In today's rapidly developing artificial intelligence landscape, the phenomenon of AI 'hallucinations' has become increasingly frequent, causing significant inconvenience for many companies. Customer service chatbots confidently describe non-existent products, financial AIs fabricate market data, and medical bots provide dangerous medical advice. These issues are no longer mere curiosities; they pose serious risks that are affecting company reputation and profitability. To tackle this challenge, San Francisco-based startup Patronus AI has announced the launch of the world's first self-service platform aimed at real-time detection and prevention.

Recently, some media reports indicated that WeChat is conducting gray testing of the AI Q&A feature, allowing some users to input questions in the WeChat search box to trigger the AI Q&A function. The AI Q&A can generate an approximately 400-word response in about 20 seconds and provide 5 reference materials from WeChat official accounts. The test results show that the AI Q&A feature currently mainly cites WeChat official accounts as reference materials.

Apple has recently launched the testing of its new artificial intelligence software suite, 'Apple Intelligence,' in the U.S., attracting attention from tech enthusiasts. This software rollout marks a significant advancement for Apple in the field of artificial intelligence, particularly in writing assistance and voice assistant capabilities. Apple plans to expand these features to other English-speaking markets by December and expects to roll them out to users in the European Union by April 2025. Image source note: Image generated by AI, image licensed by Midj.

Recently, many users reported that Kimi has begun internal testing of the "Exploration Version". Users who have obtained testing privileges can enter " / " in the input box to see the new feature options. From the screenshots of the Kimi Exploration Version interface, the page prompts: Enter " / " to use Kimi Exploration Version to solve complex search problems. It is worth mentioning that since it is in functional testing, each account with testing privileges can only use it 5 times a day. It is understood that Kimi Exploration Version has significantly improved its intelligence and reasoning capabilities.

Ding Jian, the president of Xianyu, announced at the '2024 E-commerce Conference' during the 2024 China International Fair for Trade in Services that Xianyu will launch two new AI products this year, namely 'Smart Publishing' and 'Smart Hosting'. These products are currently undergoing small-scale gray testing within the Xianyu app and are expected to officially launch this year. 'Smart Publishing' allows users to upload just one product image, and the system will use image recognition and natural language generation technology to automatically create a product description that fits the Xianyu style.

Worth Buying Technology announced its 2024 semi-annual report on the evening of August 28, 2024. The report shows that the company achieved operating revenue of 715.81 million yuan during the reporting period, a year-on-year increase of 5.69%; the net profit attributable to shareholders of the listed company was 7.66 million yuan, while the net profit after deducting non-recurring gains and losses was 11.98 million yuan.

Xincheng Lingo is the first end-to-end speech large model in China, with internal testing reservations now open. This model is considered to be close in capability to GPT-4, marking a significant breakthrough for China in the field of speech AI. The core advantages of Xincheng Lingo include native speech understanding, diverse speech style expression, and efficient speech modality compression. It can recognize textual information in speech and capture other important features, providing a natural and vibrant interaction experience, and adjusting speech styles flexibly according to context and user instructions. With a high compression rate speech codec, Xincheng Lingo...

Amidst the widespread adoption of cloud computing and SaaS in the digital transformation of enterprises, startup Opkey has announced the completion of a $47 million Series B funding round, aimed at accelerating the development of its AI-driven ERP testing platform. Developed by Pankaj Goel and his team, the platform addresses functionality issues that may arise from frequent updates to ERP systems. Opkey has collaborated with over 200 major companies, covering well-known platforms such as Oracle, Workday, Salesforce, and SAP, through automated testing.

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.

Geekbench AI is a brand new cross-platform tool designed to evaluate device performance under AI-intensive workloads. It tests the CPU, GPU, and NPU to determine the device's ability to handle machine learning applications. It supports various AI frameworks such as ONNX, CoreML, TensorFlow Lite, and OpenVINO, and evaluates performance from both speed and accuracy dimensions, providing scores in full precision, half precision, and quantized precision. Additionally, Geekbench AI's testing results...

To test the 'stress resistance' of these AIs, a group of researchers embarked on an 'adventure'. They created a test called Self-Contradictory Instructions (SCI), which is essentially the 'death challenge' for AI. This test includes 20,000 self-contradictory instructions, covering both language and visual domains. For example, you might be shown a picture of a cat and asked to describe this 'dog'. Isn't that a challenge for humans? Oh no, it's a challenge for AI.

OpenAI recently released the GPT-4o System Card, detailing the safety measures and risk assessments prior to the launch of the new model. GPT-4o was officially launched in May, and the assessment showed an overall risk rating of 'medium', with major risks concentrated in cybersecurity, biological threats, persuasive capability, and model autonomy. Researchers found that while GPT-4o may be more persuasive in influencing reader opinions, it did not surpass humans overall. At the time of releasing the system card, OpenAI faced criticism from internal employees and state senators questioning its decision-making.

Long context understanding is a key challenge in the field of natural language processing, especially when large language models (LLMs) handle text that exceeds their context window size. To address this issue, researchers developed the LooGLE benchmark test, aimed at assessing LLMs' capability to understand extremely long documents (averaging 19.3k words, comprising 776 articles across multiple domains). LooGLE includes 7 tasks that cover both short and long dependencies, evaluating model understanding across texts of varying lengths. The test data is sourced from open-access content after 2022.

The futuristic scenario depicted in the movie 'Her' is becoming a reality, as OpenAI's GPT-4o model has launched a gradual testing phase for its voice feature, providing users with a more natural and real-time conversational experience. This feature allows users to interrupt freely, can sense and respond to user emotions, and includes 'face-to-face' communication via video and screen-sharing capabilities. GPT-4o's output capability has significantly increased, allowing for an output range of 4000 to 64000 tokens in one go, equivalent to about 200 pages of a novel. During testing, users can use it to tell jokes.

Researchers from the Department of Psychology at the University of Cambridge have recently made a significant breakthrough in the early diagnosis of Alzheimer's disease. They have developed an advanced artificial intelligence tool that can predict the progression of Alzheimer's disease with an accuracy of 80%. This innovative approach aims to reduce reliance on invasive and expensive dementia diagnostic tests while also hoping to significantly improve treatment outcomes in the early stages of th

At the 2024 World Artificial Intelligence Conference (WAIC), SenseTime Technology launched a controllable character video generation large model named Vimi. The Vimi model utilizes SenseTime's advanced large model technology to generate character videos consistent with the target action from a single photo, achieving precise facial and limb control. The model supports various driving methods, including videos, animations, sounds, and text. It boasts years of accumulated facial tracking technolog