Google seeks to monetize AI, rival Microsoft with new Gemini model | Mint

New Delhi: Google on Wednesday launched Gemini 2.0, its new generation of artificial intelligence (AI) models, that it says is equipped to break down complex queries, burnishing the tool’s appeal as a more useful, practical product for businesses to adopt.

“Over the last year, we have been investing in developing more agentic models, meaning they can understand more about the world around you, think multiple steps ahead, and take action on your behalf, with your supervision. Today, we’re excited to launch our next era of models built for this new agentic era—Gemini 2.0, our most capable model yet,” Google and Alphabet chief executive Sundar Pichai said in a blog post on Wednesday. “With new advances in multimodality, like native image and audio output, and native tool use, it will enable us to build new Al agents that bring us closer to our vision of a universal assistant.”

Agentic AI helps enterprises automate complex tasks by using models trained on specific domains such as pharma, e-commerce and many others. For instance, in human resources, agentic AI algorithms help executives get a qualitative analysis of job applicants.

Experts believe that the new, more powerful model is Google’s attempt to catch up with Microsoft that has taken early lead in enterprise AI adoption.

Product testing with Gemini 2.0 with a select number of developers commenced on Wednesday, Google said in a pre-launch roundtable. “We are working quickly to get it into our products, leading with Gemini and Search. Starting today, our Gemini 2.0 Flash experimental model will be available to all Gemini users. We’re also launching a new feature called Deep Research, which uses advanced reasoning and long context capabilities to act as a research assistant, exploring complex topics and compiling reports on your behalf,” Pichai added.

Improved performance

A set of technical benchmarks published by Google during Gemini 2.0’s launch said that the new model is nearly 9% better than its predecessor, Gemini 1.5 Pro. It also performed 4% better in mathematical problems, and 5% better in factual authenticity and verification. The new model is also available in the company’s experimental Project Astra, which uses generative AI to offer live ambient object recognition and more.

However, Google, during its roundtable, did not respond to Mint’s queries on the size of the model, the incremental cost of development of the new model, and how significantly have datasets used to train Gemini 2.0, versus its predecessor, differed. The new model is natively multimodal to support live audio and video streams, Google said in several of its blog posts on Wednesday, indicating significant effort in the development of the new model.

Gemini 2.0 will be a key competitor to Microsoft and OpenAI’s partnership on the latter’s Generative Pre-trained Transistors (GPT) AI models, which industry stakeholders said is key to monetizing investments in AI.

“For Google, the dilemma is that of trying to match up with its own pioneering role in foundational AI research, and keeping up with OpenAI’s rapid pace of innovation and leadership globally. Google’s product adoption across enterprises has not been up to the scale that the company would like it to be. This is to the same effect as the company’s market position in the cloud environment, which in turn justifies its push towards enterprise adoption through agentic AI,” said Kashyap Kompella, AI expert and founder of tech consultancy firm RPA2AI Research.

Jayanth Kolla, co-founder of consultancy firm Convergence Catalyst, said that the enterprise push for agentic AI adoption is a natural move. “The likes of Google and OpenAI build large language models that are massive in size. Tweaking models to make them suitable for agentic AI is one or two levels down the pecking order of customization. Google’s pitch with the new models is to push its AI models to corporate generative AI development—in the same way that Microsoft claimed fine tweaks to make Azure the preferred cloud platform for enterprises. This can help it monetize its AI investments better, where OpenAI has a significant lead despite Google’s foundational work in this field,” he said.

Dave Citron, senior director of product management for Gemini at Google, largely concurred with this analysis in his blog post on the release of the new models—but added that the company also seeks to use its own models across its own products, as Kolla said.

“These releases are a significant step forward to our goal to build the world’s most helpful personal AI assistant. They also represent a step towards making our products more agentic; as Gemini gets better, it can act on your behalf to get even more done,” Citron said.