General-Purpose In-Context Learning by Meta-Learning Transformers

Modern machine learning requires system designers to specify aspects of the learning pipeline, such as losses, architectures, and optimizers. Meta-learning, or learning-to-learn, instead aims to learn those aspects, and promises to unlock greater capabilities with less manual effort. One particularly ambitious goal of meta-learning is to train general-purpose in-context learning algorithms from scratch, using only black-box models with minimal inductive bias. Such a model takes in training data, and produces test-set predictions across a wide range of problems, without any explicit definition of an inference model, training loss, or optimization algorithm. In this paper we show that Transformers and other blackbox models can be meta-trained to act as general-purpose in-context learners. We characterize transitions between algorithms that generalize, algorithms that memorize, and algorithms that fail to meta-train at all, induced by changes in model size, number of tasks, and meta-optimization. We further show that the capabilities of meta-trained algorithms are bottlenecked by the accessible state size (memory) determining the next prediction, unlike standard models which are thought to be bottlenecked by parameter count. Finally, we propose practical interventions such as biasing the training distribution that improve the meta-training and metageneralization of general-purpose in-context learning algorithms. Meta-learning is the process of automatically discovering new learning algorithms instead of designing them manually (Schmidhuber, 1987). An important quality of human-engineered learning algorithms, such as backpropagation and gradient descent, is their applicability to a wide range of tasks or environments. For learning-to-learn to exceed those capabilities, the meta-learned learning algorithms must be similarily general-purpose. Recently, there has been significant progress toward this goal (Kirsch et al., 2019; Oh et al., 2020). The improved generality of the discovered learning algorithms has been achieved by introducing inductive bias, such as by bottlenecking the architecture or by hiding information, which encourage learning over memorization. Methods include restricting learning rules to use gradients (Metz et al., 2019; Kirsch et al., 2019; Oh et al., 2020), symbolic graphs (Real et al., 2020; Co-Reyes et al., 2021), or parameter sharing (Kirsch & Schmidhuber, 2020; Kirsch et al., 2021). While enabling generalization, these inductive biases come at the cost of increasing the effort to design these systems and potentially restrict the space of discoverable learning algorithms. Instead, we seek to explore general-purpose meta-learning systems with minimal inductive bias.