The problem
Today, we stand at a crossroads. The vast majority of the world's high-value content, from cutting-edge research papers to insightful news articles and transformative YouTube videos, is not accessible to most people, because it's in a language that they cannot understand. Despite the global prevalence of English, 6 billion people worldwide cannot access this content due to language barriers. This limitation is not just a blocker for their lives but for human progress.
But hasn't AI solved this?
It has already been seven years since the publication of the seminal "Attention is All You Need" paper by Google Research, introducing the Transformer model that revolutionized natural language processing. Following this, the advent of ChatGPT just two years ago promised a future where language barriers could be effortlessly overcome. Yet, the translation landscape remains largely unchanged. The potential of AI in translation has not been fully realized in the tools used by translation teams or in platforms that host high-value content. This stagnation is unacceptable.
Our vision of accelerated translation
We propose a new concept: Accelerated Translation. This approach is based on AI to instantly generate and verify translations that learns from human feedback (RLHF). By doing so, we aim to create a world where language is no longer a barrier to knowledge and progress. Accelerated Translation is not just about translating text; it's about empowering every individual with the ability to share and access information across linguistic divides.
Our call to action
The time for incremental improvements is over. We need a paradigm shift that brings the best of AI advancements into the hands of those who bridge cultures through translation. This manifesto is a call to developers, researchers, and content creators to prioritize the development of tools that harness the power of AI for translation in a way that is accessible, efficient, and, above all, inclusive.