One belief underlying the power-hungry approach to machine learning advanced by OpenAI and Mistral AI is that an artificial intelligence model must review its entire dataset before spitting out new insights. Sepp Hochreiter, an early pioneer of the technology who runs an AI lab at Johannes Kepler University in Linz, Austria, has a different view, one that requires far less cash and computing power. He's interested in teaching AI models how to efficiently forget. Hochreiter holds a special place in the world of artificial intelligence, having scaled the technology's highest peaks long before most computer scientists. As a university student in Munich during the 1990s, he came up with the conceptual framework that underpinned the first generation of nimble AI models used by Alphabet, Apple and Amazon.

We present FORGE, a method that enables sim-to-real transfer of contact-rich manipulation policies in the presence of significant pose uncertainty. FORGE combines a force threshold mechanism with a dynamics randomization scheme during policy learning in simulation, to enable the robust transfer of the learned policies to the real robot. At deployment, FORGE policies, conditioned on a maximum allowable force, adaptively perform contact-rich tasks while respecting the specified force threshold, regardless of the controller gains. Additionally, FORGE autonomously predicts a termination action once the task has succeeded. We demonstrate that FORGE can be used to learn a variety of robust contact-rich policies, enabling multi-stage assembly of a planetary gear system, which requires success across three assembly tasks: nut-threading, insertion, and gear meshing. Project website can be accessed at https://noseworm.github.io/forge/.

AI-Generated Content (AIGC) is gaining great popularity, with many emerging commercial services and applications. These services leverage advanced generative models, such as latent diffusion models and large language models, to generate creative content (e.g., realistic images and fluent sentences) for users. The usage of such generated content needs to be highly regulated, as the service providers need to ensure the users do not violate the usage policies (e.g., abuse for commercialization, generating and distributing unsafe content). A promising solution to achieve this goal is watermarking, which adds unique and imperceptible watermarks on the content for service verification and attribution. Numerous watermarking approaches have been proposed recently. However, in this paper, we show that an adversary can easily break these watermarking mechanisms. Specifically, we consider two possible attacks. (1) Watermark removal: the adversary can easily erase the embedded watermark from the generated content and then use it freely bypassing the regulation of the service provider. (2) Watermark forging: the adversary can create illegal content with forged watermarks from another user, causing the service provider to make wrong attributions. We propose Warfare, a unified methodology to achieve both attacks in a holistic way. The key idea is to leverage a pre-trained diffusion model for content processing and a generative adversarial network for watermark removal or forging. We evaluate Warfare on different datasets and embedding setups. The results prove that it can achieve high success rates while maintaining the quality of the generated content. Compared to existing diffusion model-based attacks, Warfare is 5,050~11,000x faster.

Membership inference (MI) attack is currently the most popular test for measuring privacy leakage in machine learning models. Given a machine learning model, a data point and some auxiliary information, the goal of an MI attack is to determine whether the data point was used to train the model. In this work, we study the reliability of membership inference attacks in practice. Specifically, we show that a model owner can plausibly refute the result of a membership inference test on a data point $x$ by constructing a proof of repudiation that proves that the model was trained without $x$. We design efficient algorithms to construct proofs of repudiation for all data points of the training dataset. Our empirical evaluation demonstrates the practical feasibility of our algorithm by constructing proofs of repudiation for popular machine learning models on MNIST and CIFAR-10. Consequently, our results call for a re-evaluation of the implications of membership inference attacks in practice.

Manufacturers are counting on technologies and systems known collectively as Industry 4.0 to drive efficiencies and unleash new revenue streams in their sectors. Industry 4.0 leverages sensor technologies, advanced analytics, artificial intelligence, and machine learning capabilities to collect and analyze data. Benefits include reducing production downtime, improving product quality, and enabling proactive maintenance. But important factors, including legacy operations and failure to effectively engage frontline workers in operational redesign, often stand in the way of such large-scale transformation. Momentum for Industry 4.0 is growing -- one report by Statista projects the market will swell from $263 billion in 2021 to $1.1 trillion by 2028.

November may also bring a long-awaited fix demanded by the community: new ways to earn experience points. As of now, the game relies on completing challenges that many (including myself) have criticized as too restrictive for a video game meant to be played at a relaxed pace of our choosing. These challenges force players into modes and play styles that they may not necessarily want to play. Going forward, players will experience a new "Match XP beta," along with a new free battle pass. The studio did not expand on how this match experience system works, nor did they detail what this non-season battle pass might look like, but promised updates -- regularly, and soon.

In an effort to create a standard set of tools that would help data science teams collaborate on AI development, an infrastructure initiative launched this week will promote a unified stack for developing and scaling machine learning models. The AI Infrastructure Alliance said this week it will initially focus on creating Canonical Stack for AI envisioned as a development platform for machine learning models destined for enterprise applications. As with previous hardware and software stacks, the machine learning initiative seeks to forge an AI development infrastructure that would free developers to address more complex problems. As machine learning models move to the edge, the alliance said it would create a single platform that integrates existing AI technologies into a common framework that would accelerate and improve MLOps and edge applications. Establishing a so-called canonical AI stack for machine learning and MLOps would include developing best practices and architectures used to scale machine learning models in edge and other applications.

The integration and adoption of technology into every industry is inevitable, even in industries that are traditionally relatively slower to adopt like real estate. Tech developments are enabling professionals in the space to be able to facilitate easier, faster, efficient, and more secure deals for all parties involved. Below are some of the emerging technology trends that are being integrated into real estate now, and in the coming years, many of which have been accelerated by the outbreak of COVID-19. The AR/VR spaces have seen tremendous growth, and it seems with the onset of COVID-19 the demand and urgency to roll out these capabilities has only increased. With the need for limited contact during the pandemic, restricted travel, and work from home trends, it is only natural that AR/VR trends will continue to surface in the real estate industry.

Streamlining an inference pipeline on the latest fine-tuned NLP model is a must for fast prototyping. However, with the plethora of diverse model architectures and NLP libraries to choose from, it can make prototyping a time-consuming task. As such, we've created The NLP Model Forge. A database/code generator for 1,400 fine-tuned models that were carefully curated from top NLP research companies such as Hugging Face, Facebook (ParlAI), DeepPavlov, and AI2. The Forge is your destination for generating inference code for your NLP model of choice.

"QC Ware estimates that with Forge Data Loaders, the industry's 10-to-15-year timeline for practical applications of QML will be reduced significantly," said Yianni Gamvros, Head of Product and Business Development at QC Ware. "What our algorithms team has achieved for the quantum computing industry is equivalent to a quantum hardware manufacturer introducing a chip that is 10 to 100 times faster than their previous offering. This exciting development will require business analysts to update their quad charts and innovation scouts to adjust their technology timelines." Apart from the Forge Data Loaders, the latest release of Forge includes tools for GPU acceleration, which allows algorithms testing to be completed in seconds versus hours, and turnkey algorithms implementations on a choice of simulators and quantum hardware. Quantum hardware integrations include D-Wave Systems, and IonQ and Rigetti architectures through Amazon Braket.