Gemini 2.0 — Is it finally good?

Day 348 / 366

Up until a few years ago, Google was the clear Tech Giant. It was supposed to be the best in AI as well. When you are building a good AI model, apart from the actual model logic, what you need is good data. And who’s got more data about us than Google?

But sadly, in the new age of LLMs, Google has failed to maintain its top spot. OpenAI, Meta, and Anthropic, all came up with models that performed far better than Google’s flagship Gemini models. What made it worse was Google’s Left-wing ideology that crept into its models, leading to such a PR nightmare that Google had to scrap the image generation features of Gemini altogether.

Two days ago, Google introduced Gemini 2.0, a new set of models focusing more on AI agent-like features.

https://www.youtube.com/watch?v=Fs0t6SdODd8

AI Agents are one step ahead of the regular LLMs. Not only can they generate stuff, but they can use tools such as search, web browsing, code execution etc.

For instance, they have Project Astra, an assistant that lets you understand the world around you. Basically all you do is open your camera and point it at stuff while asking the assistant questions. It will analyze the image as well as your voice in real-time and answer you.

They have also announced similar agentic tools to help developers. Like always though, all of this right now is just demos and nothing for the general public to use and test. So we will have to wait and see if any of these tools are actually any good.