Gemini 1.5 Pro, an update from Google, now lets the model listen to audio files and extract information without needing a written transcript.
At the Google Next event, they mentioned making Gemini 1.5 Pro accessible to everyone through their platform, Vertex AI, for building AI applications. Gemini 1.5 Pro was initially introduced in February..
The latest Gemini Pro model outperforms the largest model, Gemini Ultra. Google says Gemini 1.5 Pro can understand complex commands and doesn’t require fine-tuning. However, it’s only accessible to those with access to Vertex AI and AI Studio.
Currently, many interact with Gemini models through the Gemini chatbot. Although Gemini Ultra powers the Gemini Advanced chatbot and understands lengthy commands, it’s not as speedy as Gemini 1.5 Pro.
Google is updating Imagen 2, a text-to-image model used in Gemini’s image-generation capabilities. This update includes features like inpainting and outpainting, allowing users to add or remove elements from images.
Additionally, Google is making its SynthID digital watermarking feature available on all pictures created through Imagen models. SynthID adds an invisible watermark to images, indicating their source when viewed with a detection tool.
Similar features, such as inpainting and outpainting, are already available in other text-to-image models like Stability AI’s Stable Cascade and Getty’s Generative AI by iStock. These features are also becoming more widely accessible on newer Samsung Galaxy phones.
Google is now offering a public preview of a feature that connects its AI responses with Google Search, ensuring they provide current information. This is important because responses generated by large language models may not always be up-to-date, sometimes intentionally.
For example, Google has intentionally prevented Gemini from answering questions about the 2024 US election. Recently, Gemini faced criticism for producing photos featuring historically inaccurate individuals.
Leave a Reply