Google releases Gemini 1.5 Flash
Gemini is an artificial intelligence model released by Google DeepMind on December 6, 2023. The initial version comprises three models of varying capacities: Gemini Ultra for handling highly complex tasks, Gemini Nano for managing multiple tasks, and Gemini Pro for addressing specific tasks on terminal devices.
On May 14, Google's 2024 I/O developer conference kicked off as scheduled in Mountain View, California. This year's theme revolves entirely around artificial intelligence, particularly generative AI and its practical applications across various domains. During the event, Google unveiled the latest AI model in the Gemini series: Gemini 1.5 Flash. Sundar Pichai, CEO of Google, emphasized Google's commitment to advancing artificial intelligence technology responsibly to ensure positive impacts on society.
The brand new 1.5 Flash
The latest addition to the Gemini series, 1.5 Flash, has been optimized for higher speed and efficiency. Despite being lighter-weight compared to 1.5 Pro, it excels in multimodal reasoning and features a groundbreaking long-context window. 1.5 Flash performs exceptionally well in tasks such as summarization, chat applications, image and video captioning, as well as data extraction from long documents and tables. It undergoes a unique process called "distillation," trained by 1.5 Pro, transferring essential knowledge and skills to a smaller, more efficient model.
Significantly improved 1.5 Pro
In recent months, Google has made significant improvements to 1.5 Pro. Not only has it expanded its context window to 2 million tokens, but it has also enhanced its capabilities in code generation, logical reasoning and planning, as well as multi-turn conversations through advancements in data and algorithms. Today, 1.5 Pro can follow increasingly complex and nuanced instructions, including those involving product-level behaviors such as roles, formats, and styles. Additionally, Google has added audio understanding functionality in the Gemini API and Google AI Studio and continues to integrate 1.5 Pro into Google products, including Gemini Advanced and Workspace applications.
Challenges and Future
In the field of artificial intelligence, OpenAI is undoubtedly a formidable competitor that cannot be ignored. With the high-profile release of Gemini on December 7th last year, Google's market value has continuously increased. However, this has been accompanied by ongoing controversies, such as allegations of fraud and exaggerated claims. Gemini had claimed to surpass GPT-4 in several standard performance metrics. However, in reality, the difference is negligible, and such comparisons are not entirely fair, as each excels in different standards.
This year, the competition in AI has intensified. In February, Google released Gemini 1.5 Pro, but at the same time, Sora also gained traction in the market. Just as the May conference ended, GPT-4o made its debut. In various aspects such as response time, emotional richness of speech, and input-output, GPT-4o's interactive experience seems slightly superior. However, Gemini has not halted its progress, and perhaps in the future, it will introduce an even more powerful Ultra version. This battle is far from over.
