Standard artificial neural networks suffer from the well-known issue of catastrophic forgetting, making continual or lifelong learning problematic. Recently, numerous methods have been proposed for continual learning, but due to differences in evaluation protocols it is difficult to directly compare their performance. To enable more meaningful comparisons, we identified three distinct continual learning scenarios based on whether task identity is known and, if it is not, whether it needs to be inferred. Performing the split and permuted MNIST task protocols according to each of these scenarios, we found that regularization-based approaches (e.g., elastic weight consolidation) failed when task identity needed to be inferred. In contrast, generative replay combined with distillation (i.e., using class probabilities as "soft targets") achieved superior performance in all three scenarios. In addition, we reduced the computational cost of generative replay by integrating the generative model into the main model by equipping it with generative feedback connections. This Replay-through-Feedback approach substantially shortened training time with no or negligible loss in performance. We believe this to be an important first step towards making the powerful technique of generative replay scalable to real-world continual learning applications.

Multiple query criteria active learning (MQCAL) methods have a higher potential performance than conventional active learning methods in which only one criterion is deployed for sample selection. A central issue related to MQCAL methods concerns the development of an integration criteria strategy (ICS) that makes full use of all criteria. The conventional ICS adopted in relevant research all facilitate the desired effects, but several limitations still must be addressed. For instance, some of the strategies are not sufficiently scalable during the design process, and the number and type of criteria involved are dictated. Thus, it is challenging for the user to integrate other criteria into the original process unless modifications are made to the algorithm. Other strategies are too dependent on empirical parameters, which can only be acquired by experience or cross-validation and thus lack generality; additionally, these strategies are counter to the intention of active learning, as samples need to be labeled in the validation set before the active learning process can begin. To address these limitations, we propose a novel MQCAL method for classification tasks that employs a third strategy via weighted rank aggregation. The proposed method serves as a heuristic means to select high-value samples of high scalability and generality and is implemented through a three-step process: (1) the transformation of the sample selection to sample ranking and scoring, (2) the computation of the self-adaptive weights of each criterion, and (3) the weighted aggregation of each sample rank list. Ultimately, the sample at the top of the aggregated ranking list is the most comprehensively valuable and must be labeled. Several experiments generating 257 wins, 194 ties and 49 losses against other state-of-the-art MQCALs are conducted to verify that the proposed method can achieve superior results.

The US government is only dabbling in artificial intelligence at the moment. It might make a larger commitment before long, however. A bipartisan group of senators (Brian Schatz, Cory Gardner and Rob Portman) have introduced an AI in Government Act that would increase federal AI adoption by both including AI in data-related plans and supplying the resources to make those plans a reality. Thankfully, this isn't just a question of throwing money at the problem -- it would have multiple government organizations shift more attention to the emerging technology. The General Services Administration would have additional powers to both research AI policy and provide relevant expertise to agencies.

Despite its promise, the growing field of Artificial Intelligence (AI) is experiencing a variety of growing pains. In addition to the problem of bias I discussed in a previous article, there is also the'black box' problem: if people don't know how AI comes up with its decisions, they won't trust it. In fact, this lack of trust was at the heart of many failures of one of the best-known AI efforts: IBM Watson – in particular, Watson for Oncology. If oncologists had understood how Watson had come up with its [diagnoses] – what the industry refers to as'explainability' – their trust level may have been higher. "The more complex a system is, the less explainable it will be," says John Zerilli, postdoctoral fellow at University of Otago and researcher into explainable AI. "If you want your system to be explainable, you're going to have to make do with a simpler system that isn't as powerful or accurate."

