Continual learning aims to incrementally acquire new concepts in data streams while resisting forgetting previous knowledge.With the rise of powerful pre-trained models (PTMs), there is a growing interest in training incremental learning systems using these foundation models, rather than learning from scratch. Existing works often view PTMs as a strong initial point and directly apply parameter-efficient tuning (PET) in the first session for adapting to downstream tasks.In the following sessions, most methods freeze model parameters for tackling forgetting issues. However, applying PET directly to downstream data cannot fully explore the inherent knowledge in PTMs.Additionally, freezing the parameters in incremental sessions hinders models' plasticity to novel concepts not covered in the first session. To solve the above issues, we propose a Slow And Fast parameter-Efficient tuning (SAFE) framework.In particular, to inherit general knowledge from foundation models, we include a transfer loss function by measuring the correlation between the PTM and the PET-applied model.After calibrating in the first session, the slow efficient tuning parameters can capture more informative features, improving generalization to incoming classes.Moreover, to further incorporate novel concepts, we strike a balance between stability and plasticity by fixing slow efficient tuning parameters and continuously updating the fast ones.Specifically, a cross-classification loss with feature alignment is proposed to circumvent catastrophic forgetting.During inference, we introduce an entropy-based aggregation strategy to dynamically utilize the complementarity in the slow and fast learners.Extensive experiments on seven benchmark datasets verify the effectiveness of our method by significantly surpassing the state-of-the-art.

The career path of Colombian physicist Luisa Fernanda Chaparro Sierra took her from studying the Higgs Boson at CERN, to using similar machine learning techniques to gauge perceptions of crime in the Colombian capital of Bogota. Chaparro, currently a Research Professor at Tecnológico de Monterrey in Monterrey, México, says that after finishing her Phd, she had the opportunity to be part of the DataLab (Laboratorio de Datos) of the Universidad Nacional de Colombia where she used the techniques of handling large databases to help understand the problem of the perception of security in Bogota via machine learning methods. "At CERN, we handled large amounts of data and to differentiate between signal and background; we used supervised machine learning techniques, so I used similar methods and adapted others for the case of perception of security," she says, adding that DataLab was composed of mathematicians, physicists, and engineers with knowledge in programming and statistics. "We used Twitter as our data source and reviewed tweets that talked about security in the city for a year," Chaparro says, "The goal was to design a model that would allow us to quantify something as subjective as perception." The researchers were also hoping to find a relationship between it and real crimes by comparing the results with the databases provided by the National Police.

… studying the Higgs Boson at CERN, to using similar machine learning techniques to gauge perceptions of crime in the Colombian capital of Bogota.

A near-term solution is for autonomous vehicles to prove themselves in demonstrations and pilots not unlike the way teenagers drive first with parents and instructors. To this end, RMI is developing autonomous vehicle pilots and programs on the ground in Austin, as a means to promote their commercial deployment and legality, and drive adoption by consumers. Determining the bar for driverless vehicle legality is critical. It must be fair when compared to our bar for human drivers. Once an autonomous vehicle and software meet this bar, it should graduate to a full license and be allowed full access to the roads.

Analysts estimate that by 2030, self-driving cars and trucks (autonomous vehicles) could account for as much as 60 percent of US auto sales. But autonomous vehicles are basically computers on wheels, and computers crash all the time. Besides that, computers get hacked every day. So you gotta ask, "Can self-driving cars ever really be safe?" The Short Answer No. Self-driving cars can never really be safe.

Artificial intelligence is an amazing technology that's changing the world in fantastic ways, but anybody who has ever seen the movie Terminator knows that there are some dangers associated with advanced A.I. That's why Elon Musk, Stephen Hawking, and hundreds of other researchers, tech leaders, and scientists have endorsed a list of 23 guiding principles that should steer A.I. development in a productive, ethical, and safe direction. The Asilomar A.I. Principles were developed after the Future of Life Institute brought dozens of experts together for their Beneficial A.I. 2017 conference. The experts, whose ranks consisted of roboticists, physicists, economists, philosophers, and more had fierce debates about A.I. safety, economic impact on human workers, and programming ethics, to name a few. In order to make the final list, 90 percent of the experts had to agree on its inclusion. "What remained was a list of 23 principles ranging from research strategies to data rights to future issues including potential super-intelligence, which was signed by those wishing to associate their name with the list," Future of Life's website explains.

Continued reliance on human operators for managing data centers is a major impediment for them from ever reaching extreme dimensions. Large computer systems in general, and data centers in particular, will ultimately be managed using predictive computational and executable models obtained through data-science tools, and at that point, the intervention of humans will be limited to setting high-level goals and policies rather than performing low-level operations. Data-driven autonomics, where management and control are based on holistic predictive models that are built and updated using generated data, opens one possible path towards limiting the role of operators in data centers. In this paper, we present a data-science study of a public Google dataset collected in a 12K-node cluster with the goal of building and evaluating a predictive model for node failures. We use BigQuery, the big data SQL platform from the Google Cloud suite, to process massive amounts of data and generate a rich feature set characterizing machine state over time. We describe how an ensemble classifier can be built out of many Random Forest classifiers each trained on these features, to predict if machines will fail in a future 24-hour window. Our evaluation reveals that if we limit false positive rates to 5%, we can achieve true positive rates between 27% and 88% with precision varying between 50% and 72%. We discuss the practicality of including our predictive model as the central component of a data-driven autonomic manager and operating it on-line with live data streams (rather than off-line on data logs). All of the scripts used for BigQuery and classification analyses are publicly available from the authors' website.

In large and complex planning problems, there will almost inevitably be aspects that are not relevant to a specific problem instance. Thus, identifying and removing irrelevant parts from an instance is one of the most important techniques for scaling up automated planning. We examine the path-based relevance analysis method, which is safe (preserves plan existence and cost) and powerful but has exponential time complexity, and show how to make it run in polynomial time with only a minimal loss of pruning power.