Viewpoint: When Will AI Exceed Human Performance? Evidence from AI Experts

Authors: Katja Grace, John Salvatier, Allan Dafoe, Baobao Zhang, Owain Evans

JAIR 2018 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Here we report the results from a large survey of machine learning researchers on their beliefs about progress in AI. Researchers predict AI will outperform humans in many activities in the next ten years, such as translating languages (by 2024), writing high-school essays (by 2026), driving a truck (by 2027), working in retail (by 2031), writing a bestselling book (by 2049), and working as a surgeon (by 2053).
Researcher Affiliation Collaboration Katja Grace EMAIL John Salvatier EMAIL AI Impacts, Berkeley, USA Allan Dafoe EMAIL Baobao Zhang EMAIL Owain Evans EMAIL Future of Humanity Institute, University of Oxford, UK
Pseudocode No The paper describes survey methods and presents statistical analysis of the responses, but does not include any structured pseudocode or algorithm blocks.
Open Source Code Yes (The Qualtrics file will be shared to enable replication.)
Open Datasets No The paper analyzes data collected from a survey of machine learning researchers. While it mentions that 'The Qualtrics file will be shared to enable replication' (referring to the survey instrument), it does not provide concrete access information (link, DOI, repository) for the collected survey response data itself.
Dataset Splits No The paper describes a survey and its results, categorizing respondents by demographic features like region, time in field, and citation count. It does not involve traditional machine learning dataset splits (e.g., training/test/validation) for model evaluation.
Hardware Specification No The paper describes a survey of AI experts and analyzes the collected data. This type of research does not typically require specialized hardware, and no specific hardware specifications are mentioned for the analysis conducted.
Software Dependencies No The paper states, 'The significance test for the effect of region on HLMI date (Table C.2) was done using robust linear regression using the R function rlm from the MASS package to do the regression and then the f.robtest function from the sfsmisc package to do a robust F-test significance.' While specific R packages (MASS and sfsmisc) are mentioned, their version numbers, as well as the version of R itself, are not provided.
Experiment Setup No The paper describes a survey and its statistical analysis. It does not involve machine learning models and therefore does not include specific experimental setup details such as hyperparameter values, batch sizes, or optimizer settings typically found in model training experiments.