today – Fashion, Business, Sports, Manufacturing, Defence, Medicine and more. But, can they bring solutions to all the earth's problems? Can AI frame policies to fight climate change? The answer might be an artificially intelligent robot that can help nations to effectively derive a data-based policy framework. AI is a powerful tool that can address the challenges posed by climate change.

However, with stricter regulations and developments in Artificial Intelligence (AI) and Machine Learning (ML), CISOs may be in a stronger position …

The government of Tamil Nadu on Wednesday issued orders to modernise its forest force by equipping them with modern technologies. The Tamil Nadu Forest Force modernisation project has six components, including a Human Resource Management component to be implemented at a cost of ₹8.55 crore. This would include in-service training programmes for frontline forest personnel and improvements to the training infrastructure at the Tamil Nadu Forest Academy, according to a release. Forest personnel shall be given specialised training and exposure to best practices around the country and abroad. Under component II, the government has proposed to improve the Forest Department's information and communication technology at a cost of ₹40 lakh.