51
0今日内容一览：
[] Anthropic进军AI制药领域，计划自主开发新药
[] 特斯拉限制员工AI支出，每周额度上限设为200美元
[] AI先驱杨立昆：大语言模型在机器人领域毫无希望
[] 马斯克降低Optimus量产预期：这跟造车完全不同
---
[] Hello and welcome to the Daily Tech Briefing. I'm your host. Today, we have four exciting stories. First, Anthropic enters the drug discovery space with a new platform. Second, Tesla sets a weekly budget limit on employee AI usage. Third, AI pioneer Yann LeCun explains why current language models are not smart. And finally, Elon Musk warns that early Optimus robot production will be very slow. Let's get started.
大家好，欢迎收听科技每日简报。我是主持人。今天我们为您带来四个精彩的科技故事。首先，Anthropic推出了一个新平台，宣布进军AI制药领域；第二，特斯拉限制员工使用外部AI工具的每周支出；第三，AI先驱杨立昆解释了为什么当前的大语言模型并不聪明；最后，埃隆·马斯克警告称，Optimus人形机器人的早期量产速度将非常缓慢。让我们开始吧。
---
[] Anthropic Aims to Develop Its Own Drugs Using AI. We start with a major move in medical technology. Anthropic has announced Claude Science. This is a new AI platform designed to help scientists do research. But Anthropic wants to go even further. The company plans to develop its own drugs to treat neglected (被忽视的，被遗忘的) diseases. AI is already changing drug discovery (发现，研发). Large companies use AI to find new molecules. However, experts say that AI-designed drugs are still far from reaching patients. AI cannot replace real-world testing. AI can speed up the search, but new drugs must still prove themselves through slow, methodical (有条不紊的，有条理的，系统性的) experiments in the real world. This process requires a lot of money and time.
Anthropic进军AI制药领域，计划自主开发新药。我们首先来看医疗技术领域的一项重大举措。Anthropic发布了Claude Science。这是一个旨在帮助科学家进行科学研究的新型AI平台。但Anthropic的目标还不止于此。该公司计划自行开发药物，以治疗被忽视的疾病。AI目前已经改变了药物研发的进程。大型制药公司正在利用AI来寻找新的分子。然而，专家表示，AI设计的药物距离真正造福患者还有很长的路要走。AI无法取代现实世界中的测试。AI可以加快筛选速度，但新药仍然必须通过现实世界中缓慢而系统性的实验来证明自己。这个过程需要耗费大量的资金和时间。（来源：The Verge Tech & Culture）
---
[] Tesla Caps Employee AI Spending to Control Rising Token Costs. Next, Tesla is placing a limit on how much money its employees can spend on external (外部的) AI tools. A new internal rule limits software engineers to 200 dollars per week for AI tokens. Before this cap, some engineers spent thousands of dollars weekly. Tesla recently launched its own internal platform called Bottle Rocket. This platform lets employees use models from OpenAI, Anthropic, and xAI. CEO Elon Musk has urged employees to use their own tools. However, many engineers still prefer (更喜欢，更倾向于) to use Anthropic's Claude. Meanwhile, SpaceX is planning to acquire Anysphere, the company behind the popular AI coding tool Cursor.
特斯拉限制员工AI支出，每周额度上限设为200美元。接下来，特斯拉限制了其员工在外部AI工具上的支出金额。一项新的内部规定限制软件工程师每周最多只能花费200美元来购买AI代币。在此限制之前，一些工程师每周的花费高达数千美元。特斯拉最近推出了名为Bottle Rocket的内部平台。该平台允许员工使用来自OpenAI、Anthropic和xAI的模型。首席执行官埃隆·马斯克敦促员工使用自己的工具。然而，许多工程师仍然更倾向于使用Anthropic's Claude。与此同时，SpaceX正计划收购热门AI编程工具Cursor的母公司Anysphere。（来源：The Decoder）
---
[] AI Pioneer (开拓者，先驱) Yann LeCun Claims LLMs Are Hopeless for Robotics. Now, let's talk about the nature of artificial intelligence. Yann LeCun, a pioneer in AI, argues that current models are not truly smart. He left Meta to start AMI Labs. LeCun believes that Large Language Models cannot deal with real-world physical data. He says they are hopeless for robotics. LLMs only predict words based on statistical patterns. They do not understand the physical laws of our world. To solve this, his team is building a new AI architecture. This new system aims to create a flexible (灵活的，可适应的) model that understands physical actions. Other researchers are also working on world models to help robots learn from mental simulations.
AI先驱杨立昆：大语言模型在机器人领域毫无希望。现在，我们来聊聊人工智能的本质。AI领域的先驱杨立昆认为，当前的AI模型并不是真正的聪明。他离开了Meta，创立了AMI Labs。杨立昆相信，大语言模型无法处理现实世界中的物理数据。他说，这些模型对于机器人开发来说是毫无希望的。大语言模型仅仅基于统计规律来预测词语。它们并不理解我们世界的物理法则。为了解决这一问题，他的团队正在构建一种全新的AI架构。这个新系统旨在创建一个能理解物理动作的灵活模型。其他研究人员也正在开发世界模型，以帮助机器人通过内部模拟来学习。（来源：BBC Technology）
---
[] Musk Tempers Expectations for Tesla Optimus Production Speed. Finally, Tesla CEO Elon Musk has lowered expectations for the Optimus humanoid robot. Musk stated that early production of the robot will be extremely slow. He explained that building a humanoid robot is not like building a car because everything is completely new. Tesla is converting space at its Fremont factory to start limited (有限的) production this summer. The robot has about 10,000 unique parts. This complexity (复杂性，复杂程度) makes early output rates impossible to predict. However, Tesla is building a larger factory in Texas for volume production in 2027. Early units will perform simple tasks in Tesla factories before helping consumers.
马斯克降低Optimus量产预期：这跟造车完全不同。最后，特斯拉首席执行官埃隆·马斯克降低了外界对Optimus人形机器人的预期。马斯克表示，该机器人的早期生产速度将会极其缓慢。他解释说，制造人形机器人并不像制造汽车，因为一切都是全新的。特斯拉正在改造其弗里蒙特工厂的部分空间，以便在今年夏天开始有限度量产。这种机器人拥有大约1万个独特的零部件。这一复杂性导致早期的产量几乎无法预测。不过，特斯拉正在德克萨斯州建造一座更大的工厂，以准备在2027年进行大规模量产。在服务普通消费者之前，早期的机器人将被安排在特斯拉工厂内执行简单的任务。（来源：Teslarati）
---
[] 📚 今日单词：
methodical
- /əˈθɒdɪkl/
- adjective 有条不紊的，有条理的，系统性的
[] 词组/例句
- methodical experiments 系统性/有条不紊的实验
- AI can speed up the search, but new drugs must still prove themselves through slow, methodical experiments in the real world.
AI可以加快筛选速度，但新药仍然必须通过现实世界中缓慢而系统性的实验来证明自己。
---
[] That is all for today's briefing. As we discussed Yann LeCun's skepticism about language models and the challenges of humanoid robots, we want to ask you: do you agree that current AI models are hopeless for robotics? Or do you think humanoid robots will become a part of our daily lives sooner than expected? Let us know your thoughts in the comments. Thank you for listening, and see you next time.
以上就是今天简报的全部内容。正如我们今天讨论的，杨立昆对语言模型持怀疑态度，以及人形机器人面临的诸多挑战。我们想问问大家：您是否同意当前的AI模型对于机器人开发来说是毫无希望的？或者您认为人形机器人会比预期更快地融入我们的日常生活？欢迎在评论区分享您的想法。感谢收听，我们下期再见。
