Upload Full-Length Films as Training Material with Google’s Gemini Version 1.5

Google has been rapidly updating its Gemini AI, previously known as Google Bard, over the past few months. Today, the company revealed an updated Gemini 1.5 model currently in testing, with the ability to process more information at once and use videos as input.

Google has revealed Gemini 1.5, an updated version of the model that currently powers the Gemini chatbot and other AI features and services. It’s currently only available in preview for software developers and enterprise customers, but it will presumably roll out to the Gemini chatbot in the near future.

The main promised improvement is a significantly larger “token context window”—the data that can be input information in the AI prompt. Gemini 1.5 Pro has a standard 128,000 token context window, a significant leap from the 32,000 token limit in Gemini 1.0. Google is allowing a limited group of developers and companies to use context windows of up to 1 million tokens, which is enough for one hour of video, 11 hours of audio, or over 700,000 words. Google said it also successfully tested a 10 million token limit.

Google said in a blog post, “[Gemini] 1.5 Pro can perform highly-sophisticated understanding and reasoning tasks for different modalities, including video. For instance, when given a 44-minute silent Buster Keaton movie, the model can accurately analyze various plot points and events, and even reason about small details in the movie that could easily be missed.”

The updated Gemini model is also supposedly better at “understanding and reasoning tasks,” outperforming the earlier 1.0 Pro model in 87% of the benchmarks Google uses to test large language models. That still leaves room for Gemini AI to create incorrect data, like every other generative AI solution, so it remains an imperfect solution for data analysis and facts.

Google didn’t confirm when the regular Gemini chatbot and other Google services will be updated to use Gemini 1.5, but the company did say that “significant improvements in speed are also on the horizon,” so Google might be waiting for that before rolling it out to all users. The new model is available in preview for developers and enterprise customers using AI Studio and Vertex AI.

Source: Google