Detecting hallucinations in large language models using semantic entropy.

Large language model (LLM) systems, such as ChatGPT1 or Gemini2, can show impressive reasoning and question-answering capabilities but often 'hallucinate' false outputs and unsubstantiated answers3,4. Answering unreliably or without the necessary information prevents adoption in diverse fields, with problems including fabrication of legal precedents5 or untrue facts in news articles6 and even posing a risk to human life in medical domains such as radiology7. Encouraging truthfulness through supervision or reinforcement has been only partially successful8. Researchers need a general method for detecting hallucinations in LLMs that works even with new and unseen questions to which humans might not know the answer. Here we develop new methods grounded in statistics, proposing entropy-based uncertainty estimators for LLMs to detect a subset of hallucinations-confabulations-which are arbitrary and incorrect generations. Our method addresses the fact that one idea can be expressed in many ways by computing uncertainty at the level of meaning rather than specific sequences of words. Our method works across datasets and tasks without a priori knowledge of the task, requires no task-specific data and robustly generalizes to new tasks not seen before. By detecting when a prompt is likely to produce a confabulation, our method helps users understand when they must take extra care with LLMs and opens up new possibilities for using LLMs that are otherwise prevented by their unreliability.

AI models collapse when trained on recursively generated data.

Stable diffusion revolutionized image creation from descriptive text. GPT-2 (ref. 1), GPT-3(.5) (ref. 2) and GPT-4 (ref. 3) demonstrated high performance across a variety of language tasks. ChatGPT introduced such language models to the public. It is now clear that generative artificial intelligence (AI) such as large language models (LLMs) is here to stay and will substantially change the ecosystem of online text and images. Here we consider what may happen to GPT-{n} once LLMs contribute much of the text found online. We find that indiscriminate use of model-generated content in training causes irreversible defects in the resulting models, in which tails of the original content distribution disappear. We refer to this effect as 'model collapse' and show that it can occur in LLMs as well as in variational autoencoders (VAEs) and Gaussian mixture models (GMMs). We build theoretical intuition behind the phenomenon and portray its ubiquity among all learned generative models. We demonstrate that it must be taken seriously if we are to sustain the benefits of training from large-scale data scraped from the web. Indeed, the value of data collected about genuine human interactions with systems will be increasingly valuable in the presence of LLM-generated content in data crawled from the Internet.

Learning from prepandemic data to forecast viral escape.

Effective pandemic preparedness relies on anticipating viral mutations that are able to evade host immune responses to facilitate vaccine and therapeutic design. However, current strategies for viral evolution prediction are not available early in a pandemic-experimental approaches require host polyclonal antibodies to test against1-16, and existing computational methods draw heavily from current strain prevalence to make reliable predictions of variants of concern17-19. To address this, we developed EVEscape, a generalizable modular framework that combines fitness predictions from a deep learning model of historical sequences with biophysical and structural information. EVEscape quantifies the viral escape potential of mutations at scale and has the advantage of being applicable before surveillance sequencing, experimental scans or three-dimensional structures of antibody complexes are available. We demonstrate that EVEscape, trained on sequences available before 2020, is as accurate as high-throughput experimental scans at anticipating pandemic variation for SARS-CoV-2 and is generalizable to other viruses including influenza, HIV and understudied viruses with pandemic potential such as Lassa and Nipah. We provide continually revised escape scores for all current strains of SARS-CoV-2 and predict probable further mutations to forecast emerging strains as a tool for continuing vaccine development ( evescape.org ).

Disease variant prediction with deep generative models of evolutionary data.

Quantifying the pathogenicity of protein variants in human disease-related genes would have a marked effect on clinical decisions, yet the overwhelming majority (over 98%) of these variants still have unknown consequences1-3. In principle, computational methods could support the large-scale interpretation of genetic variants. However, state-of-the-art methods4-10 have relied on training machine learning models on known disease labels. As these labels are sparse, biased and of variable quality, the resulting models have been considered insufficiently reliable11. Here we propose an approach that leverages deep generative models to predict variant pathogenicity without relying on labels. By modelling the distribution of sequence variation across organisms, we implicitly capture constraints on the protein sequences that maintain fitness. Our model EVE (evolutionary model of variant effect) not only outperforms computational approaches that rely on labelled data but also performs on par with, if not better than, predictions from high-throughput experiments, which are increasingly used as evidence for variant classification12-16. We predict the pathogenicity of more than 36 million variants across 3,219 disease genes and provide evidence for the classification of more than 256,000 variants of unknown significance. Our work suggests that models of evolutionary information can provide valuable independent evidence for variant interpretation that will be widely useful in research and clinical settings.

Mask wearing in community settings reduces SARS-CoV-2 transmission.

