Google recently launched Gemini AI, marking a major milestone in generative AI and signaling its intent to compete with OpenAI’s ChatGPT, which revolutionized the AI landscape last year.
Sundar Pichai, CEO of Google and Alphabet, in a blog post, outlined the company’s approach towards AI development. He stressed the potential benefits and risks associated with AI technology. Pichai emphasized that Google is just beginning to explore the full potential of AI, adopting a bold yet responsible approach. This involves ambitious research, developing capabilities to benefit society, and establishing safeguards in collaboration with governments and experts to mitigate any emerging risks.
Gemini 1.0: A Leap Beyond Text
Differing from ChatGPT’s text-only interface, Gemini 1.0 has been trained on Google’s extensive data archives and is capable of processing text, images, and videos. The first iteration, Gemini 1.0 Pro, has been integrated into Google’s Bard chatbot in English and is set to expand to over 170 countries.
Starting December 13, Gemini will be accessible to developers through Google’s Cloud API. It will be released in three versions: Pro, already surpassing an earlier version of ChatGPT in several benchmarks; Ultra, designed for data centers and enterprise applications, coming in 2024; and Nano, tailored for Google’s Android platform, including Pixel smartphones. Google plans to introduce more products, pending rigorous trust and safety evaluations.
Bard Advanced: The Next Step in Chatbot Evolution
In 2024, Google is set to introduce Bard Advanced, a new version of its Bard chatbot, enhancing its capabilities and user experience.
The Multimodal Edge of Gemini
Gemini stands out with its multimodal capabilities, adding image and video analysis to the existing text-based AI platforms. As reported by The Verge, Gemini’s strength lies in its unified model that comprehends and interacts with audio and visuals, not just text, a strategic decision from the outset. Unlike OpenAI’s separate models for different modalities, such as DALL-E for images and Whisper for voice, Google has developed a single multisensory model in Gemini.
News source: https://deepmind.google/technologies/gemini/#introduction
