In supervised classification tasks, models are trained to predict a label for each data point. In real-world datasets, these labels are often noisy due to annotation errors. While the impact of label noise on the performance of deep learning models has been widely studied, its effects on the networks' hidden representations remain poorly understood. We address this gap by systematically comparing hidden representations using the Information Imbalance, a computationally efficient proxy of conditional mutual information. Through this analysis, we observe that the information content of the hidden representations follows a double descent as a function of the number of network parameters, akin to the behavior of the test error. We further demonstrate that in the underparameterized regime, representations learned with noisy labels are more informative than those learned with clean labels, while in the overparameterized regime, these representations are equally informative. Our results indicate that the representations of overparameterized networks are robust to label noise. We also found that the information imbalance between the penultimate and pre-softmax layers decreases with cross-entropy loss in the overparameterized regime. This offers a new perspective on understanding generalization in classification tasks. Extending our analysis to representations learned from random labels, we show that these perform worse than random features. This indicates that training on random labels drives networks much beyond lazy learning, as weights adapt to encode labels information.

Forecasting oil production from oilfields with multiple wells is an important problem in petroleum and geothermal energy extraction, as well as energy storage technologies. The accuracy of oil forecasts is a critical determinant of economic projections, hydrocarbon reserves estimation, construction of fluid processing facilities, and energy price fluctuations. Leveraging generative AI techniques, we model time series forecasting of oil and water productions across four multi-well sites spanning four decades. Our goal is to effectively model uncertainties and make precise forecasts to inform decision-making processes at the field scale. We utilize an autoregressive model known as TimeGrad and a variant of a transformer architecture named Informer, tailored specifically for forecasting long sequence time series data. Predictions from both TimeGrad and Informer closely align with the ground truth data. The overall performance of the Informer stands out, demonstrating greater efficiency compared to TimeGrad in forecasting oil production rates across all sites.

SPAN develops products that accelerate the rapid adoption of renewable energy in the home. The flagship SPAN Smart Panel is the first true evolution for the traditional home electric panel, harnessing enhanced technology for metering, monitoring, and control. An expanded product suite of intelligent, integrated solutions radically lowers the cost and complexity of energy upgrades–including solar, batteries and EVs–empowering homeowners to be active, resilient and informed players in the energy market. We are seeking a Staff Data Engineer to join our team building the cloud based glue that gives our users access to the rich information and controls provided by the SPAN Panel. Our system collects a large volume of energy monitoring data that needs to be stored, processed, and exposed in different ways for different end users.

To assess the state of the art and identify issues requiring further investigation, a workshop on qualitative and abstract probability was held during the third week of November 1993. This workshop brought together a mix of active researchers from academia, industry, and government interested in the practical and theoretical impact of these abstractions on techniques, methods, and tools for solving complex AI tasks. The result was a set of specific recommendations on the most promising and important avenues for future research. The workshop, entitled "Putting Qualitative and Abstract Probability to Work," gathered active researchers from university, industry, and government to assess the state of the art and make recommendations for future research. The event was sponsored by the Palo Alto Laboratory of Rockwell Science Center.

The Sixth Annual Knowledge-Based Software Engineering Conference (KBSE-91) was held at the Sheraton University Inn and Conference Center in Syracuse, New York, from Sunday afternoon, 22 September, through midday Wednesday, 25 September. The KBSE field is concerned with applying knowledge-based AI techniques to the problems of creating, understanding, and maintaining very large software systems. The Sixth Annual Knowledge-Based Software Engineering Conference (KBSE-91) was held at the Sheraton University Inn and Conference Center in Syracuse, New York, from Sunday afternoon, 22 September, through midday Wednesday, 25 September. This conference was sponsored by Rome Laboratory (previously Rome Air Development Center) and was held in cooperation with the Association for Computing Machinery and the American Association for Artificial Intelligence. The origin of KBSE-91 is as follows: In 1983, Rome Air Development Center published a report calling for the development of a knowledgebased software assistant (KBSA) that would use AI techniques to support all phases of the software development process (Green et al. 1986).