The effectiveness of mask wearing at controlling severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) transmission has been unclear. While masks are known to substantially reduce disease transmission in healthcare settings [D. K. Chu et al., Lancet 395, 1973­1987 (2020); J. Howard et al., Proc. Natl. Acad. Sci. U.S.A. 118, e2014564118 (2021); Y. Cheng et al., Science eabg6296 (2021)], studies in community settings report inconsistent results [H. M. Ollila et al., medRxiv (2020); J. Brainard et al., Eurosurveillance 25, 2000725 (2020); T. Jefferson et al., Cochrane Database Syst. Rev. 11, CD006207 (2020)]. Most such studies focus on how masks impact transmission, by analyzing how effective government mask mandates are. However, we find that widespread voluntary mask wearing, and other data limitations, make mandate effectiveness a poor proxy for mask-wearing effectiveness. We directly analyze the effect of mask wearing on SARS-CoV-2 transmission, drawing on several datasets covering 92 regions on six continents, including the largest survey of wearing behavior (n= 20 million) [F. Kreuter et al., https://gisumd.github.io/COVID-19-API-Documentation (2020)]. Using a Bayesian hierarchical model, we estimate the effect of mask wearing on transmission, by linking reported wearing levels to reported cases in each region, while adjusting for mobility and nonpharmaceutical interventions (NPIs), such as bans on large gatherings. Our estimates imply that the mean observed level of mask wearing corresponds to a 19% decrease in the reproduction number R. We also assess the robustness of our results in 60 tests spanning 20 sensitivity analyses. In light of these results, policy makers can effectively reduce transmission by intervening to increase mask wearing.

Mixtures of large-scale dynamic functional brain network modes.

Accurate temporal modelling of functional brain networks is essential in the quest for understanding how such networks facilitate cognition. Researchers are beginning to adopt time-varying analyses for electrophysiological data that capture highly dynamic processes on the order of milliseconds. Typically, these approaches, such as clustering of functional connectivity profiles and Hidden Markov Modelling (HMM), assume mutual exclusivity of networks over time. Whilst a powerful constraint, this assumption may be compromising the ability of these approaches to describe the data effectively. Here, we propose a new generative model for functional connectivity as a time-varying linear mixture of spatially distributed statistical "modes". The temporal evolution of this mixture is governed by a recurrent neural network, which enables the model to generate data with a rich temporal structure. We use a Bayesian framework known as amortised variational inference to learn model parameters from observed data. We call the approach DyNeMo (for Dynamic Network Modes), and show using simulations it outperforms the HMM when the assumption of mutual exclusivity is violated. In resting-state MEG, DyNeMo reveals a mixture of modes that activate on fast time scales of 100-150 ms, which is similar to state lifetimes found using an HMM. In task MEG data, DyNeMo finds modes with plausible, task-dependent evoked responses without any knowledge of the task timings. Overall, DyNeMo provides decompositions that are an approximate remapping of the HMM's while showing improvements in overall explanatory power. However, the magnitude of the improvements suggests that the HMM's assumption of mutual exclusivity can be reasonable in practice. Nonetheless, DyNeMo provides a flexible framework for implementing and assessing future modelling developments.

Machine learning for functional protein design.

Recent breakthroughs in AI coupled with the rapid accumulation of protein sequence and structure data have radically transformed computational protein design. New methods promise to escape the constraints of natural and laboratory evolution, accelerating the generation of proteins for applications in biotechnology and medicine. To make sense of the exploding diversity of machine learning approaches, we introduce a unifying framework that classifies models on the basis of their use of three core data modalities: sequences, structures and functional labels. We discuss the new capabilities and outstanding challenges for the practical design of enzymes, antibodies, vaccines, nanomachines and more. We then highlight trends shaping the future of this field, from large-scale assays to more robust benchmarks, multimodal foundation models, enhanced sampling strategies and laboratory automation.

ProteinNPT: Improving Protein Property Prediction and Design with Non-Parametric Transformers.

Protein design holds immense potential for optimizing naturally occurring proteins, with broad applications in drug discovery, material design, and sustainability. However, computational methods for protein engineering are confronted with significant challenges, such as an expansive design space, sparse functional regions, and a scarcity of available labels. These issues are further exacerbated in practice by the fact most real-life design scenarios necessitate the simultaneous optimization of multiple properties. In this work, we introduce ProteinNPT, a non-parametric transformer variant tailored to protein sequences and particularly suited to label-scarce and multi-task learning settings. We first focus on the supervised fitness prediction setting and develop several cross-validation schemes which support robust performance assessment. We subsequently reimplement prior top-performing baselines, introduce several extensions of these baselines by integrating diverse branches of the protein engineering literature, and demonstrate that ProteinNPT consistently outperforms all of them across a diverse set of protein property prediction tasks. Finally, we demonstrate the value of our approach for iterative protein design across extensive in silico Bayesian optimization and conditional sampling experiments.

