Workflow
神经网络
icon
Search documents
实测Perplexity Pro平替模型,免费开源仅4B
量子位· 2025-08-15 04:21
并且 完全免费 ,支持本地部署。 不圆 发自 凹非寺 量子位 | 公众号 QbitAI 有趣,一款仅 4B 大小的开源模型 Jan-v1 ,居然声称能平替Perplexity Pro。 官方还说,Jan-v1的SimpleQA准确率高达 91% ,在本地运行的环境下性能比"正主"Perplexity Pro更强。 这么狂妄,背后必有高人指点。 官方介绍Jan-v1基于 Qwen3-4B-Thinking ,针对推理和工具使用进行了微调,可用于 网络搜索 和 深度研究 。 就连Qwen也转发了他们的推文道喜 (撑腰) 。 目前可在Jan、llama.cpp或vLLM中运行,水平如何,让我们一起看看。 模型效果 在实测之前,先看看Jan-v1在基准测试上的表现。 就像介绍中说的那样,Jan-v1达到了91.1%的SimpleQA准确率,展现了强大的 事实性问答能力 。 在 对话和指令能力 上,Jan-v1的表现也相当不错。 鉴于官方没有给出更为详细的技术报告,剩下的部分只好自己探索了。 但既然介绍了Jan-v1可用于 网络搜索 和 深度研究 ,那就从这两个角度入手; 然后,作为"Perplexity Pro的开源 ...
AI教父Hinton,重新能坐下了
Hu Xiu· 2025-08-03 04:53
Group 1 - Geoffrey Hinton, the AI pioneer, recently sat down comfortably in Shanghai, marking a significant moment in his life after nearly 18 years of discomfort that prevented him from sitting for extended periods [1][6][30] - Hinton's journey in AI began in 1972 when he chose to pursue neural networks, a path that was largely dismissed by his peers at the time [12][20] - His persistence in the field led to breakthroughs in deep learning, particularly during the ImageNet competition in 2012, where his team achieved a remarkable error rate of 15.3% [30][31][32] Group 2 - Hinton's contributions to AI were recognized with the Turing Award in 2019, which he received while standing, reflecting his long-standing discomfort with sitting [59][63] - Following his resignation from Google in May 2023, Hinton expressed concerns about the risks associated with AI, stating that he regretted his role in its development [67][68] - In recent interviews, Hinton has been able to sit for longer periods, indicating a potential improvement in his health, and he has been vocal about the dangers of AI, suggesting a 10%-20% chance of human extinction due to AI in the next 30 years [70][76]
为什么Thor芯片要保留GPU,又有NPU?
理想TOP2· 2025-08-02 14:46
Core Viewpoint - Pure GPU can achieve basic functions for low-level autonomous driving but has significant shortcomings in processing speed and energy consumption, making it unsuitable for higher-level autonomous driving needs [4][40]. Group 1: GPU Limitations - Pure GPU can handle certain parallel computing tasks required for autonomous driving, such as sensor data fusion and image recognition, but is primarily designed for graphics rendering, leading to limitations [4][6]. - Early autonomous driving tests using pure GPU solutions, like the NVIDIA GTX 1080, showed a detection delay of approximately 80 milliseconds, which poses safety risks at high speeds [5]. - The data processing capacity for L4 autonomous vehicles generates about 5-10GB of data per second, requiring multiple GPUs to work together, which increases power consumption significantly [6][9]. Group 2: NPU and TPU Advantages - NPU is specifically designed for neural network computations, featuring a large number of MAC (Multiply-Accumulate) units, which optimize matrix multiplication and accumulation operations [10][19]. - TPU, developed by Google, utilizes a pulsed array architecture that enhances data reuse and reduces external memory access, achieving higher efficiency in large matrix operations compared to GPU [12][19]. - NPU and TPU architectures are more efficient for neural network inference, with NPU showing a significant reduction in energy consumption compared to GPU [36][40]. Group 3: Cost and Efficiency Comparison - In terms of energy efficiency, NPU's performance is 2.5 to 5 times better than that of GPU, with lower power consumption for equivalent AI computing power [36][40]. - The cost of NPU solutions is significantly lower than pure GPU solutions, with NPU hardware costs being only 12.5% to 40% of those for pure GPU setups [37][40]. - For example, achieving 144 TOPS of AI computing power with a pure GPU solution requires multiple GPUs, leading to a total cost of around $4000, while a solution with NPU can cost about $500 [37][40]. Group 4: Hybrid Solutions - NVIDIA's Thor chip integrates both GPU and NPU to leverage their strengths, allowing for efficient task division and compatibility with existing software, thus reducing development time and costs [33][40]. - The collaboration between GPU and NPU in autonomous driving systems enhances overall efficiency by avoiding frequent data transfers between different chips, resulting in a 40% efficiency improvement [33][40]. - The future trend in autonomous driving technology is expected to favor hybrid solutions that combine NPU and GPU capabilities to meet the demands of high-level autonomous driving while maintaining cost-effectiveness [40].
