DeepSeek’s newly released OCR model presents a significant advancement in AI memory retention techniques. This model extracts text from images and converts it into machine-readable words while utilizing fewer tokens than traditional large language models. Researchers claim that this method of processing could alleviate the problem of ‘context rot’, enabling models to remember relevant information during extended user interactions.
The approach utilizes image-based tokens instead of highlighting only text elements. Such innovation may lead to more efficient AI memory storage, which is likened to how humans retain memories at varying clarity. The potential benefits extend beyond memory to creating more interactive AI agents, as the model can generate substantial training data on a daily basis. Nonetheless, experts urge that further research into memory dynamics is crucial to enhance AI’s capabilities.
👉 Pročitaj original: MIT Technology Review – AI
