A naukri.com initiative
TechnologyReview
1d
192
Image Credit: TechnologyReview
A Google Gemini model now has a “dial” to adjust how much it reasons
- Google DeepMind’s latest update to the Gemini AI model includes a dial to control reasoning intensity in responses to save money for developers and prevent overthinking.
- The shift towards reasoning models aims to handle complex tasks by logically working through problems, enhancing existing models without building from scratch.
- However, excessive thinking can lead to increased costs, with some tasks costing over $200 to complete, impacting both finances and energy usage.
- Gemini Flash 2.5 introduces a slider to dial down unnecessary thinking, addressing the issue of models overthinking and reducing running expenses.
- Despite improved performance in specific tasks, overthinking remains prominent, leading to inefficient outcomes and increased computing expenditures.
- The introduction of the reasoning dial aims to balance the need for intensive thinking, allowing developers to manage computing resources based on task requirements.
- The focus on reasoning models signifies a shift in improving model efficiency through longer thinking times rather than just scaling up model sizes.
- Google DeepMind's approach to reasoning leverages human-readable language for intuitive communication, even though AI models don't reason like humans do.
- The rise of reasoning models challenges the traditional paradigm of making models larger, indicating a transformation in AI development strategies.
- The trend towards reasoning models impacts both computing expenses and environmental footprints, emphasizing the need for efficient AI solutions.
- While reasoning models continue to evolve, the availability of open-source models like DeepSeek presents cost-effective alternatives for developers seeking powerful AI solutions.
Read Full Article
11 Likes
For uninterrupted reading, download the app