Additional details about Gemini were provided by DeepMind's CEO, Demis Hassabis, who stated that Gemini will be a series of models with different sizes and capabilities. The system may also use memory, fact-checking, and improved reinforcement learning to enhance accuracy. Pichai indicated that Gemini and future iterations will become universal personal assistants integrated into daily life. Despite the promising early results, there is competition from other companies like Meta, which is also reportedly working on an AI model to compete with OpenAI.
Key takeaways:
- Google's upcoming artificial intelligence system, Gemini, is being developed by the Google DeepMind division and is expected to compete with AI systems like ChatGPT from OpenAI.
- Gemini is designed to be multimodal, integrating text, images, and other data types, and may have future capabilities like memory and planning.
- Google Chief Scientist Jeffrey Dean stated that Gemini will utilize Google’s new AI infrastructure, Pathways, for scaling up training on diverse datasets, possibly making it the largest language model created to date.
- Google CEO Sundar Pichai envisions Gemini and future iterations as "incredible universal personal assistants" integrated throughout people’s daily lives in areas like travel, work, and entertainment.