In modern relational databases, Cardinality Estimation (CE) techniques are crucial for optimizing query execution plans, directly affecting query efficiency and database performance. Traditional cardinality estimation methods are based on simplified assumptions and often perform poorly on complex query predictions, whereas learning-based CE models can provide more accurate predictions but face challenges such as long training times, requirement of large amounts of data, and lack of systematic evaluation. To address this challenge, the Google research team has launched the CardBench benchmarking framework, integrating over 20 real-world databases and thousands of queries.

Google AI Studio has released a new native prompt library that integrates features such as long-text context, multimodal inputs (images, videos, audio), and structured outputs. The aim is to provide users with a rich and diverse experience across various fields including education, programming, and travel. This library includes practical features like recipe generators, math tutors, math homework generators, treasure hunt design, unit test generators, and travel suggestions, all designed to enhance user exploration and innovation through intuitive displays and comprehensive information. Users only need to log into Google AI Studio.

In the recently concluded International Mathematical Olympiad, Google's DeepMind AI systems, AlphaProof and AlphaGeometry2, showcased remarkable problem-solving abilities, finishing just one point shy of the championship while surpassing human competitors in speed and accuracy. The AI achieved perfect scores on four problems and solved one in an astonishing 19 seconds, adhering to mathematical competition standards with rigorous clarity. This com....

Recently, Google has made a remarkable breakthrough in the field of weather forecasting. They have developed a new atmospheric circulation model called NeuralGCM, which boasts a computational efficiency that is 100,000 times higher than traditional physical models, equivalent to the progress made in high-performance computing over the past 25 years.By combining machine learning with physical modeling, NeuralGCM performs both rapidly and accurately in simulating Earth's atmosphere. The results of

Google's new AI search feature may have issues that could promote fraudulent websites and malware. Users have reported search results pointing to junk sites, posing potential malware risks. Google claims to be addressing the problem, but this incident highlights its challenges in handling junk websites. This issue has raised concerns, showing that Google still faces challenges and areas for improvement in the AI search field.

Google researchers have introduced a pre-trained scoring model called Cappy, aimed at improving the performance of large multitask language models. Cappy's architecture is based on RoBERTa and is pre-trained on a diverse collection of datasets to improve the performance and efficiency of multitask law master's models. By introducing the lightweight pre-trained scoring model Cappy, it addresses the challenge of effectively utilizing large language models in multitask scenarios. Researchers have proposed a data construction method that meets the diversity requirements of labels in pre-training data, generating a large and effective regression pre-training dataset.

Google introduces Gemini Pro, the best model for text and image inference tasks in Android applications. Prototype and run prompts using Google AI Studio, and export the model as code. Android applications can access the Gemini model through the Google AI Client SDK. Gemini also offers multimodal models and a smaller version, Gemini Nano. Gemini Nano can run on-device, ensuring predictable latency and data security for offline applications.

Google AR&VR, in collaboration with a research team from the University of Central Florida, has released a virtual character library called 'VALID' to promote diversity and inclusion. The library includes 210 validated, fully equipped virtual characters representing seven different races, including Asian, Black, and White. Research findings show consistent recognition of Asian, Black, and White virtual characters among participants of various races, but the representation of races such as Native American, Alaskan Native, Hispanic, Middle Eastern, North African, Hawaiian, and Pacific Islander is still lacking.

OpenAI is enticing top AI researchers from Google with a billion-dollar equity incentive plan, with the company's valuation expected to surpass $80 billion. The recruitment is focused on senior AI researchers for the Gemini AI model, with annual salaries primarily offered in the form of stock ranging from $5 million to $10 million. The market is fiercely competitive, as OpenAI not only offers attractive compensation packages to senior staff but also increases salaries for junior employees, leading to an intense talent war with Google.

Demis Hassabis, head of Google AI, believes that the development of artificial intelligence poses a threat to humanity similar to that of climate change, expressing concerns about the potential for uncontrolled superintelligent systems. He emphasized that the risks associated with AI must be taken as seriously as climate change and called for the establishment of an independent body to supervise AI, similar to the Intergovernmental Panel on Climate Change (IPCC). Google, Microsoft, OpenAI, and Anthropic announced the creation of a $10 million AI safety fund aimed at advancing the most capable AIs.

Google's AI chatbot Bard has recently been updated to include support for plugins for Gmail, Docs, and Drive. The updated Google Bard now supports real-time information retrieval from applications like Google Maps, YouTube, hotels, and flights, and allows users to turn it off at any time. Users can quickly select plugins by clicking the plugin icon or typing the plugin name prefix '@'. Google Bard also introduces a public link sharing feature, allowing users to share with Bard.

The Google AI research team has proposed a general method for generating personalized text using large language models. This method involves a multi-stage, multi-task structure that includes retrieval, ranking, summarization, synthesis, and generation, training the large language models for personalized text production. This approach has been validated on three public datasets, showing significant improvements over baseline models.

Google's AI search function has serious issues with geographic knowledge. The AI search cannot correctly answer questions about countries that start with certain letters. Google acknowledges that the AI search is still in the experimental stage and stated that it will continue to improve and optimize.