Optical Character Recognition (OCR) is an essential building block supporting many research activities and products. Consequently, Google strives for better OCR. Google has been developing an in-house OCR system supporting many languages, covering various domains, and running on multiple platforms. In this talk, the algorithms, design, and philosophy behind the Google's OCR system are presented. The talk also refers to the interactions between OCR and Spoken Language Processing (SLP) studies. Although OCR and SLP are in different domains, we can find analogous machine learning problems in them. Finally, unsolved problems for OCR are discussed. There may be an opportunity to share technologies to solve the unsolved problems in both fields.