current location:Home > Technical Articles > Technology peripherals > AI
- Direction:
- All web3.0 Backend Development Web Front-end Database Operation and Maintenance Development Tools PHP Framework Daily Programming WeChat Applet Common Problem Other Tech CMS Tutorial Java System Tutorial Computer Tutorials Hardware Tutorial Mobile Tutorial Software Tutorial Mobile Game Tutorial
- Classify:
-
- Search results are organized in 10 seconds, and brain map tables are generated with one click. Netizen: Search finally looks like it should.
- Recently, there is an AI search tool that is very popular in small circles. When I was crazy about Amway, it was filled with several tags: clean and refreshing, ad-free and smart. Just enter a question and you can quickly filter out high-quality relevant information across the entire network, and provide one-click sorting, summarizing and summarizing functions. Such a magical AI tool, we Qubits must try it, and the result... The preliminary trial conclusion is that this tool can not only be used as a regular search engine, but also can be used as a learning and office tool. Whether you're making a mind map or generating a problem outline, whether you're looking for relevant events or organizations, it has it all. Technology enthusiasts can learn this way. For AI enthusiasts, Sora is a topic that has attracted much attention recently. A little understanding of it will tell you that it is a
- AI 764 2024-03-26 19:06:23
-
- Tsinghua Microsoft open sourced a new prompt word compression tool, the length dropped by 80%! GitHub gets 3.1K stars
- In natural language processing, a lot of information is actually repeated. If the prompt words can be effectively compressed, it is equivalent to expanding the length of the context supported by the model to some extent. Existing information entropy methods reduce this redundancy by deleting certain words or phrases. However, the calculation based on information entropy only covers the one-way context of the text and may ignore key information required for compression; moreover, the calculation method of information entropy is not completely consistent with the actual purpose of compressing prompt words. To meet these challenges, researchers from Tsinghua University and Microsoft jointly proposed a new data processing process called LLMLingua-2. It aims to extract knowledge from large language models (LLM) and achieve information refinement by compressing prompt words while ensuring that key information is not
- AI 314 2024-03-26 18:36:02
-
- Apple chooses Baidu AI, which always feels like a rumor
- On March 25, reporters learned from people familiar with the matter that Baidu will provide AI functions for iPhone 16, Mac system and iOS 18 released by Apple this year. Apple had negotiated with Alibaba and another large domestic model company, and finally decided that Baidu would provide this service. Apple is expected to adopt API interface method for billing. Apple uses domestic large-model AI functions in devices such as the state-owned iPhone mainly for compliance needs. The company cannot solve the compliance problem in the short term. However, the AI functions of foreign versions of iPhones and other devices all come from Apple's own large-scale models. (Science and Technology Innovation Board Daily) Of course, last week, there was another rumor that if you are abroad, Apple chose to cooperate with Google’s Gemini large model to connect to Apple mobile phones.
- AI 492 2024-03-26 17:51:44
-
- Tencent Robot Research Tops the Issue! It can help programmers install monitors and work together like real people
- A new breakthrough for domestic robots: two independent robotic arms can already cooperate smoothly! If you don’t believe it, look at these hands twisting the bottle cap: after twisting, they pick up the cup and pour water: lifelike, like a real person. Now, it can also lend a hand to help programmers install the display: it can even take over the big box from a "colleague": it can be said that objects with various geometric and physical properties can be held firmly. (What else can be done next, I dare not think about it) This is the latest achievement of Tencent’s RoboticsX laboratory: a universal dual-arm cooperative dexterous operation framework. It has been published in the top journal "IEEE Transactions on Robotics" in the field of robotics. Since you are working, you must have the ability to resist interference: Since you are working with both hands, your arms cannot be "stirred" together: the following is not
- AI 621 2024-03-26 17:41:54
-
- CIO shares: How to harness generative AI in the enterprise
- Generative AI is bringing innovative opportunities to enterprises, but in this new era, senior managers need to pay close attention to the application of generative AI to ensure code quality and reduce technical risks. Executives should carefully evaluate the reliability and security of AI solutions and develop effective monitoring measures to detect and correct potential problems in a timely manner. By establishing strict technical standards and oversight mechanisms, companies can better leverage generative AI, which can begin to transform organizations at an early stage and have a profound impact on IT strategy. While large language models accelerate engineering agility, they also create issues with technical debt. Stephen O'Grady, principal analyst and co-founder of RedMonk, pointed out: "Generative systems may increase code generation.
