Google Gemini 1.5: Best AI for PrestaShop ? Written on . Posted in Generative AI and LLM.
Google has unveiled Gemini 1.5 on Thursday, the new generation of its AI model with enhanced performances. Alongside, Gemini 1.5 Pro emerges with robust long-context understanding capabilities. This innovation may soon enrich the WizardAI module for PrestaShop, offering eCommerce platforms advanced AI functionalities.
Just a few days after transitioning Bard into Google Gemini, the search engine giant has disclosed its latest strides in artificial intelligence with Gemini 1.5. Focusing on long-context understanding, it expands its information processing capability. Gemini 1.5 Pro, a medium-sized model unveiled concurrently, matches Gemini 1.0 Ultra's performances but with lower computational demand. Let's delve into this further.
Spectacular Performances for Gemini 1.5
Google's ambition with Gemini 1.5 is substantial: to "revolutionize its AI approach" with a new generation that exhibits enhanced performances. This is supported by a Mixture of Experts (MoE) architecture, increasing the model's precision and parameter count. Specifically, this MoE model divides the neural network into "experts," optimizing the model's efficiency by selectively activating the most relevant expert pathways. This pioneering approach by Google will enable Gemini 1.5 to learn complex tasks more swiftly while maintaining quality.
The Pro version, akin to the Ultra 1.0 model, excels in long-context understanding, processing up to 1 million tokens. Efficiency gains allow for quicker iteration and easier deployment. Gemini 1.5 marks a significant advancement, offering developers and businesses the chance to extend their AI capabilities.
Handling Huge Amounts of Data Simultaneously with Gemini 1.5 Pro
Google propels its first multimodal model with preliminary tests around Gemini 1.5 Pro, featuring a standard contextual window of 128,000 tokens, akin to the Ultra 1.0 model. We mentioned a contextual window extending up to 1 million tokens. This capacity enables the model to process 1 hour of video, 11 hours of audio, codebases over 30,000 lines, or more than 700,000 words in a single go. The prospects are indeed promising.
Google also aims to reassure on the security and ethics in its AI models' development. Rigorous testing ensures a thorough risk evaluation, incorporating lessons into governance and development processes. Since the Ultra 1.0 version, the company has committed to enhancing security, employing Red Teaming techniques to assess potential hacking risks. Notably, Gemini 1.5 Pro is available from this Thursday, February 15, in a limited preview via AI Studio and Vertex AI, for developers and businesses.
