Google has been using AI behind the scenes for a decade but, like most of the rest of the world, failed to anticipate the generative AI craze set off by ChatGPT.
The company released its answer in the form of Bard in March of 2023 but that release was wildly overshadowed by ChatGPT 4.0 which was a huge leap forward for generative AI and took all the press.
Bard struggled to attract any attention in the gen AI wars and Google was seen as lagging.
Behind the scenes, Google was working on a new large language model (LLM) called Gemini which was released to the world recently.
Google DeepMind CEO Demis Hassabis discussed how Gemini was internally benchmarked against ChatGPT 4.0 saying, “I think we’re substantially ahead on 30 out of 32” of the tests.
Gemini comes in three sizes for different use cases.
First is a lightweight version called Nano, which is being used for offline generative AI for Android phone devices like the Google Pixel.
Next is the Pro version which was immediately moved to underpin Google’s generative AI chatbot, Bard, and is similar to ChatGPT 4.0.
Finally is a Gemini Ultra which is, by the company’s accounts, the most advanced large language model ever, and is capable of input across multiple types of media from text, to pictures, to video, sound, and even touch.
Gemini Ultra is too slow and costly at the moment for wide release, but Google intends to move it under Bard in the next year.
This is all well and good except for the fact that OpenAI is also working on ChatGPT 5.0 which will also include multiple types of media based on its DALL-E image generating model.
What Google has that OpenAI does not is the world’s leading search engine and with a comparable chatbot to ChatGPT it makes this new release very relevant.