ProteinGym: Large-Scale Benchmarks for Protein Design and Fitness Prediction.

Predicting the effects of mutations in proteins is critical to many applications, from understanding genetic disease to designing novel proteins that can address our most pressing challenges in climate, agriculture and healthcare. Despite a surge in machine learning-based protein models to tackle these questions, an assessment of their respective benefits is challenging due to the use of distinct, often contrived, experimental datasets, and the variable performance of models across different protein families. Addressing these challenges requires scale. To that end we introduce ProteinGym, a large-scale and holistic set of benchmarks specifically designed for protein fitness prediction and design. It encompasses both a broad collection of over 250 standardized deep mutational scanning assays, spanning millions of mutated sequences, as well as curated clinical datasets providing high-quality expert annotations about mutation effects. We devise a robust evaluation framework that combines metrics for both fitness prediction and design, factors in known limitations of the underlying experimental methods, and covers both zero-shot and supervised settings. We report the performance of a diverse set of over 70 high-performing models from various subfields (eg., alignment-based, inverse folding) into a unified benchmark suite. We open source the corresponding codebase, datasets, MSAs, structures, model predictions and develop a user-friendly website that facilitates data access and analysis.

Propagating Uncertainty Across Cascaded Medical Imaging Tasks for Improved Deep Learning Inference.

Although deep networks have been shown to perform very well on a variety of medical imaging tasks, inference in the presence of pathology presents several challenges to common models. These challenges impede the integration of deep learning models into real clinical workflows, where the customary process of cascading deterministic outputs from a sequence of image-based inference steps (e.g. registration, segmentation) generally leads to an accumulation of errors that impacts the accuracy of downstream inference tasks. In this paper, we propose that by embedding uncertainty estimates across cascaded inference tasks, performance on the downstream inference tasks should be improved. We demonstrate the effectiveness of the proposed approach in three different clinical contexts: (i) We demonstrate that by propagating T2 weighted lesion segmentation results and their associated uncertainties, subsequent T2 lesion detection performance is improved when evaluated on a proprietary large-scale, multi-site, clinical trial dataset acquired from patients with Multiple Sclerosis. (ii) We show an improvement in brain tumour segmentation performance when the uncertainty map associated with a synthesised missing MR volume is provided as an additional input to a follow-up brain tumour segmentation network, when evaluated on the publicly available BraTS-2018 dataset. (iii) We show that by propagating uncertainties from a voxel-level hippocampus segmentation task, the subsequent regression of the Alzheimer's disease clinical score is improved.

Technology readiness levels for machine learning systems.

The development and deployment of machine learning systems can be executed easily with modern tools, but the process is typically rushed and means-to-an-end. Lack of diligence can lead to technical debt, scope creep and misaligned objectives, model misuse and failures, and expensive consequences. Engineering systems, on the other hand, follow well-defined processes and testing standards to streamline development for high-quality, reliable results. The extreme is spacecraft systems, with mission critical measures and robustness throughout the process. Drawing on experience in both spacecraft engineering and machine learning (research through product across domain areas), we've developed a proven systems engineering approach for machine learning and artificial intelligence: the Machine Learning Technology Readiness Levels framework defines a principled process to ensure robust, reliable, and responsible systems while being streamlined for machine learning workflows, including key distinctions from traditional software engineering, and a lingua franca for people across teams and organizations to work collaboratively on machine learning and artificial intelligence technologies. Here we describe the framework and elucidate with use-cases from physics research to computer vision apps to medical diagnostics.

A dataset of non-pharmaceutical interventions on SARS-CoV-2 in Europe.

During the second half of 2020, many European governments responded to the resurging transmission of SARS-CoV-2 with wide-ranging non-pharmaceutical interventions (NPIs). These efforts were often highly targeted at the regional level and included fine-grained NPIs. This paper describes a new dataset designed for the accurate recording of NPIs in Europe's second wave to allow precise modelling of NPI effectiveness. The dataset includes interventions from 114 regions in 7 European countries during the period from the 1st August 2020 to the 9th January 2021. The paper includes NPI definitions tailored to the second wave following an exploratory data collection. Each entry has been extensively validated by semi-independent double entry, comparison with existing datasets, and, when necessary, discussion with local epidemiologists. The dataset has considerable potential for use in disentangling the effectiveness of NPIs and comparing the impact of interventions across different phases of the pandemic.

Towards global flood mapping onboard low cost satellites with machine learning.

