Man + Machine: How AI Supercharges Us, Says Google DeepMind’s Manish Gupta

Manish Gupta, Senior Director, Google DeepMind believes that Google Gemini will take Indian creativity to new heights, especially in light of a recent Kantar report finding that a large number of Indian people want to improve productivity (72 per cent), their creativity (77 per cent), and communication (73 per cent) using GenAI tools.

"Thus, there is a very big gap between the probable availability of content (in local languages) and its digital availability... hence millions of people's languages in which there are no digital contents," Gupta clarified, stating that Google is attempting to unlock these sources and also inviting the government to make its archives available to train the models.

Gupta elaborates that his team in DeepMind also are attempting to bridge that language chasm wherein they know the response will differ if the identical query is posted in English versus in a regional language. "If you are trained in maths through Tamil medium, you would be able to reply even when an issue has been asked using another language. People do the same thing so comfortably.". But AI systems tend to suffer because it's trapped in their abilities between languages," he explained, noting that his team has achieved "pretty significant progress in ensuring Gemini actually does well on a whole bunch of languages."

Google has also released a new Gemini 2.5 version and a text-to-video Veo 2 product called Deep Research that searches the whole web to provide.

When asked if the entire set of new content creation capabilities will lead to a vicious cycle of synthetic content feeding into training data, Gupta pointed out that Google DeepMind's SynthID Google is now working with Nvidia to insert a signature in all AI-generated content-from text to video.