Perception, performance, and detectability of conversational artificial intelligence across 32 university courses.

The emergence of large language models has led to the development of powerful tools such as ChatGPT that can produce text indistinguishable from human-generated work. With the increasing accessibility of such technology, students across the globe may utilize it to help with their school work-a possibility that has sparked ample discussion on the integrity of student evaluation processes in the age of artificial intelligence (AI). To date, it is unclear how such tools perform compared to students on university-level courses across various disciplines. Further, students' perspectives regarding the use of such tools in school work, and educators' perspectives on treating their use as plagiarism, remain unknown. Here, we compare the performance of the state-of-the-art tool, ChatGPT, against that of students on 32 university-level courses. We also assess the degree to which its use can be detected by two classifiers designed specifically for this purpose. Additionally, we conduct a global survey across five countries, as well as a more in-depth survey at the authors' institution, to discern students' and educators' perceptions of ChatGPT's use in school work. We find that ChatGPT's performance is comparable, if not superior, to that of students in a multitude of courses. Moreover, current AI-text classifiers cannot reliably detect ChatGPT's use in school work, due to both their propensity to classify human-written answers as AI-generated, as well as the relative ease with which AI-generated text can be edited to evade detection. Finally, there seems to be an emerging consensus among students to use the tool, and among educators to treat its use as plagiarism. Our findings offer insights that could guide policy discussions addressing the integration of artificial intelligence into educational frameworks.

Privacy-preserving cancer type prediction with homomorphic encryption.

Cancer genomics tailors diagnosis and treatment based on an individual's genetic information and is the crux of precision medicine. However, analysis and maintenance of high volume of genetic mutation data to build a machine learning (ML) model to predict the cancer type is a computationally expensive task and is often outsourced to powerful cloud servers, raising critical privacy concerns for patients' data. Homomorphic encryption (HE) enables computation on encrypted data, thus, providing cryptographic guarantees to protect privacy. But restrictive overheads of encrypted computation deter its usage. In this work, we explore the challenges of privacy preserving cancer type prediction using a dataset consisting of more than 2 million genetic mutations from 2713 patients for several cancer types by building a highly accurate ML model and then implementing its privacy preserving version in HE. Our solution for cancer type inference encodes somatic mutations based on their impact on the cancer genomes into the feature space and then uses statistical tests for feature selection. We propose a fast matrix multiplication algorithm for HE-based model. Our final model achieves 0.98 micro-average area under curve improving accuracy from 70.08 to 83.61% , being 550 times faster than the standard matrix multiplication-based privacy-preserving models. Our tool can be found at https://github.com/momalab/octal-candet .

Privacy-preserving genotype imputation with fully homomorphic encryption.

Genotype imputation is the inference of unknown genotypes using known population structure observed in large genomic datasets; it can further our understanding of phenotype-genotype relationships and is useful for QTL mapping and GWASs. However, the compute-intensive nature of genotype imputation can overwhelm local servers for computation and storage. Hence, many researchers are moving toward using cloud services, raising privacy concerns. We address these concerns by developing an efficient, privacy-preserving algorithm called p-Impute. Our method uses homomorphic encryption, allowing calculations on ciphertext, thereby avoiding the decryption of private genotypes in the cloud. It is similar to k-nearest neighbor approaches, inferring missing genotypes in a genomic block based on the SNP genotypes of genetically related individuals in the same block. Our results demonstrate accuracy in agreement with the state-of-the-art plaintext solutions. Moreover, p-Impute is scalable to real-world applications as its memory and time requirements increase linearly with the increasing number of samples. p-Impute is freely available for download here: https://doi.org/10.5281/zenodo.5542001.

Fast and Scalable Private Genotype Imputation Using Machine Learning and Partially Homomorphic Encryption.

The recent advances in genome sequencing technologies provide unprecedented opportunities to understand the relationship between human genetic variation and diseases. However, genotyping whole genomes from a large cohort of individuals is still cost prohibitive. Imputation methods to predict genotypes of missing genetic variants are widely used, especially for genome-wide association studies. Accurate genotype imputation requires complex statistical methods. Due to the data and computing-intensive nature of the problem, imputation is increasingly outsourced, raising serious privacy concerns. In this work, we investigate solutions for fast, scalable, and accurate privacy-preserving genotype imputation using Machine Learning (ML) and a standardized homomorphic encryption scheme, Paillier cryptosystem. ML-based privacy-preserving inference has been largely optimized for computation-heavy non-linear functions in a single-output multi-class classification setting. However, having a large number of multi-class outputs per genome per individual calls for further optimizations and/or approximations specific to this application. Here we explore the effectiveness of linear models for genotype imputation to convert them to privacy-preserving equivalents using standardized homomorphic encryption schemes. Our results show that performance of our privacy-preserving genotype imputation method is equivalent to the state-of-the-art plaintext solutions, achieving up to 99% micro area under curve score, even on real-world large-scale datasets up to 80,000 targets.