Spaceborne Earth observation is a key technology for flood response, offering valuable information to decision makers on the ground. Very large constellations of small, nano satellites- 'CubeSats' are a promising solution to reduce revisit time in disaster areas from days to hours. However, data transmission to ground receivers is limited by constraints on power and bandwidth of CubeSats. Onboard processing offers a solution to decrease the amount of data to transmit by reducing large sensor images to smaller data products. The ESA's recent PhiSat-1 mission aims to facilitate the demonstration of this concept, providing the hardware capability to perform onboard processing by including a power-constrained machine learning accelerator and the software to run custom applications. This work demonstrates a flood segmentation algorithm that produces flood masks to be transmitted instead of the raw images, while running efficiently on the accelerator aboard the PhiSat-1. Our models are trained on WorldFloods: a newly compiled dataset of 119 globally verified flooding events from disaster response organizations, which we make available in a common format. We test the system on independent locations, demonstrating that it produces fast and accurate segmentation masks on the hardware accelerator, acting as a proof of concept for this approach.

Inferring the effectiveness of government interventions against COVID-19.

Governments are attempting to control the COVID-19 pandemic with nonpharmaceutical interventions (NPIs). However, the effectiveness of different NPIs at reducing transmission is poorly understood. We gathered chronological data on the implementation of NPIs for several European and non-European countries between January and the end of May 2020. We estimated the effectiveness of these NPIs, which range from limiting gathering sizes and closing businesses or educational institutions to stay-at-home orders. To do so, we used a Bayesian hierarchical model that links NPI implementation dates to national case and death counts and supported the results with extensive empirical validation. Closing all educational institutions, limiting gatherings to 10 people or less, and closing face-to-face businesses each reduced transmission considerably. The additional effect of stay-at-home orders was comparatively small.

Understanding the effectiveness of government interventions against the resurgence of COVID-19 in Europe.

European governments use non-pharmaceutical interventions (NPIs) to control resurging waves of COVID-19. However, they only have outdated estimates for how effective individual NPIs were in the first wave. We estimate the effectiveness of 17 NPIs in Europe's second wave from subnational case and death data by introducing a flexible hierarchical Bayesian transmission model and collecting the largest dataset of NPI implementation dates across Europe. Business closures, educational institution closures, and gathering bans reduced transmission, but reduced it less than they did in the first wave. This difference is likely due to organisational safety measures and individual protective behaviours-such as distancing-which made various areas of public life safer and thereby reduced the effect of closing them. Specifically, we find smaller effects for closing educational institutions, suggesting that stringent safety measures made schools safer compared to the first wave. Second-wave estimates outperform previous estimates at predicting transmission in Europe's third wave.

Changing composition of SARS-CoV-2 lineages and rise of Delta variant in England.

BACKGROUND: Since its emergence in Autumn 2020, the SARS-CoV-2 Variant of Concern (VOC) B.1.1.7 (WHO label Alpha) rapidly became the dominant lineage across much of Europe. Simultaneously, several other VOCs were identified globally. Unlike B.1.1.7, some of these VOCs possess mutations thought to confer partial immune escape. Understanding when and how these additional VOCs pose a threat in settings where B.1.1.7 is currently dominant is vital. METHODS: We examine trends in the prevalence of non-B.1.1.7 lineages in London and other English regions using passive-case detection PCR data, cross-sectional community infection surveys, genomic surveillance, and wastewater monitoring. The study period spans from 31st January 2021 to 15th May 2021. FINDINGS: Across data sources, the percentage of non-B.1.1.7 variants has been increasing since late March 2021. This increase was initially driven by a variety of lineages with immune escape. From mid-April, B.1.617.2 (WHO label Delta) spread rapidly, becoming the dominant variant in England by late May. INTERPRETATION: The outcome of competition between variants depends on a wide range of factors such as intrinsic transmissibility, evasion of prior immunity, demographic specificities and interactions with non-pharmaceutical interventions. The presence and rise of non-B.1.1.7 variants in March likely was driven by importations and some community transmission. There was competition between non-B.1.17 variants which resulted in B.1.617.2 becoming dominant in April and May with considerable community transmission. Our results underscore that early detection of new variants requires a diverse array of data sources in community surveillance. Continued real-time information on the highly dynamic composition and trajectory of different SARS-CoV-2 lineages is essential to future control efforts. FUNDING: National Institute for Health Research, Medicines and Healthcare products Regulatory Agency, DeepMind, EPSRC, EA Funds programme, Open Philanthropy, Academy of Medical Sciences Bill,Melinda Gates Foundation, Imperial College Healthcare NHS Trust, The Novo Nordisk Foundation, MRC Centre for Global Infectious Disease Analysis, Community Jameel, Cancer Research UK, Imperial College COVID-19 Research Fund, Medical Research Council, Wellcome Sanger Institute.