Rumors that Google is delaying the launch of their latest AI model Gemini until next year seem far off the mark after its launch just now. Gemini is the largest and most powerful AI model ever developed by Google with the help of DeepMind and Google Research.
Gemini 1.0 is offered in three different sizes to suit your needs. Gemini Ultra is the most powerful model with the ability to complete complex tasks. Gemini Pro is the best model that can be used for various tasks. Lastly is the Gemini Nano which is a model that can run on mobile devices to offer AI readily available.
According to Google, Gemini scored the highest compared to other LLM models in 30 out of 32 benchmark tests performed. It is a multi-modal model that can understand information in the form of text, images, audio, video and code.
Among the capabilities of Gemini 1.0 shown is reading over 250,000 research journals for a specific topic in the same time taken for lunch. Then the user can instruct Gemini to filter the journals most relevant to the topic being studied. With multi-modal capabilities, all the graphics in the journal can be understood by him.
For parents, the feature of helping children complete their schoolwork may be the most attractive. Homework and math homework can be checked by Gemini based on uploaded pictures. If the child's math answer is wrong, Gemini can explain where the mistake was made and show the correct way of doing things.
Gemini can also generate code based on popular programming languages such as Python, Java, C++, and Go. Using Gemini, AI AlphaCode 2.0 has been successfully developed with higher capabilities compared to AlphaCode shown last year which is capable of producing code as good as humans.
At this time Gemini Pro will begin to be integrated into products such as Bard. It is accessible to Bard users in 170 countries starting today but in English only. This seems to confirm earlier reports that Gemini has issues generating answers in languages other than English.
Gemini Nano will be given to Pixel 8 Pro devices with the Summarize feature in the Recorder app. But starting next year it will also improve the ability of Smart Reply on Gboard on WhatsApp before expanding to other Google products in stages.
Finally, Gemini Ultra is still in the process of security testing and last-minute details by a small number of customers, business partners and developers before being offered to business customers starting in early 2024.