- AI 979 2024-03-26 17:31:44
-
- What is the difference between AI inference and training? do you know?
- If I want to sum up the difference between AI training and reasoning in one sentence, I think "one minute on stage, ten years off stage" is the most appropriate. Xiao Ming has been dating his long-cherished goddess for many years and has quite a lot of experience in the techniques and tips for asking her out, but he is still confused about the mystery. Can accurate predictions be achieved with the help of AI technology? Xiao Ming thought over and over again and summarized the variables that may affect whether the goddess accepts the invitation: whether it is a holiday, the weather is bad, too hot/cold, in a bad mood, sick, he has another appointment, relatives are coming to the house... ..etc. The picture weights and sums these variables. If it is greater than a certain threshold, the goddess must accept the invitation. So, how much weight do these variables have, and what are the thresholds? This is a very complex question and difficult to pass
- AI 786 2024-03-26 14:40:15
-
- TensorFlow deep learning framework model inference pipeline for portrait cutout inference
- Overview In order to enable ModelScope users to quickly and conveniently use various models provided by the platform, a set of fully functional Python libraries are provided, which includes the implementation of ModelScope official models, as well as the necessary tools for using these models for inference, finetune and other tasks. Code related to data pre-processing, post-processing, effect evaluation and other functions, while also providing a simple and easy-to-use API and rich usage examples. By calling the library, users can complete tasks such as model reasoning, training, and evaluation by writing just a few lines of code. They can also quickly perform secondary development on this basis to realize their own innovative ideas. The algorithm model currently provided by the library is:
- AI 682 2024-03-26 13:00:39
-
- CLIP-BEVFormer: Explicitly supervise the BEVFormer structure to improve long-tail detection performance
- Written above & the author’s personal understanding: At present, in the entire autonomous driving system, the perception module plays a vital role. The autonomous vehicle driving on the road can only obtain accurate perception results through the perception module. The downstream regulation and control module in the autonomous driving system makes timely and correct judgments and behavioral decisions. Currently, cars with autonomous driving functions are usually equipped with a variety of data information sensors including surround-view camera sensors, lidar sensors, and millimeter-wave radar sensors to collect information in different modalities to achieve accurate perception tasks. The BEV perception algorithm based on pure vision is favored by the industry because of its low hardware cost and easy deployment, and its output results can be easily applied to various downstream tasks.
- AI 525 2024-03-26 12:41:28
-
- Computer vision is changing the retail industry
- Retail business owners often face inventory management issues that hinder the development of long-term customer relationships. Leveraging the application of computer vision in retail inventory management is an innovative solution that helps establish a robust operating model to achieve business goals. The retail industry is highly sensitive to customer needs and therefore requires continuous investment to improve the consumer experience. The application of computer vision in retail has helped enhance inventory management, a vital aspect of the retail industry. This technology provides an ideal solution as it involves multiple interdependent processes that ultimately impact the delivery of the product. Even small errors in this chain of processes can pose a threat to customer satisfaction and corporate reputation. The benefits of computer vision in retail inventory management
- AI 314 2024-03-26 12:31:19
-
- How to empower industrial applications with machine learning?
