Learning robust audio representations currently demands extensive datasets of real-world sound recordings. By applying artificial transformations to these recordings, models can learn to recognize similarities despite subtle variations through techniques like contrastive learning. However, these transformations are only approximations of the true diversity found in real-world sounds, which are generated by complex interactions of physical processes, from vocal cord vibrations to the resonance of musical instruments. We propose a solution to both the data scale and transformation limitations, leveraging synthetic audio. By randomly perturbing the parameters of a sound synthesizer, we generate audio doppelgängers—synthetic positive pairs with causally manipulated variations in timbre, pitch, and temporal envelopes. These variations, difficult to achieve through augmentations of existing audio, provide a rich source of contrastive information. Despite the shift to randomly generated synthetic data, our method produces strong representations, outperforming real data on several standard audio classification tasks. Notably, our approach is lightweight, requires no data storage, and has only a single hyperparameter, which we extensively analyze. We offer this method as a complement to existing strategies for contrastive learning in audio, using synthesized sounds to reduce the data burden on practitioners.

Progress in AI is driven largely by the scale and quality of training data. Despite this, there is a deficit of empirical analysis examining the attributes of well-established datasets beyond text. In this work we conduct the largest and first-of-its-kind longitudinal audit across modalities --- popular text, speech, and video datasets --- from their detailed sourcing trends and use restrictions to their geographical and linguistic representation. Our manual analysis covers nearly 4000 public datasets between 1990-2024, spanning 608 languages, 798 sources, 659 organizations, and 67 countries. We find that multimodal machine learning applications have overwhelmingly turned to web-crawled, synthetic, and social media platforms, such as YouTube, for their training sets, eclipsing all other sources since 2019. Secondly, tracing the chain of dataset derivations we find that while less than 33% of datasets are restrictively licensed, over 80% of the source content in widely-used text, speech, and video datasets, carry non-commercial restrictions. Finally, counter to the rising number of languages and geographies represented in public AI training datasets, our audit demonstrates measures of relative geographical and multilingual representation have failed to significantly improve their coverage since 2013. We believe the breadth of our audit enables us to empirically examine trends in data sourcing, restrictions, and Western-centricity at an ecosystem-level, and that visibility into these questions are essential to progress in responsible AI. As a contribution to ongoing improvements in dataset transparency and responsible use, we release our entire multimodal audit, allowing practitioners to trace data provenance across text, speech, and video.

Augmentative and alternative communication devices (AACs) are designed to assist humans with complex communication needs. Recently, AAC use has been reported in non-human animals. Such tools may potentially provide enrichment and increase interspecies connection. However, there is no evaluation framework and little data available to assess AAC potential. Here, we examine seven months of a single parrot’s sustained use of a tablet-based AAC totaling 129 sessions within 190 days. After devising a coding schema, we propose a framework to explore the expressive potential and enrichment value for the parrot. Our results suggest that the choice of destination words cannot be simply explained based on random selection or icon location alone, and 92% of corroborable selections are validated by behaviors. The parrot interactions also appear significantly skewed toward social and cognitive enrichment. This work is a first step toward assessment of AAC use for parrot enrichment.

Audiovisual representation learning typically relies on the correspondence between sight and sound. However, there are often multiple audio tracks that can correspond with a visual scene. Consider, for example, different con-versations on the same crowded street. The effect of such counterfactual pairs on audiovisual representation learning has not been previously explored. To investigate this, we use dubbed versions of movies and television shows to augment cross-modal contrastive learning. Our approach learns to represent alternate audio tracks, differing only in speech, similarly to the same video. Our results, from a comprehensive set of experiments investigating different training strategies, show this general approach improves performance on a range of downstream auditory and audio-visual tasks, without majorly affecting linguistic task per-formance overall. These findings highlight the importance of considering speech variation when learning scene-level audiovisual correspondences and suggest that dubbed au-dio can be a useful augmentation technique for training au-diovisual models toward more robust performance on di-verse downstream tasks.

Neural audio synthesis methods now allow specifying ideas in natural language. However, these methods produce results that cannot be easily tweaked, as they are based on large latent spaces and up to billions of uninterpretable parameters. We propose a text-to-audio generation method that leverages a virtual modular sound synthesizer with only 78 parameters. Synthesizers have long been used by skilled sound designers for media like music and film due to their flexibility and intuitive controls. Our method, CTAG, iteratively updates a synthesizer's parameters to produce high-quality audio renderings of text prompts that can be easily inspected and tweaked. Sounds produced this way are also more abstract, capturing essential conceptual features over fine-grained acoustic details, akin to how simple sketches can vividly convey visual concepts. Our results show how CTAG produces sounds that are distinctive, perceived as artistic, and yet similarly identifiable to recent neural audio synthesis models, positioning it as a valuable and complementary tool.

High-quality alt text is crucial for making scientific figures accessible to blind and low-vision readers. Crafting complete, accurate alt text is challenging even for domain experts, as published figures often depict complex visual information and readers have varied informational needs. These challenges, along with high diversity in figure types and domain-specific details, also limit the usefulness of fully automated approaches. Consequently, the prevalence of high-quality alt text is very low in scientific papers today. We investigate whether and how human-AI collaborative editing systems can help address the difficulty of writing high-quality alt text for complex scientific figures. We present FigurA11y, an interactive system that generates draft alt text and provides suggestions for author revisions using a pipeline driven by extracted figure and paper metadata. We test two versions, motivated by prior work on visual accessibility and writing support. The base Draft+Revise version provides authors with an automatically generated draft description to revise, along with extracted figure metadata and figure-specific alt text guidelines to support the revision process. The full Interactive Assistance version further adds contextualized suggestions: text snippets to iteratively produce descriptions, and hypothetical user questions with possible answers to reveal potential ambiguities and resolutions. In a study of authors (N=14), we found the system assisted them in efficiently producing descriptive alt text. Generated drafts and interface elements enabled authors to quickly initiate and edit detailed descriptions. Additionally, interactive suggestions from the full system prompted more iteration and highlighted aspects for authors to consider, resulting in greater deviation from the drafts without increased average cognitive load or manual effort.

