Google's Gemini AI Model
Google recently introduced its new generative AI model called Gemini. In a prerecorded video demo, Google showcased Gemini's unique features, such as its ability to engage in audible conversations. However, it was later revealed that the video was edited to enhance the AI model's capabilities beyond reality. Google acknowledged the editing in the YouTube description of the demo, stating that the latency was reduced, and Gemini outputs were shortened for brevity.
The staged nature of the demo did not go unnoticed, with critiques coming from various sources like Bloomberg and TechCrunch. In response, Google conducted "live demos" to showcase its new Gemini Pro tools, AI Studio and Vertex AI, allowing developers to create generative AI applications. AI Studio is a web-based tool for prompt creation, while Vertex AI offers customization options for Gemini, including fine-tuning with proprietary data and enhanced security measures.
Demonstrations and Features
During the media previews on Dec. 11 and 12, Google presented live demonstrations of AI Studio and Vertex AI. Katie Camacho, a Google spokesperson, confirmed the live nature of the demos, emphasizing transparency in showcasing Google's generative AI capabilities amid fierce competition from companies like OpenAI and Anthropic.
Thomas Kurian, CEO of Google Cloud, introduced the "live demonstrations" and reiterated the Gemini Pro announcements. Josh Woodward, VP of Google Labs, showcased AI Studio's capabilities by generating real estate property descriptions from images uploaded from Google Drive. Nenshad Bardoliwalla, Director of Product Management of Vertex AI, demonstrated the platform's features, including multilingual support and real-time web data utilization for tasks like adjusting rental prices.
Competition and Market Dynamics
With OpenAI leading the AI landscape, Google faces immense pressure to surpass its competitors. Google claims that Gemini's most advanced model outperforms OpenAI's GPT-4 in seven out of eight industry benchmarks. Despite this, Google needs to attract developers to adopt Gemini Pro over existing solutions like GPT-4.
Google offers free access to Gemini Pro through AI Studio and Vertex AI, with limitations, until its full release early next year. Following the initial period, developers will be charged based on character count or image usage across both platforms. Google aims to entice developers with competitive pricing, as highlighted by Kurian during the press briefing.
