Google has announced its latest AI model, Gemini 1.5 Pro, which promises significant improvements in performance compared to its predecessor, Gemini 1.0 Ultra. This new version not only matches but exceeds the capabilities of the previous model, all while utilizing less computing power. The standout feature of the Gemini 1.5 Pro is its ability to process large files, including text, video, and computer code.
While Gemini 1.0 Ultra was limited to processing 32,000 “tokens” or approximately 20,000 words per query, the new Gemini 1.5 Pro can handle an impressive 1 million tokens per query. This means it can efficiently process vast amounts of information in a single go, such as one hour of video, 11 hours of audio, codebases with over 30,000 lines of code, or over 700,000 words.
Google demonstrated the improved capabilities of Gemini 1.5 Pro with two impressive demos. In one demo, the model successfully searched through a 402-page Apollo 11 Moon landing transcript to find specific text snippets. In another demo, Gemini 1.5 Pro visually searched through a 44-minute Buster Keaton film, accurately identifying the precise moment each scene appeared in the film.
Google’s CEO of Deepmind, Demis Hassabis, emphasized that Gemini 1.5 Pro outperforms its predecessor in 87% of the benchmarks used for developing large language models. The significant upgrade in performance is attributed to an AI modeling technique called Mixture-of-Experts (MoE), which selectively activates the most relevant expert pathways in the model’s neural network based on the type of input given.
However, one downside to the Gemini 1.5 Pro model is that it can take some time to respond to requests involving the processing of large files. In Google’s demos, it took up to a minute to produce answers. Nevertheless, Google plans to refine the model to decrease processing times.
While Google has not provided a specific launch date for Gemini 1.5 Pro, it will initially roll out access to developers and enterprise customers through Google’s AI Studio and Vertex AI services. The company also indicated that the model will not be free for processing large files, and pricing tiers will be introduced in the future.
Overall, Google’s Gemini 1.5 Pro AI model represents a significant leap forward in performance and processing capabilities. Its ability to handle large files opens up new possibilities for information analysis across various formats, making it an invaluable tool for developers and enterprises alike.
Frequently Asked Questions (FAQs):
1. What is Google’s latest AI model called, and how does it compare to its predecessor?
– Google’s latest AI model is called Gemini 1.5 Pro, and it promises significant improvements in performance compared to Gemini 1.0 Ultra. It not only matches but exceeds the capabilities of the previous model while utilizing less computing power.
2. What standout feature does Gemini 1.5 Pro have?
– The standout feature of Gemini 1.5 Pro is its ability to process large files, including text, video, and computer code.
3. How many tokens per query can Gemini 1.5 Pro handle?
– Gemini 1.5 Pro can handle an impressive 1 million tokens per query, allowing it to efficiently process vast amounts of information in a single go.
4. How did Google demonstrate the improved capabilities of Gemini 1.5 Pro?
– Google demonstrated the improved capabilities of Gemini 1.5 Pro with two impressive demos. In one demo, the model successfully searched through a 402-page Apollo 11 Moon landing transcript to find specific text snippets. In another demo, Gemini 1.5 Pro visually searched through a 44-minute Buster Keaton film, accurately identifying the precise moment each scene appeared.
5. What AI modeling technique contributes to the significant upgrade in performance of Gemini 1.5 Pro?
– The significant upgrade in performance is attributed to an AI modeling technique called Mixture-of-Experts (MoE), which selectively activates the most relevant expert pathways in the model’s neural network based on the type of input given.
6. What is a downside to the Gemini 1.5 Pro model?
– One downside to the Gemini 1.5 Pro model is that it can take some time to respond to requests involving the processing of large files. In Google’s demos, it took up to a minute to produce answers. However, Google plans to refine the model to decrease processing times.
7. How can developers and enterprise customers access Gemini 1.5 Pro?
– Initially, access to Gemini 1.5 Pro will be rolled out to developers and enterprise customers through Google’s AI Studio and Vertex AI services.
8. Will processing large files with Gemini 1.5 Pro be free?
– No, Google has indicated that the model will not be free for processing large files, and pricing tiers will be introduced in the future.
Key Terms:
– AI model: A system or software that mimics human intelligence and can perform tasks or solve problems that typically require human intelligence.
– Tokens: Units of text that can be individual words or even smaller components, used to represent language data in natural language processing tasks.
– Neural network: A computational model inspired by the biological neural networks of the human brain, used in machine learning to perform complex tasks.
Related Links:
– Google AI
– Google AI Blog
The source of the article is from the blog lokale-komercyjne.pl