General-purpose artificial intelligence (AI) systems are built on massive swathes of public web data, assembled into corpora such as C4, RefinedWeb, and Dolma. To our knowledge, we conduct the first, large-scale, longitudinal audit of the consent protocols for the web domains underlying AI training corpora. Our audit of 14,000 web domains provides an expansive view of crawlable web data and how codified data use preferences are changing over time. We observe a proliferation of AI-specific clauses to limit use, acute differences in restrictions on AI developers, as well as general inconsistencies between websites' expressed intentions in their Terms of Service and their robots.txt. We diagnose these as symptoms of ineffective web protocols, not designed to cope with the widespread re-purposing of the internet for AI. Our longitudinal analyses show that in a single year (2023-2024) there has been a rapid crescendo of data restrictions from web sources, rendering ~5\%+ of all tokens in C4, or 28%+ of the most actively maintained, critical sources in C4, fully restricted from use. For Terms of Service crawling restrictions, a full 45% of C4 is now restricted. If respected or enforced, these restrictions are rapidly biasing the diversity, freshness, and scaling laws for general-purpose AI systems. We hope to illustrate the emerging crises in data consent, for both developers and creators. The foreclosure of much of the open web will impact not only commercial AI, but also non-commercial AI and academic research.

We present a new dataset for learning to solve, explain, and generate university-level STEM questions from 27 courses across a dozen departments in seven universities. We scale up previous approaches to questions from courses in the departments of Mechanical Engineering, Materials Science and Engineering, Chemistry, Electrical Engineering, Computer Science, Physics, Earth Atmospheric and Planetary Sciences, Economics, Mathematics, Biological Engineering, Data Systems, and Society, and Statistics. We visualize similarities and differences between questions across courses. We demonstrate that a large foundation model is able to generate questions that are as appropriate and at the same difficulty level as human-written questions.

While developing a story, novices and published writers alike have had to look outside themselves for inspiration. Language models have recently been able to generate text fluently, producing new stochastic narratives upon request. However, effectively integrating such capabilities with human cognitive faculties and creative processes remains challenging. We propose to investigate this integration with a multimodal writing support interface that offers writing suggestions textually, visually, and aurally. We conduct an extensive study that combines elicitation of prior expectations before writing, observation and semi-structured interviews during writing, and outcome evaluations after writing. Our results illustrate the individual and situational variation in machine-in-the-loop writing approaches, suggestion acceptance, and ways the system is helpful. Centrally, we report how participants perform integrative leaps , by which they do cognitive work to integrate suggestions of varying semantic relevance into their developing stories. We interpret these findings, offering modeling and design recommendations for future creative writing support technologies.

Can a machine learn Machine Learning? This work trains a machine learning model to solve machine learning problems from a University undergraduate level course. We generate a new training set of questions and answers consisting of course exercises, homework, and quiz questions from MIT’s 6.036 Introduction to Machine Learning course and train a machine learning model to answer these questions. Our system demonstrates an overall accuracy of 96% for open-response questions and 97% for multiple-choice questions, compared with MIT students’ average of 93%, achieving grade A performance in the course, all in real-time. Questions cover all 12 topics taught in the course, excluding coding questions or questions with images. Topics include: (i) basic machine learning principles; (ii) perceptrons; (iii) feature extraction and selection; (iv) logistic regression; (v) regression; (vi) neural networks; (vii) advanced neural networks; (viii) convolutional neural networks; (ix) recurrent neural networks; (x) state machines and MDPs; (xi) reinforcement learning; and (xii) decision trees. Our system uses Transformer models within an encoder-decoder architecture with graph and tree representations. An important aspect of our approach is a data-augmentation scheme for generating new example problems. We also train a machine learning model to generate problem hints. Thus, our system automatically generates new questions across topics, answers both open-response questions and multiple-choice questions, classifies problems, and generates problem hints, pushing the envelope of AI for STEM education.

Measuring the acoustic characteristics of a space is often done by capturing its impulse response (IR), a representation of how a full-range stimulus sound excites it. This work generates an IR from a single image, which can then be applied to other signals using convolution, simulating the reverberant characteristics of the space shown in the image. Recording these IRs is both time-intensive and expensive, and often infeasible for inaccessible locations. We use an end-to-end neural network architecture to generate plausible audio impulse responses from single images of acoustic environments. We evaluate our method both by comparisons to ground truth data and by human expert evaluation. We demonstrate our approach by generating plausible impulse responses from diverse settings and formats including well known places, musical halls, rooms in paintings, images from animations and computer games, synthetic environments generated from text, panoramic images, and video conference backgrounds.

Software tools for media production have largely been adapted from physical media paradigms, offering blank canvases upon which to import, combine, and process content. In music production, this increasingly involves meticulous manual assembly of audio clips often carefully curated from diverse sources. As collections of audio content scale upwards in sample size, diversity, and number, creative projects require exponentially more time, effort, and attention to effectively shape them. New tools must find new ways to contend with this abundance of content. We propose the Sound Sketchpad, an algorithm-in-the-loop audio-graphical system and interface for combining sounds from a database into new music. It allows a user to sketch broad musical ideas by making sound, and then interactively modify and refine the resulting composition by drawing visual paths. We discuss the design, implementation, and advantages of this approach.