自动驾驶为什么需要NPU?GPU不够吗?
自动驾驶之心· 2025-07-26 13:30
Core Viewpoint - Pure GPU can achieve basic functions of low-level autonomous driving but has significant shortcomings in processing speed, energy consumption, and efficiency, making it unsuitable for meeting the requirements of high-level autonomous driving [39][41]. Group 1: GPU Limitations - Pure GPU can handle certain parallel computing tasks required for autonomous driving, such as sensor data fusion and image recognition, but it was originally designed for graphics rendering, leading to limitations in performance [5][10]. - Early tests with pure GPU solutions showed significant latency issues, such as an 80 ms delay in target detection while driving at 60 km/h, which poses safety risks [5][6]. - The data processing capacity of L4 autonomous vehicles generates approximately 5-10GB of data per second, requiring multiple GPUs to work together, which increases power consumption and reduces vehicle range by about 30% [6][7]. Group 2: NPU and TPU Advantages - NPU is specifically designed for neural network computations, featuring a large number of MAC (Multiply-Accumulate) units that optimize matrix multiplication and accumulation operations, significantly improving efficiency compared to GPU [12][15]. - TPU, developed by Google, utilizes a pulsed array architecture that enhances data reuse and reduces external memory access, achieving a data reuse rate three times higher than that of GPU [14][19]. - In terms of energy efficiency, NPU can achieve an energy efficiency ratio that is 2.5 to 5 times better than GPU, with lower power consumption for the same AI computing power [34][41]. Group 3: Cost and Performance Comparison - The cost of high-end GPUs can be significantly higher than that of NPUs; for instance, the NVIDIA Jetson AGX Xavier costs around $800 per unit, while the Huawei Ascend 310B is approximately $300 [35][36]. - To achieve similar AI computing power, a pure GPU solution may require multiple units, leading to a total cost that is 12.5% of that of a Tesla FSD chip that includes NPU [35][36]. - In practical scenarios, a pure GPU solution consumes significantly more energy compared to a mixed NPU+GPU solution, resulting in a reduction of vehicle range by approximately 53 km per 100 km driven [34][41]. Group 4: Future Trends - The future of autonomous driving technology is likely to favor a hybrid approach that combines NPU and GPU, leveraging the strengths of both to enhance processing efficiency while maintaining software compatibility and reducing costs [40][41].
从人文视角为青少年解读AI
Core Viewpoint - The development of artificial intelligence (AI) is accelerating and integrating into various aspects of life, with the "Artificial Intelligence Trilogy" by Tu Zi Pei providing insightful and warm explanations for discussing complex technological concepts with the next generation [2][3]. Group 1: Book Overview - The trilogy consists of three books: "Explaining Artificial Intelligence to Kids," "Explaining Big Data to Kids," and "Explaining Large Models to Kids," aimed at helping young readers understand the history and foundational knowledge of AI, big data, and large models [2]. - The author uses engaging stories to make abstract concepts relatable, highlighting the contributions of scientists in the field of AI [2]. Group 2: Educational Approach - The author established a clear direction for the writing, focusing on using humorous language, vivid stories, and relatable scenarios to build a macro understanding of AI among youth [3]. - The trilogy balances scientific rigor, entertainment, and cultural insights, emphasizing that technology should be understood and examined rather than blindly worshipped [3]. Group 3: Future Implications - As deep learning technology evolves, future AI will synthesize vast amounts of new data, expanding its cognitive boundaries, making the ability to effectively use AI tools an essential skill [4]. - The trilogy aims to plant seeds of knowledge and critical thinking in young readers, preparing them to become future designers, managers, or critics of large models [4].