- Equipment failures pose serious problems to the industrial sector, leading to production losses and unplanned downtime. This situation represents a serious challenge to process manufacturers worldwide, causing losses that can run into billions of dollars annually. For example, if a key production equipment suddenly fails, it may cause the entire production line to shut down for several hours, thus affecting the operation of the entire supply chain. Fortunately, modern machine learning (ML) offers a breakthrough solution. By analyzing large amounts of sensor data, ML algorithms can predict failures and backlogs before they occur, enabling proactive repairs and drastically reducing downtime. But that’s not all, ML also reveals hidden patterns in production data, optimizing processes, reducing waste and improving overall efficiency. in group
- AI 781 2024-03-26 12:16:02
-
- The model will evolve after merging, and directly win SOTA! Transformer author's new entrepreneurial achievements are popular
- Use the ready-made models on Huggingface to "save up" - can you directly combine them to create new powerful models? ! The large Japanese model company sakana.ai was very creative (it was the company founded by one of the "8 Transformers") and came up with such a clever way to evolve and merge models. Not only does the method automatically generate new base models, but its performance is anything but: they achieved state-of-the-art results on relevant benchmarks using a large model of Japanese mathematics with 7 billion parameters, surpassing the 70 billion parameter Llama- 2 and other previous models. Most importantly, deriving such a model does not require any gradient training and therefore requires significantly less computing resources. NVIDIA scientist JimFan praised it after reading it
- AI 332 2024-03-26 11:30:14
-
- Efficient LLM tuning on local GPU using GaLore
- Training large language models (LLMs) is a computationally intensive task, even those with "only" 7 billion parameters. This level of training requires resources beyond the capabilities of most individual enthusiasts. To bridge this gap, parameter-efficient methods such as low-rank adaptation (LoRA) have emerged, allowing fine-tuning of a large number of models on consumer-grade GPUs. GaLore is an innovative approach that uses optimized parameter training to reduce VRAM requirements rather than simply reducing the number of parameters. This means that GaLore is a new model training strategy that allows the model to fully utilize all parameters for learning and save memory more efficiently than LoRA. GaLore maps these gradients into a low-dimensional space,
- AI 825 2024-03-26 08:26:35
-
- Cambridge team's open source: empowering multi-modal large model RAG applications, the first pre-trained universal multi-modal post-interactive knowledge retriever
- Paper link: https://arxiv.org/abs/2402.08327DEMO link: https://u60544-b8d4-53eaa55d.westx.seetacloud.com:8443/Project homepage link: https://preflmr.github.io/paper Title: PreFLMR: ScalingUpFine-GrainedLate-InteractionMulti-modalRetrievers Background Although multi-modal large models (such as GPT4-Vision, Gemini, etc.) demonstrate powerful general image and text understanding capabilities,
- AI 327 2024-03-25 20:50:47
-
- AI model training: reinforcement algorithm and evolutionary algorithm
- Reinforcement learning algorithm (RL) and evolutionary algorithm (EA) are two unique algorithms in the field of machine learning. Although they both belong to the category of machine learning, there are obvious differences in the methods and concepts of problem solving. Reinforcement learning algorithm: Reinforcement learning is a machine learning method whose core lies in the agent interacting with the environment and learning optimal behavioral strategies through trial and error to maximize cumulative rewards. The key to reinforcement learning is that the agent constantly tries various behaviors and adjusts its strategy based on reward signals. By interacting with the environment, the agent gradually optimizes its decision-making process to achieve the established goal. This method imitates the way humans learn, improving performance through continuous trial and error and adjustments, allowing the agent to perform complex tasks, including the main components of reinforcement learning.
- AI 566 2024-03-25 19:21:18
-
- CVPR 2024 | Zero-sample 6D object pose estimation framework SAM-6D, one step closer to embodied intelligence
- Object pose estimation plays a key role in many practical applications, such as in areas such as embodied intelligence, robotic manipulation, and augmented reality. In this field, the first task to receive attention is instance-level 6D pose estimation, which requires annotated data about the target object for model training, making the deep model object-specific and unable to be transferred to new objects. Later, the research focus gradually turned to category-level 6D pose estimation, which is used to process unseen objects, but requires that the object belongs to a known category of interest. Zero-sample 6D pose estimation is a more generalized task setting. Given a CAD model of any object, it aims to detect the target object in the scene and estimate its 6D pose. Despite its significance, this zero-sample task setting
- AI 459 2024-03-25 18:56:18