Pretrained Transformers as Universal Computation Engines
This work addresses the challenge of cross-modal generalization for AI models, showing potential for more efficient and versatile systems, though it is incremental as it builds on existing transformer architectures.
The paper tackles the problem of whether transformers pretrained on natural language can generalize to non-language tasks like numerical computation, vision, and protein fold prediction with minimal finetuning, and finds that they achieve strong performance and improved compute efficiency.
We investigate the capability of a transformer pretrained on natural language to generalize to other modalities with minimal finetuning -- in particular, without finetuning of the self-attention and feedforward layers of the residual blocks. We consider such a model, which we call a Frozen Pretrained Transformer (FPT), and study finetuning it on a variety of sequence classification tasks spanning numerical computation, vision, and protein fold prediction. In contrast to prior works which investigate finetuning on the same modality as the pretraining dataset, we show that pretraining on natural language can improve performance and compute efficiency on non-language downstream tasks. Additionally, we perform an analysis of the architecture, comparing the performance of a random initialized transformer to a random LSTM. Combining the two insights, we find language-pretrained transformers can obtain strong performance on a variety of non-language tasks.