That's quite like a tech interview instead of a f2f talk.
@readword-cn8 күн бұрын
喷特斯拉纯视觉有问题大家怎么看
@paulpaul77776 күн бұрын
期待在大陆比拼一下
@huiwan29596 күн бұрын
质疑,批评,但不能说是喷,马斯克是神吗?他过往说得话没错过?
@大宅门-t1z6 күн бұрын
没有喷吧,理想的激光雷达的方向主要是摆设,主要方向是特斯拉的纯视觉
@allen3744Күн бұрын
亦步亦趋特斯拉😂
@codedoc9558 күн бұрын
他的口癖是我觉得,我也有这个现象,不知道要怎么避免
@hanchisun61648 күн бұрын
说明口癖不影响做千亿企业,不如把时间精力花在更重要的事情上
@karlmin84717 күн бұрын
总比口癖是额、啊、这个、那个的好吧。
@lc.sin.15 күн бұрын
This podcast episode features a conversation between two individuals, "Xiaojun" and "Guangmi", discussing the global large language model (LLM) landscape, with a particular focus on the shift from the technology itself to the products and applications built on top of it. The conversation moves into predictions about AI trends in 2025 and beyond. Key Insights and Takeaways: The "Next Google" Race: The main theme of the podcast is the race to become "the next Google," with ChatGPT being seen as just one step along the way rather than the final goal. Many companies, such as Anthropic, XAI, Perplexity, and even domestic Chinese LLMs like Doubao, Kimi, and coding-focused platforms like Cursor and Devin, are all following different paths but ultimately converging on the same goal of becoming the dominant platform. Guangmi emphasizes that the competition isn't just to surpass ChatGPT, but to fundamentally change information access, distribution, and organization. Evolution of Information Architecture: The discussion highlights the evolution of how information is organized, beginning with the early days of Yahoo-style portals with manual curated lists, then progressing to Google’s keyword search indexing, and later to recommender systems, and now, to AI-driven task and agent focused systems. The podcast delves into the history of computer science to explain the current state of LLMs. It highlights the shift from the curated lists of the early internet to Google's keyword indexing, to the current AI era where the focus is on "action" and agent-based applications. They see a clear evolution from websites to content to AI tokens as the basic building block of the internet. The Rise of the "Agent": The podcast introduces the idea of AI agents and task-based output. They see a future where AI is not just a chatbot, but a personal assistant, completing complex tasks for users. This shift from static content to dynamic actions is considered a major change in AI development. It's suggested that this means the small unit of interaction is no longer a web page, but tokens and tasks, requiring the re-organization of intelligence. The Importance of "Context": The conversation underscores the critical role of context in AI applications. The podcast emphasizes that AI requires access to user data, information, and past behaviors to function effectively, and that companies that can best leverage context will ultimately succeed. The idea of context as the new currency of the internet is introduced. Just as logistics and payments were crucial for e-commerce, context is the most important foundation for next generation AI applications. Business Models and Challenges: The discussion questions whether ChatGPT’s business model of subscriptions alone will be sustainable, highlighting the difference in monetization potential between tools, content platforms, and task oriented platforms. The podcast analyzes the different business models and highlights that current chat interfaces have limited monetization power compared to platforms like TikTok, Taobao, WeChat. They question whether ChatGPT or similar platforms should start showing ads or whether they will evolve into something else. Full Stack vs. Building on Others: The podcast explores the debate between building full-stack solutions (like Google's approach), and building applications on top of other companies' LLMs. Building a full stack vertical integration like Apple or Tesla may give more control, but the podcast also sees benefit in companies like Anthropic building operating system-like infrastructure. The Importance of Coding: The podcast identifies coding as a key battleground, with companies that perform well at coding related tasks having the potential to win big, as evidenced by the popularity of tools like Cursor and Devin. Predictions for the Near Future: In the next few years, they predict the rise of more practical, agent-based AI, capable of completing complicated tasks. They mention that AI agents could even take on the jobs of a manager in certain areas. They predict the growth of companies specializing in specific areas of AI such as coding, or research based AI. The podcast also emphasizes the importance of building high-quality data sets for training AI, particularly in relation to tasks like coding. The importance of data efficiency is emphasized. The Importance of Human-AI Interaction The presenters point out that the current chat interfaces limit the interaction and intelligence of the AI, and that there will be a need for better human-computer interfaces that do not limit context in the future. Open Source vs. Closed Models The podcast highlights the growing power of the open-source model of AI development as being key to democratization of AI, while also citing challenges in some core functions like Google. The Role of Large Tech Companies: They suggest that some tech giants might end up as major players due to the high costs associated with developing AI technology from scratch. They believe Google, Amazon, and Apple still have a chance to make a big impact in this space. They point out the importance of infrastructure (AWS), the OS platform (Android) and the business model for these companies to win. Overall Themes: Beyond the Chatbot: The focus has shifted from the impressive conversational abilities of LLMs to how these capabilities can be integrated into practical applications and products. The Value of Action and Context: The podcast stresses that future AI advancements will be centered around agents capable of performing complex actions and having a full and rich understanding of their context. The Long Game: The podcast acknowledges that despite the fast pace of change, major transformations will likely take time and that new business models must emerge to support long term growth of the AI sector. Report Generation Notes: The summary above tries to capture the essence of the conversation, without direct translation, but preserving the key concepts and themes discussed. The original podcast uses many industry specific terms, and the summary uses the most precise English equivalent when possible. The presenters make many references to other companies, models, and events. The references are not included here for brevity, but are available in the audio if required. The summary emphasizes predictions about the future of AI from the presenters' perspective.