AutoML stands for Automated Machine Learning, and it describes the process of automating various tasks in model training pipelines, including the autonomous exploration and evaluation of different combinations of model architectures and hyperparameters.
The goal of AutoML is to reduce the amount of manual effort required to build and deploy machine learning models, thereby making it easier for organizations to leverage the power of machine learning.
AutoML techniques are now the state-of-the-art approach for hyperparameter tuning of large deep neural networks. AutoML is also used with less success (and more cost) by some platforms to find an appropriate model architecture (model selection). For example, DataRobot can train in parallel (and then compare) a random forest model with a gradient-boosted decision tree with a feed-forward neural network. Where AutoML has had less success to date is in data pre-processing, feature engineering, and model deployments. In contrast, ChatGPT and GPT-4 have had more success in generating source code for data pre-processing using well-structured prompts. Deep learning has had huge success in automating feature engineering in fields such as computer vision and NLP.