Jeff Dean, Google's head scientist, has worked with the company for more than 24 years and has contributed to some of the world's most influential inventions, such as Google's search bar. He is co-leading Google's Gemini project, including a multi-modal big language model family.
Potential of these models to work across multiple languages
On a recent trip to India, Jeff stated that he is highly concerned about the potential of these models to work across multiple languages - up to 1,000 - and that Google is actively working on the subject. "We want the models we're developing today and tomorrow to work in the world's thousand most widely spoken languages. And that includes spoken languages, not just written ones because many languages have little written material but are still widely spoken by millions."
According to Jeff, such a language paradigm could have far-reaching repercussions for India and other African countries. "We will be able to provide access to information in previously underserved areas for people who speak various languages and live in various communities. And that is why we are actively working on this in over a thousand languages."
India's massive army of technical talent positions it ideally to capitalise on the new digital era
Furthermore, India's massive army of technical talent positions it ideally to capitalise on the new digital era that we are all entering, which has far-reaching global ramifications. "India has immense technical talent; there is a long history of computer science students in India producing incredible things. Anyone interested in computer science focuses on machine learning and what it can do. India is well-positioned in this subject because it has many students and professionals in this business. It's a big global issue."
And, as vital as pricey computational equipment is in this new AI era, Jeff believes that enhancing algorithm quality can have a significant impact on progressing this technology - which is critical for countries like India that lack advanced AI infrastructure.
Benefits of the LLM language model
An LLM for 1,000 languages could bridge the digital divide and empower people using lesser-known languages to access information and participate online. Such a model could facilitate communication and knowledge exchange across diverse cultures and languages, helping break down language barriers. By including less common languages, Google could contribute to preserving linguistic diversity and cultural heritage.
While there is no official confirmation or schedule, Google's interest and current capabilities suggest they are actively investigating this potential. However, developing such a complicated model requires tremendous time, resources, and technical challenges.