AI Models Exceeding Compute Thresholds

该数据集包含了基于一种预测未来基础模型数量的模型，对超过特定计算阈值（分别是10的25次方和10的26次方浮点运算）的人工智能模型数量的预测。数据集中不仅包括了达到计算阈值的模型数量的第5、第50和第95百分位预测，还包含了历史回溯数据以验证模型的准确性。该数据集的时间跨度为2024年至2028年，其任务是预测超过计算阈值的人工智能模型数量。

This dataset provides forecasts for the number of artificial intelligence (AI) models that exceed two specific computational thresholds: 10^25 and 10^26 floating-point operations, respectively. All forecasts are generated using a model specialized in forecasting the count of future foundation models. The dataset includes not only the 5th, 50th, and 95th percentile predictions of the number of models meeting these computational thresholds, but also historical backtesting data to verify the accuracy of the forecasting framework. Covering the time span from 2024 to 2028, the core task of this dataset is to forecast the quantity of AI models that surpass the predefined computational thresholds.