Google’s AI system, Gemini, makes significant advancements

Google unveiled its highly anticipated next-generation foundation model, Gemini, on Tuesday. According to Google, Gemini has been tested against GPT-4 and has outperformed OpenAI’s LLM in 30 out of 32 benchmarks, including reasoning, math, reading comprehension, and Python code generation. The integration of Gemini Pro into Google Bard has further enhanced its capabilities. Since its launch two days ago, experts and users on social media platforms have expressed both positive and negative reactions to Google’s latest AI innovation.

Google Gemini: What is it?

Google DeepMind has developed Google Gemini in collaboration with other teams such as Google Research. It is a general-purpose AI that can help build a variety of AI services that can work in many different fields to automate tasks. It has been released in three sizes – Nano, Pro and Ultra.

While the Gemini Nano model can run natively on Android devices like the Google Pixel, the Gemini Pro is integrated with Google Bard, the company’s own AI chatbot. On the other hand, Gemini Ultra is the top version. Google claims it is even more powerful than ChatGPT and the most powerful Large Language Model (LLM) ever created.

Google Gemini: Reactions from the tech community

For X, MosaicML research engineer Mihir Patel posted screenshots of a comparison between Gemini-based Google Bard and GPT-based ChatGPT. Patel asked the question “What is Mamba in deep learning?”

After reviewing the response, Patel said: “Gemini is so good. SO much better and SO much faster than GPT-4. I’m sold – switch.”

However, not all users have been impressed with Google’s new basic model. In particular, coders have expressed their disappointment with Gemini Pro-powered Bard’s features in X.

Another X user Filip Piekniewski posted: “Tried gemini-based Bard and still can’t write intersection of two polygons. It’s one of the few relatively simple functions that was never implemented in python, no stack overflow, and all these models fail at it .

Another user named Mohsen Azimi said: “Try Gemini Pro: it’s quite disappointing for my example. I asked it to make an analog clock using HTML, like this one by ChatGPT. It can refer to the code on Github, but it’s a few ms away… “

Author Edison Ade, who writes about AI and startup growth, said, “GPT 4 is still better than Gemini Pro. Created a Tic Tac Toe game with ChatGPT and Bard (used on Gemini Pro).

Posting a video of the created game, he wrote: “ChatGPT wrote the code on the first try (first video). Bard is 3 tries (second video).