The Canadian government's proposed use of artificial intelligence to assess refugee claims and immigration applications could jeopardize applicants' human rights, says a group of privacy experts. The warnings are raised in a new report from the Citizen Lab, a group of civic-minded technological and privacy policy researchers at the University of Toronto's Munk School of Global Affairs, and the University of Toronto Faculty of Law's International Human Rights Program. The federal government is already taking steps to make use of artificial intelligence. In May, The Globe and Mail reported that Justice Canada and Immigration, Refugees and Citizenship Canada (IRCC) were piloting an artificial-intelligence program to assist with preremoval risk assessments and immigration applications on humanitarian grounds. The report being released Wednesday cautions that experimenting with these technologies in the immigration and refugee system amounts to a "high-risk laboratory," as many of these applications come from some of the world's most vulnerable people, including those fleeing persecution and war zones.

Pessimists on the left tell us the unfolding revolution in mobile supercomputing, intelligent robotics, self-driving vehicles and the like will eliminate most middle class jobs, relegate the masses to waiting on tables and drive inequality to unconscionable dimensions. Don't worry, invest in a skills-based education and you will still be able to find good paying work -- the real threat is to your private thoughts and conscience. All these innovations are driven by artificial intelligence. Computer programs that can collect and process huge amounts of data at lightening speeds and most importantly, independently learn -- modify their own algorithms to better accomplish assigned goals. They can perform tasks as simple assembling and dispatching your next Amazon Prime shipment or as complex as providing your prospective in-laws with a detailed assessment of your future earnings prospects, political views, winning virtues and troubling faults.

The U.S. could face heightened national security threats and lose its economic edge if the government doesn't step up its game when it comes to artificial intelligence, according to a pair of oversight lawmakers. Will Hurd, R-Texas, and Robin Kelly, D-Ill., on Tuesday published a report detailing the current state of the country's artificial intelligence ecosystem and offering recommendations for how government could steer and accomodate the technology's development in the years ahead. The report is based on a series of hearings examining the government's role in advancing AI hosted earlier this year by the House Oversight Subcommittee on Information Technology, on which Hurd chairs and Kelly serves as ranking member. "[Artificial intelligence] is a topic that's going to transcend and be important beyond this Congress," Hurd said Tuesday during a call with reporters. "I think this report [will] lay a foundation for future focus by Congress and other parts of the government."

Search giant Google is celebrating its own 20th birthday today with a trademark Doodle. Replacing its logo with an occasional animation paying tribute to eminent figures from the worlds of science, the arts and history on their anniversaries is just one of the ways in which the site's programmers can express themselves. Their quirky sense of humour is actually embedded within the software's DNA. If you tell Google to "do a barrel roll", the whole page will spin clockwise at 90 degrees before juddering to a stop. If you search for "the answer to life the universe and everything", you'll be presented with Google's calculator displaying the number 42, an in-joke alluding to Douglas Adams' cult science fiction series The Hitchhiker's Guide to the Galaxy (1978).

Yet the combination of these factors created a milestone in AI history, as it had a profound impact on real-world applications and the successful deployment of various AI techniques that have been in the works for a very long time, particularly neural networks.g I shared these remarks in various contexts during the course of preparing this article. The audiences ranged from AI and computer science to law and public-policy researchers with an interest in AI. What I found striking is the great interest in this discussion and the comfort, if not general agreement, with the remarks I made. I did get a few "I beg to differ" responses though, all centering on recent advancements relating to optimizing functions, which are key to the successful training of neural networks (such as results on stochastic gradient descent, dropouts, and new activation functions). The objections stemmed from not having named them as breakthroughs (in AI). My answer: They all fall under the enabler I outlined earlier: "increasingly sophisticated statistical and optimization techniques for fitting functions." Follow up question: Does it matter that they are statistical and optimization techniques, as opposed to classical AI techniques?

Ask poverty attorney Joanna Green Brown for an example of a client who fell through the cracks and lost social services benefits they may have been eligible for because of a program driven by artificial intelligence (AI), and you will get an earful. There was the "highly educated and capable" client who had had heart failure and was on a heart and lung transplant wait list. The questions he was presented in a Social Security benefits application "didn't encapsulate his issue" and his child subsequently did not receive benefits. "It's almost impossible for an AI system to anticipate issues related to the nuance of timing," Green Brown says. Then there's the client who had to apply for a Medicaid recertification, but misread a question and received a denial a month later.