This research presents Ankh, a protein language model trained on Google's TPU-v4 that surpasses the state-of-the-art performance with fewer parameters. Ankh can be used in the protein-specific optimization of AI models to interpret the language of life optimally. The model has been tested on a range of structure and function benchmarks and succeeds in learning protein evolutionary conservation-mutation trends and introducing functional diversity while retaining key structural-functional characteristics. This research promotes accessibility to research innovation via attainable resources.

Msanii is a novel diffusion-based model for synthesizing long-context, high-fidelity music efficiently. The model combines the expressiveness of mel spectrograms, the generative capabilities of diffusion models, and the vocoding capabilities of neural vocoders. The model has been tested by synthesizing tens of seconds of stereo music at high sample rates without the use of concatenative synthesis, cascading architectures, or compression techniques. This research opens up new possibilities for cost-effective high-fidelity music synthesis.