The Third International Conference on Artificial Intelligence Planning Systems (AIPS-96) was held in Edinburgh, Scotland, from 29 to 31 May 1996. The main gathering of researchers in AI and planning and scheduling, the conference promoted the practical applications of planning technologies. Details of the conference papers and sessions are provided as well as information on the Defense Advanced Research Projects Agency-Rome Laboratory Planning Initiative. Previous conferences were held at the University of Maryland in June 1992 (AIPS-92), organized by Jim Hendler and Drew McDermott, and the University of Chicago in June 1994 (AIPS-94), organized by Kristian Hammond. The generation of plans and related fields, such as scheduling, resource allocation, and reasoning about action, have a long research tradition in AI.

We thank Kaz Kulikowski and Priscilla Rasmussen of Rutgers Universitv for their areat helD in organization of the Workshop One of the particularly sat,isfying aspects of this Workshop was the attendance by a large number of graduate students and medical fellows active in AIM research; this was made possible by a generous grant, from AAAI Chris Putnam and OS17 AI graduate students worked very hard in t,aking care of a number of details. A nurnber of systems for medical decision making, experimenting with new ideas for knowledge organization and problem solving, have been built there. The College of Medicine has just started a center for research in knowledge-based systems in medicine. Thus, after a number of years when the Workshop had been hosted by the AIM groups of MIT, TJniversity of Pittsburgh, Rutgers, and Stanford, sometimes in conjunction with major AI and medical computiug conferences, holding the Workshop at Ohio State University was an indication of a broader base for research activities in AIM. This report gives an overview of the Workshop discussions, without any claim of being complete or even representative-a report, of this kind can only be an impressionistic account.

The emergence of the blackboard architecture as a widely used paradigm for problem solving led us and other members of the blackboard research community to organize a workshop. The workshop was held during the 1987 American Association for Artificial Intelligence Conference in Seattle. The main purpose of the workshop was to highlight the advances in blackboard architectures since the introduction of the paradigm in Hearsay-II and identify issues relevant to future blackboard system research. This article describes the issues raised and the discussions in each of the five workshop panels. Highlights of the discussions follow.

The author acknowledges the kind cooperation of Professor Aravind Joshi, IJCAI program chairman, in extending the opportunity to produce this timely panel discussion The panelists included Dr Jack Williams. His presentation pinpointed the world forces of change, the government role in fostering efficient technological innovation, and the need to adapt to flexible manufacturing quickly. In discussing the AI industry, he said, LLThere are many similarities between AI and biotechnology, namely, the entrepreneurship and many startup firms, few products yet, but much commercial potential, a shortage of qualified talent, and a potential to create vast social change. The aspects of world forces of change are serious in that they threaten the livelihood of the U.S. economy because 70% of the U.S. output is in world markets. Abstract The consensus of government, academic, and industry leaders widely supports the strategic positioning of U.S. and Japanese research and development in mutually beneficial, two-way flows of innovation This report is derived from the IJCAI panel titled U S and Japanese Cooperation in AI and R&D Opportunities, held August 23, 1985 at the University of California at Los Angeles This panel discussed the sensitive topic of alternatives to nationalistic competitive strategies that have contributed to an extreme trade deficit surpassing $40 billion in 1985 The ideas offered by the panelists shed light on ways our countries' respective scientific communities can blend talents to achieve the best results in reducing trade frictions Each country has designated AI research as a key to unlock years of generations of technology and has directed billions of dollars to fund this development The most recognized projects are the U.S. Microelectronics Technology Computer Consortium (MCC) and Japan's Fifth Generation Computer Project (ICOT).

In numerous distributed environments, including today's World Wide Web, enterprise data management systems, large science projects, and the emerging semantic web, applications will inevitably use the information described by multiple ontologies and schemas. We organized the Workshop on Semantic Integration at the Second International Semantic Web Conference to bring together different communities working on the issues of enabling integration among different resources. The workshop generated a lot of interest and attracted more than 70 participants. Interoperability among applications depends critically on the ability to map between them. Semantic integration issues have now become a key bottleneck in the deployment of a wide variety of information management applications.