The value proposition of artificial intelligence (AI) lies beyond consumer-based chatbots and process agents. Getting to AI that can process 10K or 100K documents at scale requires a better architecture than generative pre-trained transformers (GPT) can offer. Byte-Latent Transformer (BLT) processes raw bytes as part of a more efficient architecture, enabling the next generation of AI to eliminate the biggest cost bottleneck in AI: the tokenizer. The removal of the tokenizer also gets rid of the "token tax" that non-English AI users 'pay' to use GPT-based systems. Our architecture will enable us to serve a much larger user base (Latin-derived, 2-to-3-byte Cyrillic, 3-byte Chinese-Japanese-Korean, Sanskrit, and all UTF-8 languages). BLT can open global markets that are closed today due to the tokenization ingestion architecture.
Our focus is outside chatbots and agentic systems because growth in organizational environments will drive the future. Our goal is to emphasize the empowerment of human decision-makers rather than the wholesale replacement of human workers with machines. We call it Augmented Intelligence (AuI): Striving for deeper human enablement by empowering human decision-makers with our AuI Executive, AuI Manager, AuI PMO, and AuI Everything, offering deep knowledge of the past and present to develop the future.