Quality: This paper suffers from a few critical issues. Clarity: The experiment setting ups can be described with more details. Sec 3.2 and 3.4 is missing important information such as the datasets used for conducting the experiments. Significance: Although the quality of the proposed model remains unclear because of the previously mentioned critical issues, it's a significant work because it's the first GAN-based model for spectrogram-to-waveform conversion which seems to be working at some degree. It's significantly over-claimed: 1) claiming state-of-the-art for spectrogram-to-waveform conversion (line 6) with MOS 3.09 is surprising; many previous works are at a much higher level (e.g.

Recent studies suggest that the existing neural models have difficulty handling repeated items in sequential recommendation tasks. However, our understanding of this difficulty is still limited. In this study, we substantially advance this field by identifying a major source of the problem: the single hidden state embedding and static item embeddings in the output softmax layer. Specifically, the similarity structure of the global item embeddings in the softmax layer sometimes forces the single hidden state embedding to be close to new items when copying is a better choice, while sometimes forcing the hidden state to be close to the items from the input inappropriately. To alleviate the problem, we adapt the recently-proposed softmax alternatives such as softmax-CPR to sequential recommendation tasks and demonstrate that the new softmax architectures unleash the capability of the neural encoder on learning when to copy and when to exclude the items from the input sequence. By only making some simple modifications on the output softmax layer for SASRec and GRU4Rec, softmax-CPR achieves consistent improvement in 12 datasets. With almost the same model size, our best method not only improves the average NDCG@10 of GRU4Rec in 5 datasets with duplicated items by 10% (4%-17% individually) but also improves 7 datasets without duplicated items by 24% (8%-39%)!

Originally published on Towards AI. The Graph Convolutional Network (GCN) is a revolutionary development in the field of deep learning, demonstrating its versatility and potential for application in addressing real-world problems. One such challenge is traffic prediction, which is a critical issue in transportation. The ability to adapt GCN algorithms for traffic prediction purposes holds immense promise and has the potential to significantly impact the transportation industry. It is important to note that this post assumes a prior understanding of GCN.

Two Stanford heavyweights have weighed in on the fiery AI sentience debate -- and the duo is firmly in the "BS" corner. The wrangle recently rose to a crescendo over arguments about Google's LaMDA system. Developer Blake Lemoine sparked the controversy. Lemoine, who worked for Google's Responsible AI team, had been testing whether the large-language model (LLM) used harmful speech. The 41-year-old told The Washington Post that his conversations with the AI convinced him that it had a sentient mind.

ANJA KASPERSEN: Today I am joined by Joel Rosenthal and Stelios Vassilakis for an irreverently engaging conversation about the impact of artificial intelligence (AI) on democracy, what we can learn from the Athenian agora in preserving what it means to be human in the biodigital realm, and how ethics empower civil engagement. Stelios Vassilakis is co-directing programs and strategic initiatives at the Stavros Niarchos Foundation, which is one of the leading international philanthropic organizations. Stelios is also a classics and modern Greek studies scholar, specializing in the works of Homer. Joel Rosenthal is president of Carnegie Council for Ethics in International Affairs and a distinguished public intellectual of international relations and foreign policy. Before handing the floor over to Joel to guide us through this conversation, I am very curious about these concepts that are guiding the work of both of your institutions. For the Stavros Niarchos Foundation it is empowering humanity, and for Carnegie Council it is about empowering ethics, and obviously there is a strong link between the two. I think in today's world we live in a very distrustful world, a crowded and overheated public space--if we can even identify that space, which we have talked about is a difficult space to even find--and so what we are trying to do at the beginning to empower ethics is first of all just to identify the issues, and to identify these issues, put a name on them, label them, and show them to be issues of competing values and competing interests that would benefit from reflection, dialogue, and discussion, even that question of identification and clarification of these issues and to bring them to the fore in a way that will not necessarily lead to polarization but can lead to constructive dialogue. The second step is to provide thought leadership around these questions--there are people who have dedicated their lives to thinking about some of these issues and to studying these issues; they have great competence and some authority in speaking about these issues--and to identify those people and bring that thought leadership to bear on these questions. Critically, though, it is not just about thinking. It is also about experience. There are people who are actually working on these issues, they are working these problems. It is part of their personal and professional life, and I think that the experience that they have themselves is almost as valuable if not more valuable than those who spend their lives thinking about these issues and creating scholarship around them. So when we talk about thought leadership we're talking about both scholarship and lived experience, Carnegie Council being a place where we can bring that expertise, if you will, to bear on these questions. The third part that is also critical today is to create a community of engagement around these issues.

In the past year, many of the conversations around autonomous vehicles (AVs) have been dominated by the same question: When will self-driving cars be the norm on public roads? While industry leaders talked a big game on AVs monopolizing our roads back in 2016, today some experts have put widespread Level 4 adoption over a decade away. However, even that timeline only works if automakers overcome significant barriers -- both technical and behavioral. The challenge of bringing AVs to consumers will be tougher than anticipated, with a central part of the effort being focused on earning the public's trust. Consumer confidence and mass adoption of AVs go hand in hand.

Early last year, a large European supermarket chain deployed artificial intelligence to predict what customers would buy each day at different stores, to help keep shelves stocked while reducing costly spoilage of goods. The company already used purchasing data and a simple statistical method to predict sales. With deep learning, a technique that has helped produce spectacular AI advances in recent years--as well as additional data, including local weather, traffic conditions, and competitors' actions--the company cut the number of errors by three-quarters. It was precisely the kind of high-impact, cost-saving effect that people expect from AI. But there was a huge catch: The new algorithm required so much computation that the company chose not to use it.

Early last year, a large European supermarket chain deployed artificial intelligence to predict what customers would buy each day at different stores, to help keep shelves stocked while reducing costly spoilage of goods. The company already used purchasing data and a simple statistical method to predict sales. With deep learning, a technique that has helped produce spectacular AI advances in recent years--as well as additional data, including local weather, traffic conditions, and competitors' actions--the company cut the number of errors by three-quarters. It was precisely the kind of high-impact, cost-saving effect that people expect from AI. But there was a huge catch: The new algorithm required so much computation that the company chose not to use it.

Early last year, a large European supermarket chain deployed artificial intelligence to predict what customers would buy each day at different stores, to help keep shelves stocked while reducing costly spoilage of goods. The company already used purchasing data and a simple statistical method to predict sales. With deep learning, a technique that has helped produce spectacular AI advances in recent years--as well as additional data including local weather, traffic conditions, and competitors' actions--the company cut the number of errors by three-quarters. It was precisely the kind of high-impact, cost-saving effect that people expect from AI. But there was a huge catch: The new algorithm required so much computation that the company chose not to use it.

The world's most powerful web-based DevOps lifecycle tool GitLab has released GitLab 13.1 to track coding quality and to stay compliant with the dynamic needs of the DevOps world. GitLab 13.1 is now officially available with extended Alert Management and Automated Coding Reporting features. Those who follow GitLab closely would agree that its acquisition of Gemnasium in 2018 has helped further fortify the security and compliance in open source. The smartest enhancement in GitLab 13.1 is Alert Management; to maintain a record of all application maintenance and to address critical issues in real-time. Simplified Alert Management, Alert Assignments and Slack integration enhance DevOps productivity with faster collaboration and just-in-time principles.