Google Vertex AI Launches New Features to Provide More Fact Support and Reduce Misinformation

AIbase

Published inAI News · 3 min read · Jun 28, 2024

131

Google Cloud's development platform, Vertex AI, is designed to help enterprises leverage Google's machine learning and large language models to build services. New features are being introduced to prevent the spread of inaccurate information by applications and services.

In addition to the Grounding with Google Search feature launched in May, Google has announced that customers can opt to use specialized third-party datasets to enhance the AI results of their services. These third-party datasets are provided by data vendors such as Moody's, MSCI, Thomson Reuters, and ZoomInfo, and are expected to be available in the third quarter of this year. Google is developing these new features to encourage organizations to adopt its "enterprise-ready" generative AI experiences, reducing the frequency of misleading or inaccurate model outputs.

Furthermore, a "High Fidelity Mode" has been introduced, allowing organizations to obtain generated outputs from their own enterprise datasets rather than from Gemini's extensive knowledge base. High Fidelity Mode is powered by a specialized version of Gemini1.5Flash and is currently available for preview through Vertex AI's Experiments tool. Organizations can also allow Google's AI models to draw information from their own company datasets.

Concurrently, Google has expanded the Vector Search feature to support hybrid searches, enabling users to find images by referencing similar graphics. This update is currently available for public preview and allows vector-based searches to be paired with keyword-based text searches to enhance accuracy. Grounding with Google Search will soon offer a "Dynamic Retrieval" feature, automatically selecting whether to fetch information from Gemini's established dataset or Google Search to address prompts that may require frequently updated resources.

Key Points:
⭐ Google introduces third-party dataset functionality to help improve AI results
⭐ Launches "High Fidelity Mode," enabling organizations to obtain generated outputs from their own datasets
⭐ Expands Vector Search feature to support hybrid searches, improving accuracy

Japanese Supermarkets Use AI to Monitor Employees' Smiles, Netizens Say It's Unnecessary

In today's pursuit of efficient and humanized management, Aeon, a large Japanese supermarket chain, has sparked widespread controversy by using AI technology to monitor employee smiles.On July 16, Yahoo Japan reported that Aeon has rolled out an AI system called "Smile-Kun" in 240 stores since July 1. This system analyzes employees' smiles and volume, providing real-time feedback to enhance customer service skills.The Smile-Kun system requires employees to greet it daily, then scores them based

Musk's xAI to Launch Grok 2 Next Month, Grok 3 Expected in December

Elon Musk recently announced in an interview with psychologist and author Jordan Peterson that xAI will release its latest artificial intelligence model, Grok2, next month.Musk stated that Grok2's performance is expected to be on par with GPT-4, having been trained on approximately 15,000 H100 GPUs. Meanwhile, the release of Grok3 is scheduled for December, which Musk claims will be the most powerful AI model.During the interview, Musk further revealed that Grok3 is currently being trained at th

Xiaohongshu Launches First 'AI Convenience Store' to Support Quality AI Notes with Traffic Boosts of 30,000 to 500,000

Recently, Xiaohongshu officially announced the launch of an intriguing tech account named "TechPotato," introducing its entrepreneurial project: the world's first AI-powered convenience store.According to official information, the "AI Convenience Store" is a collaboration between TechPotato and top authors across the web, along with renowned tech companies, to curate the most authentic and visually appealing AI-related content, which is then offered as "limited edition products" for continuous o

Microsoft Research Introduces AI Framework E5-V: Simplifying Multimodal Learning with Text Pair Unimodal Training to Reduce Costs

Recently, a research team from Microsoft Research and Beihang University has jointly introduced a novel framework called E5-V, aimed at providing a more efficient solution for multi-modal embeddings. With the continuous advancement of artificial intelligence, multi-modal large language models (MLLMs) have become a focal point of research, as they are capable of understanding both textual and visual information simultaneously, thereby better handling complex data relationships. However, effective