Google's AI Chatbot Gemini— Guardrails, Capabilities, How It Stacks Up Against ChatGPT
NDTV Profit took a demo of the app, which was launched with support for nine Indian languages.
Google on Tuesday launched its Gemini AI app in India, with support for nine Indian languages, heating up the generative AI chatbot race.
The Gemini mobile app, which Google is "particularly excited about in India given the country's strong mobile-first culture," has taken on the mantle to crack Indic languages in generative AI and supports English and nine Indian languages: Hindi, Bengali, Gujarati, Kannada, Malayalam, Marathi, Tamil, Telugu, and Urdu.
"This marks a significant step forward in our journey to build a truly conversational, multimodal, and helpful AI assistant," Amar Subramanya, Google's Vice President, Engineering, Gemini Experiences, said in a blog post.
NDTV Profit took a demo of the app to explore its capabilities, guardrails and how it stacks against OpenAI's ChatGPT.
Guardrails
Gemini didn't answer questions related to politics or the recently held Lok Sabha elections in 2024.
A simple response of "I'm still learning how to answer this question. In the meantime, try Google Search." is the standard template response to questions like who Narendra Modi is, who the current prime minister of India is, and who won the 2024 Lok Sabha elections.
There are obvious guardrails on medical advice as well, given the liabilities platforms like these could bear if it acts upon users in an unfavourable way.
Like other chatbots, Gemini also provides disclaimers against misinformation and encourages users to double-check anything that's shared.
Such guardrails protect Big Tech dabbling in generative AI to not propagate misinformation. In the past, former minister of state for information technology Rajeev Chandrasekhar had said deepfakes were a menace. They also save the company from potential liabilities arising from users and governments across the world.
Capabilities
As per the industry standard, Gemini is multimodal—that is, the app allows users to type, talk, or even add an image to get necessary assistance.
The app, currently, doesn't support creative visualisations or digital images of humans. It returns the message, "We're working to improve Gemini's ability to generate images of people. We expect this feature to return soon and will notify you in release updates when it does."
OpenAI's DALL-E platform, named after surrealist artist Salvador Dali, is currently the leading platform that uses deep learning methodologies to generate digital images from natural language descriptions.
On a second attempt however, it returned an image without featuring a face.
Interestingly, images not involving faces are readily generated.
Gemini vs ChatGPT
Document analysis: In its free plan, ChatGPT allows documents to be uploaded, culling out key information. Gemini currently doesn't have an option to add attachments. However, it will summarise points if text is added to a block.
ChatGPT.
Gemini.
Conversational learning: ChatGPT learns from the conversations it has with users, whereas Gemini does this in a limited manner. "As always, privacy is a top priority. Gemini keeps your files private to you, and they’re not used to train our models," it said in today's app launch statement.
Search engine access: Google's pioneering product has been its search engine, and its chatbot is simply a supercharged, AI-integrated version of the same. Gemini has always been able to access the Internet, whereas ChatGPT has launched this only with its GPT-4o update.
