Google does not stop, it has stepped on the accelerator with its generative AI: after killing Google Bard and presenting its Gemini Advanced paid subscription, all based on version 1.0 of its model, The company already has Gemini 1.5 ready. Faster, more powerful, and with a change that can make a real difference: Google has dramatically expanded the context its AI can handle.
It’s not that Google has done little for AI, as the company is a pioneer in training, analysis and execution tools. But it was clear that Google has failed to bring this AI to the consumer, Open AI ended up taking the scepter of generative chatbots with the ubiquitous ChatGPT. And of course, Google couldn’t be left behind, which is why it stepped on the accelerator with Gemini. So much so that a new version of the multimodal model is already ready.
10 GOOGLE APPS THAT COULD HAVE SUCCEEDED
Gemini 1.5: up to 1 million tokens compared to 128,000 for GPT-4
Token comparison. Google Image
The numbers that sum up how upset Google is about the dominance of Open AI and its GPT models are the ones that reveal the impressive theoretical potential of new Gemini model, the recently released version 1.5. Google claims to have obtained very high results in the AI benchmark; with 87% superiority compared to the previous version of the model, Gemini Pro 1.0.
Gemini 1.5 can manage environments of up to a million tokens consistently. This extends the understanding of context and model response well beyond current generative AI tools. To contextualize, GPT-4 Turbo supports up to 128,000 tokens. And Gemini Pro 1.0, the version currently used by Google in its free AI open to the public, uses 32,000 tokens.
With the new multimodal model the user will get better responses to their commands, greater creativity in behavior, Gemini 1.5 understands the context better and is also faster in execution; with fewer resources needed for generation tasks.
As Google explains on its blog:
“Gemini 1.5 Pro can process large amounts of information in one go, including 1 hour of video, 11 hours of audio, code bases with more than 30,000 lines of code or more than 700,000 words. In our research, we have also successfully tested up to 10 million tokens.
The new generative AI model will not yet be available to the general public, but developers and some enterprise customers will have access to it soon. The tests will be carried out on Gemini Pro 1.5 with a limitation of 128,000 tokensthe same ones that are currently handled by Gemini Ultra 1.0.
Once the testing period is over, Google will bring the multimodal language model to its three divisions: Nano (on mobile phones, like the Pixel 8 Pro), Pro (in the free version of the Gemini chatbot) and Ultra (Gemini Advanced) . At the moment they have not given dates for updating all the tools currently available for individual and professional users.
In Xataka Android | The Gemini app for Android is here. So you can try Google’s most advanced AI on your mobile