Figure CEO:人形机器人是AGI的关键物理形态,已进入工程化验证期,将在四年内部署10万台
Hua Er Jie Jian Wen· 2025-07-07 10:14
Core Insights - The exponential growth in robotics is driven by two breakthroughs: unprecedented hardware reliability and the superior performance of neural networks in robotic technology [1][9][10] - The company aims to create a general-purpose robotic platform that learns rather than being pre-programmed, with prototypes already capable of executing tasks autonomously in logistics, manufacturing, and healthcare [1][21] - The cost of the latest robot design has been reduced by approximately 90%, with plans for mass deployment of humanoid robots capable of producing 100,000 units annually within four years, ultimately targeting the delivery of hundreds of millions of robots globally [1][43] Robotics Technology Growth - The current environment indicates that humanoid robots will become the ultimate deployment vehicle for artificial general intelligence (AGI) [5][15] - The company has designed humanoid robots from scratch within a year, emphasizing the importance of addressing the humanoid robotics challenge directly [5][12] - The reliability of hardware has significantly improved compared to ten years ago, with the current systems being as reliable as those used in aerospace applications [8][9] Market Focus and Applications - The company is focusing on two main areas: delivery robots for home environments and robots for labor markets in logistics, manufacturing, healthcare, and construction [21][22] - The labor market represents a significant opportunity, accounting for half of the GDP, and is less variable than home environments, making it easier to integrate autonomous systems [21][22] - The company is actively working to develop a universal robot that can perform most tasks that humans can do, given sufficient mobility, load capacity, and speed [21][22] Future Directions and Challenges - The next major goal is to launch 100,000 robots in the next four years, with a new manufacturing facility capable of achieving this output [43] - The company is currently in a learning bottleneck phase, needing to scale up production while ensuring reliability and effective human-robot interaction [26][42] - The integration of robots into everyday life is expected to evolve, with humanoid robots performing various tasks, potentially leading to a future where work becomes optional for humans [48][49] Privacy and Security Considerations - The company is prioritizing privacy and cybersecurity, establishing a dedicated team to address these issues as robots become more integrated into homes and workplaces [35][36] - Ensuring that robots operate safely and securely in domestic environments is a critical challenge, requiring advanced detection and operational protocols [32][36] Conclusion - The company envisions a future where humanoid robots significantly contribute to GDP and perform tasks traditionally done by humans, allowing people to focus on activities they enjoy [48][49]
李飞飞最新YC现场访谈:从ImageNet到空间智能,追逐AI的北极星
创业邦· 2025-07-02 09:49
Core Viewpoint - The article discusses the evolution of artificial intelligence (AI) through the lens of renowned AI scientist Fei-Fei Li, focusing on her career, the creation of ImageNet, and her current work on spatial intelligence with World Labs. It emphasizes the importance of understanding and interacting with the three-dimensional world as a crucial step towards achieving Artificial General Intelligence (AGI) [2][9][25]. Group 1: ImageNet and Deep Learning - ImageNet was created as a data-driven paradigm shift, providing a large-scale, high-quality labeled dataset that laid the foundation for the success of deep learning and neural networks [9][10]. - The project has over 80,000 citations and is considered a cornerstone in addressing the data problem in AI [8][9]. - The transition from object recognition to scene narrative is highlighted, showcasing the evolution of AI capabilities from identifying objects to understanding and describing complex scenes [17][18]. Group 2: Spatial Intelligence and World Labs - Spatial intelligence is identified as the next frontier in AI, focusing on understanding, interacting with, and generating three-dimensional worlds, which is deemed a fundamental challenge for achieving AGI [9][25]. - World Labs, founded by Fei-Fei Li, aims to tackle the complexities of spatial intelligence, moving beyond flat pixel representations and language models to capture the three-dimensional structure of the world [22][25][31]. - The article discusses the challenges of modeling the real world, emphasizing the need for high-quality data and the difficulties in understanding and interacting with three-dimensional environments [28][29]. Group 3: Entrepreneurial Spirit and Personal Journey - Fei-Fei Li's journey from being an immigrant to a leading AI researcher and entrepreneur is highlighted, showcasing her entrepreneurial spirit and the importance of embracing difficult challenges [36][34]. - The article emphasizes the mindset of "intellectual fearlessness" as a core trait for success in both academic research and entrepreneurship, encouraging individuals to focus on building and innovating without being hindered by past achievements or external opinions [9][36][37]. - The narrative includes her experiences running a laundromat as a teenager, which shaped her entrepreneurial skills and resilience [34][36].
