AI is undergoing a paradigm shift with the rise of models (e.g., BERT, DALL-E, GPT-3) that are trained on broad data at scale and are adaptable to a wide range of downstream tasks. We call these models foundation models to underscore their critically central yet incomplete character. This report provides a thorough account of the opportunities and risks of foundation models, ranging from their capabilities (e.g., language, vision, robotics, reasoning, human interaction) and technical principles (e.g., model architectures, training procedures, data, systems, security, evaluation, theory) to their applications (e.g., law, healthcare, education) and societal impact (e.g., inequity, misuse, economic and environmental impact, legal and ethical considerations). Though foundation models are based on standard deep learning and transfer learning, their scale results in new emergent capabilities, and their effectiveness across so many tasks incentivizes homogenization. Homogenization provides powerful leverage but demands caution, as the defects of the foundation model are inherited by all the adapted models downstream. Despite the impending widespread deployment of foundation models, we currently lack a clear understanding of how they work, when they fail, and what they are even capable of due to their emergent properties. To tackle these questions, we believe much of the critical research on foundation models will require deep interdisciplinary collaboration commensurate with their fundamentally sociotechnical nature.

Here we describe Protein Interaction Quantitation (PIQ), a computational method that models the magnitude and shape of genome-wide DNase profiles to facilitate the identification of transcription factor (TF) binding sites. Through the use of machine learning techniques, PIQ identified binding sites for >700 TFs from one DNase-seq experiment with accuracy comparable to ChIP-seq for motif-associated TFs (median AUC=0.93 across 303 TFs). We applied PIQ to analyze DNase-seq data from mouse embryonic stem cells differentiating into pre-pancreatic and intestinal endoderm. We identified (n=120) and experimentally validated eight ‘pioneer’ TF families that dynamically open chromatin, enabling other TFs to bind to adjacent DNA. Four pioneer TF families only open chromatin in one direction from their motifs. Furthermore, we identified a class of ‘settler’ TFs whose genomic binding is principally governed by proximity to open chromatin. Our results support a model of hierarchical TF binding in which directional and non-directional pioneer activity shapes the chromatin landscape for population by settler TFs.

Overparameterized neural networks can be highly accurate on average on an i.i.d.test set yet consistently fail on atypical groups of the data (e.g., by learning spurious correlations that hold on average but not in such groups). Distributionally robust optimization (DRO) allows us to learn models that instead minimize the worst-case training loss over a set of pre-defined groups. However, we find that naively applying group DRO to overparameterized neural networks fails: these models can perfectly fit the training data, and any model with vanishing average training loss also already has vanishing worst-case training loss. Instead, their poor worst-case performance arises from poor generalization on some groups. By coupling group DRO models with increased regularization-stronger-than-typical 2 regularization or early stopping-we achieve substantially higher worst-group accuracies, with 10-40 percentage point improvements on a natural language inference task and two image tasks, while maintaining high average accuracies. Our results suggest that regularization is critical for worst-group generalization in the overparameterized regime, even if it is not needed for average generalization. Finally, we introduce and give convergence guarantees for a stochastic optimizer for the group DRO setting, underpinning the empirical study above.

We propose a new generative language model for sentences that first samples a prototype sentence from the training corpus and then edits it into a new sentence. Compared to traditional language models that generate from scratch either left-to-right or by first sampling a latent sentence vector, our prototype-thenedit model improves perplexity on language modeling and generates higher quality outputs according to human evaluation. Furthermore, the model gives rise to a latent edit vector that captures interpretable semantics such as sentence similarity and sentence-level analogies.

Language models are generally trained on data spanning a wide range of topics (e.g., news, reviews, fiction), but they might be applied to an a priori unknown target distribution (e.g., restaurant reviews). In this paper, we first show that training on text outside the test distribution can degrade test performance when using standard maximum likelihood (MLE) training. To remedy this without the knowledge of the test distribution, we propose an approach which trains a model that performs well over a wide range of potential test distributions. In particular, we derive a new distributionally robust optimization (DRO) procedure which minimizes the loss of the model over the worst-case mixture of topics with sufficient overlap with the training distribution. Our approach, called topic conditional value at risk (topic CVaR), obtains a 5.5 point perplexity reduction over MLE when the language models are trained on a mixture of Yelp reviews and news and tested only on reviews.