Creative works shape our identity, values, and worldview. And there are fundamental elements of our culture that are uniquely human. Only humans are capable of communicating the endless intricacies, nuances, and complications of the human condition through art - whether it be music, performance, writing, or any other form of creativity. Developments in artificial intelligence are exciting and could advance the world farther than we ever thought possible. But AI can never replace human expression and artistry.

We are the Concept Art Association, an advocacy organization for artists working in entertainment. Our board member, Karla Ortiz, has been one of the leaders in our industry fighting back against the unethical practices happening in the AI text-to-image space. As an organization and as individuals we deeply care about this issue, not just for those actively working as visual artists, but for future generations of artists and for the preservation of our creative industries. A text-to-image model takes input from a user in the form of a natural language prompt and produces an image matching that prompt. To condition that capability the model needs to be trained on a huge collection of images, media, and text descriptions scraped from the web and collected in the form of a "dataset " in order to extract and encode an intricate statistical survey of the dataset's items.

This post is a commentary on the MLCommons article "Perspective: Unlocking ML requires an ecosystem approach" by Peter Mattson, Aarush Selvan, David Kanter, Vijay Janapa Reddi, Roger Roberts, and Jacomo Corbo. The world of artificial intelligence (AI) and machine learning (ML) is undergoing a sea change from science to engineering at scale. Over the past decade, the volume of AI research has skyrocketed as the cost to train and deploy commercial models has decreased. Between 2015 and 2021, the cost to train an image classification system fell by 64 percent, while training times improved by 94 percent in the same period.1 The emergence of foundation models--large-scale, deep learning models trained on massive, broad, unstructured data sets--has enabled entrepreneurs and business executives to see the possibility of true scale.

Currently, the dominant paradigm in AI safety is alignment with human values. Here we describe progress on developing an alternative approach to safety, based on ethical rationalism (Gewirth, 1978), and propose an inherently safe implementation path via hybrid theorem provers in a sandbox. As AGIs evolve, their alignment may fade, but their rationality can only increase (otherwise more rational ones will have a significant evolutionary advantage) so an approach that ties their ethics to their rationality has clear long-term advantages.

A computer vision system using low-resolution image sensors can provide intelligent services (e.g., activity recognition) but preserve unnecessary visual privacy information from the hardware level. However, preserving visual privacy and enabling accurate machine recognition have adversarial needs on image resolution. Modeling the trade-off of privacy preservation and machine recognition performance can guide future privacy-preserving computer vision systems using low-resolution image sensors. In this paper, using the at-home activity of daily livings (ADLs) as the scenario, we first obtained the most important visual privacy features through a user survey. Then we quantified and analyzed the effects of image resolution on human and machine recognition performance in activity recognition and privacy awareness tasks. We also investigated how modern image super-resolution techniques influence these effects. Based on the results, we proposed a method for modeling the trade-off of privacy preservation and activity recognition on low-resolution images.

Federated learning (FL), an effective distributed machine learning framework, implements model training and meanwhile protects local data privacy. It has been applied to a broad variety of practice areas due to its great performance and appreciable profits. Who owns the model, and how to protect the copyright has become a real problem. Intuitively, the existing property rights protection methods in centralized scenarios (e.g., watermark embedding and model fingerprints) are possible solutions for FL. But they are still challenged by the distributed nature of FL in aspects of the no data sharing, parameter aggregation, and federated training settings. For the first time, we formalize the problem of copyright protection for FL, and propose FedRight to protect model copyright based on model fingerprints, i.e., extracting model features by generating adversarial examples as model fingerprints. FedRight outperforms previous works in four key aspects: (i) Validity: it extracts model features to generate transferable fingerprints to train a detector to verify the copyright of the model. (ii) Fidelity: it is with imperceptible impact on the federated training, thus promising good main task performance. (iii) Robustness: it is empirically robust against malicious attacks on copyright protection, i.e., fine-tuning, model pruning, and adaptive attacks. (iv) Black-box: it is valid in the black-box forensic scenario where only application programming interface calls to the model are available. Extensive evaluations across 3 datasets and 9 model structures demonstrate FedRight's superior fidelity, validity, and robustness.

Data-driven algorithms are only as good as the data they work with, while data sets, especially social data, often fail to represent minorities adequately. Representation Bias in data can happen due to various reasons ranging from historical discrimination to selection and sampling biases in the data acquisition and preparation methods. Given that "bias in, bias out", one cannot expect AI-based solutions to have equitable outcomes for societal applications, without addressing issues such as representation bias. While there has been extensive study of fairness in machine learning models, including several review papers, bias in the data has been less studied. This paper reviews the literature on identifying and resolving representation bias as a feature of a data set, independent of how consumed later. The scope of this survey is bounded to structured (tabular) and unstructured (e.g., image, text, graph) data. It presents taxonomies to categorize the studied techniques based on multiple design dimensions and provides a side-by-side comparison of their properties. There is still a long way to fully address representation bias issues in data. The authors hope that this survey motivates researchers to approach these challenges in the future by observing existing work within their respective domains.

Fox News contributor Jonathan Turley sounds off on the latest developments on'The Story.' Hunter Biden's countersuit against the computer shopkeeper who turned over a laptop belonging to the first son to authorities and members of the press is "one of the weirdest filings I have read in some time," one legal scholar says. In his lawsuit Biden claims John Paul Mac Isaac illegally distributed his personal data and alleges he invaded his privacy. Isaac has said a man he believed to be Hunter Biden did not return after the 90-day policy window to retrieve his damaged laptop. Biden's suit reportedly claims Isaac violated a Delaware law prohibiting dissemination or exploitation of property within one year's time. George Washington University Law professor and Fox News contributor Jonathan Turley said Friday the suit is "like Alice in Wonderland." "It gets curiouser and curiouser because this is one of the weirdest filings I have read in some time because [Hunter is] telling the court this may or may not be my laptop and these files may or may not be my files, but I am egregiously injured by the invasion of my privacy," Turley explained on "The Story with Martha MacCallum."

Is imitation the sincerest form of flattery, or theft? Perhaps it comes down to the imitator. Text-to-image artificial intelligence systems such as DALL-E 2, Midjourney and Stable Diffusion are trained on huge amounts of image data from the web. As a result, they often generate outputs that resemble real artists' work and style. It's safe to say artists aren't impressed. To further complicate things, although intellectual property law guards against the misappropriation of individual works of art, this doesn't extend to emulating a person's style.

When we sit and work on legislation, I wonder whether the legislation alone would deliver the results we're hoping it will deliver, particularly when we look at competitors, such as China and the US,