Anthropic to commit 5% stake to new'public benefit fund' Anthropic cited the potential impact of artificial intelligence on workers as one area of focus for its new fund. Artificial intelligence lab Anthropic, the developer of Claude, said on Sunday it would commit 5% of its shares to a new public benefit fund so it has increased flexibility and scale to make donations for the "public good." The proposed fund would be established as a wholly owned limited liability company and would receive an additional 2% of Anthropic's shares, at the board's discretion, for every 50% increase in the company's value above a baseline that has yet to be disclosed. The move comes at a time when Anthropic is planning one of the biggest initial public offerings in history as a public benefit corporation, a legal structure that requires management to consider certain broader interests as well as turning a profit for investors. The company, which has positioned itself as a safety-first AI lab, has said AI could pose "catastrophic or existential risks to humanity" and devoted roughly 80 pages of the 261-page main body of its prospectus to laying out risk factors.

This week I've been at SXSW London . There's been music, film, and a lot--and I mean --of talk about AI. I also had the opportunity to sit down with Gloria Mark, a psychologist at the University of California, Irvine, who has spent the last 30 years studying how people interact with digital technologies. Early in her career, the biggest concerns were the potential impacts of internet and email use on our brains. We may laugh those concerns off today, but it's true that as the technologies became more ubiquitous and ingrained in our daily lives, our attention spans began to shrink.

Early large-scale audio datasets, such as LibriSpeech, were built with hundreds of individual contributors whose voices were instrumental in the development of speech technologies, including audiobooks and voice assistants. Y et, a decade later, these same contributions have exposed voice actors to a range of risks. While existing ethical frameworks emphasize Consent, Credit, and Compensation (C), they do not adequately address the emergent risks involving vocal identities that are increasingly decoupled from context, authorship, and control. Drawing on qualitative interviews with 20 professional voice actors, this paper reveals how synthetic replication of voice without clear provenance or enforceable constraints exposes individuals to both reputational and security threats. Beyond reputational harm, such as re-purposing voice data in erotic content, offensive political messaging, and meme culture, we document concerns about accountability breakdowns when their voice is leveraged to clone voices that are deployed in high-stakes scenarios such as financial fraud, misinformation campaigns, or impersonation scams. In such cases, actors face social and legal fallout without recourse, while very few of them have a legal representative or union protection. To make sense of these shifting dynamics, we introduce the PRAC framework - an expansion of C that foregrounds Privacy, Reputation, Accountability, Consent, Credit, and Compensation as interdependent pillars of data used in the synthetic voice economy. This framework captures how privacy risks are amplified through non-consensual training, how reputational harm arises from decontextualized deployment, and how accountability can be reimagined AI Data ecosystems. We argue that voice, as both a biometric identifier and creative labor, demands governance models that restore creator agency, ensure traceability, and establish enforceable boundaries for ethical reuse.

There was a reasonable amount of discussion about this paper. The author feedback clarified a variety of issues which caused some reviewers to increase their scores, while some of the discussion caused other reviewers to decrease their scores. Although there was one holdout, the majority of the reviewers leaned towards rejection of the paper. However, I believe this is one of the rare cases where the AC should recommend for the PC to accept the paper against the recommendations of the reviewers. The main reason I'm recommending acceptance is due to the broader context and potential impact of the paper.

The manuscript proposes a formalism for computing stochastic estimates of gradients for loss functions. The formalism, referred to as stochastic computation graphs, is very general, applying to models with deterministic and stochastic components, and allowing the computation of gradient estimates for a broad range of models. Methods for deep networks, variational inference, and reinforcement learning are identified as special cases of the proposed framework. The proposed stochastic computation graphs are essentially Bayesian networks which may also contain deterministic nodes, and which are interpreted as encoding distributions over loss functions that are to be minimized. The gradient estimation algorithm extends backpropagation to the partially stochastic case covered by these models, by simply composing score function estimators and pathwise derivative estimators.

This work got mixed reviews: R1 praised the potential impact of such a simple idea being shown to work remarkably well, but other reviewers had significant concerns about the empirical evaluation, which is especially important when the main contribution of the paper is to show that an idea is effective in practice. The reviewers were ultimately unable to reach a consensus about this paper, but all reviewers agreed that the core idea is promising, and R2, R3 and R4 raised their scores in light of the discussion and the author feedback. While the resulting scores still make this a difficult decision overall, I have chosen to recommend acceptance. The main point of discussion was whether the required changes to the manuscript require another review cycle or not. Indeed, the requested changes were quite broad: - demonstrate the effect of the initial variance of the linear layers - compare the model against modern autoencoder variants - compare against vanilla autoencoders with varying latent dimension - demonstrate the effect of the number of linear layers - avoid overclaiming, e.g. about the proposed model working well "with all types of optimizers" - etc.

The theoretical contributions (relationship between time-dependent exogenous intensity to average activity) appear significant, and the use of this result to derive a closed-form control algorithm appears to be a contribution to the field of shaping activities in social (and other) networks modeled as multivariate Hawkes processes. Regarding the clarity of this paper: The paper does not frame itself relative to prior work clearly. Clearly indicating where and how this work is a generalization of prior work [8] would improve clarity of the paper, and highlight the core contribution of handling *time-dependent* exogenous events. Stating that the paper "establishes theoretical foundations of optimal campaigning over social networks where user activities are modeled as multivariate Hawkes processes" does not properly localize the work relative to [8]. This could be improved in part by mentioning [8] in the introduction, and specifically stating that prior work has done optimal control with constant exogenous control, and that this paper addresses multi-stage exogenous intensity.

The new robot is designed to revolutionize the way we work and interact with machines. Chinese startup Pudu Robotics has unveiled its latest creation, the D9 humanoid robot, designed to revolutionize the way we work and interact with machines. Standing at an impressive 5.57 feet tall, this bipedal machine is not just another robot -- it's a versatile assistant ready to tackle a wide range of tasks in various settings. The D9 is no ordinary robot. With its ability to walk upright and carry loads up to 44 pounds, it's built to handle real-world challenges.