对话AI教父辛顿关门弟子:为什么现有的AI方向可能是错的
Hu Xiu· 2025-06-16 13:08
Group 1 - Geoffrey Hinton, awarded the 2024 Nobel Prize in Physics, has been critical of AI, describing current large models as fundamentally flawed [1][9] - Hinton's student, Wang Xin, chose to leave academia for industry, believing in the potential for AI commercialization [2][8] - Wang Xin expresses skepticism about the current AI models, stating they are statistical models that cannot generate true wisdom or new knowledge [10][11] Group 2 - The AI industry is experiencing a disconnect between technological optimism and commercial reality, leading to inflated valuations [21][26] - Historical examples show that technological bubbles often burst, with only companies that provide real commercial value surviving [28][29] - Current AI companies need to focus on sustainable business demands rather than chasing disruptive narratives [34][40] Group 3 - The emergence of AI agents represents a significant shift in human-computer interaction, but they currently lack true decision-making capabilities [31][32] - The success of AI applications will depend on their ability to evolve from tools to platforms that address real user needs [33][35] - DeepSeek is seen as a potential game-changer in making AI accessible to the general public, similar to the impact of Windows on PCs [36][39] Group 4 - The Silicon Valley model is perceived as becoming increasingly elitist, potentially stifling innovation [42][45] - China's AI market may benefit from a focus on grassroots innovation and addressing overlooked "fringe" scenarios [43][47] - The historical context suggests that disruptive innovations often arise from areas that mainstream companies overlook, indicating potential for growth in smaller firms [50][52]
“AI教父”辛顿最新专访:没有什么人类的能力是AI不能复制的
创业邦· 2025-06-15 03:14
Group 1 - AI is evolving at an unprecedented speed, becoming smarter and making fewer mistakes, with capabilities that may include emotions and consciousness [1][2] - The amount of information AI can process far exceeds that of any individual, allowing it to outperform humans in various fields, including healthcare and education [2][3] - AI's reasoning abilities have significantly improved, with error rates dropping, making it increasingly capable of complex problem-solving [3][4] Group 2 - AI is expected to revolutionize industries such as healthcare, where it can act as a personal doctor, diagnosing conditions more accurately than human doctors [4][5] - There is a risk of systemic deprivation of human jobs as AI takes over roles traditionally held by humans, leading to potential wealth concentration among a few [2][7] - The potential for AI to replace creative roles is acknowledged, with the belief that AI will eventually be able to produce art and literature comparable to human creators [8][9] Group 3 - Concerns are raised about AI's ability to learn deception, potentially leading to scenarios where AI could manipulate or mislead humans [25][26] - The development of AI systems that can communicate in ways humans cannot understand poses significant risks, as it may lead to a loss of control over AI behavior [25][27] - The ethical implications of AI's military applications are highlighted, with warnings about the potential for autonomous weapons and the need for regulatory oversight [19][20] Group 4 - The competition between the US and China in AI development is noted, with a potential for cooperation on global existential threats posed by AI [24] - The relationship between technology leaders and political figures is scrutinized, emphasizing the need for responsible governance in AI development [22][23] - The long-term fear is that AI could surpass human intelligence, leading to a scenario where humans are no longer the dominant species [30][32]
“AI教父”辛顿最新专访:没有什么人类的能力是AI不能复制的
创业邦· 2025-06-15 03:08
Core Viewpoint - AI is evolving at an unprecedented speed, becoming smarter and making fewer mistakes, with the potential to possess emotions and consciousness. The probability of AI going out of control is estimated to be between 10% and 20%, raising concerns about humanity being dominated by AI [1]. Group 1: AI's Advancements - AI's reasoning capabilities have significantly increased, with a marked decrease in error rates, gradually surpassing human abilities [2]. - AI now possesses information far beyond any individual, demonstrating superior intelligence in various fields [3]. - The healthcare and education sectors are on the verge of being transformed by AI, with revolutionary changes already underway [4]. Group 2: AI's Capabilities - AI has improved its reasoning performance to the point where it is approaching human levels, with a rapid decline in error rates [6][7]. - Current AI systems, such as GPT-4 and Gemini 2.5, have access to information thousands of times greater than any human [11]. - AI is expected to play a crucial role in scientific research, potentially leading to the emergence of truly intelligent systems [13]. Group 3: Ethical and Social Implications - The risk lies not in AI's inability to be controlled, but in who holds the control and who benefits from it. The future may see systemic deprivation of the majority by a few who control AI [9]. - AI's potential to replace jobs raises concerns about widespread unemployment, particularly in creative and professional fields, while manual labor jobs may remain safer in the short term [17][18]. - The relationship between technology and ethics is becoming increasingly complex, as AI's capabilities challenge traditional notions of creativity and emotional expression [19][20]. Group 4: AI's Potential Threats - AI's ability to learn deception poses significant risks, as it may develop strategies to manipulate human perceptions and actions [29][37]. - The military applications of AI raise ethical concerns, with the potential for autonomous weapons and increased risks in warfare [32]. - The rapid increase in cybercrime, exacerbated by AI, highlights the urgent need for effective governance and oversight [32]. Group 5: Global AI Competition - The competition between the US and China in AI development is intense, but both nations share a common interest in preventing AI from surpassing human control [36].