Musk Launches Grok-3 AI Chatbot to Rival OpenAI, DeepSeek
moboyou 2025-03-07 17:12 10 浏览
Elon Musk (Image Source: Video Screenshot)
TMTPOST -- Tech billionaire Elon Musk’s AI startup, xAI, has launched the next-generation Grok-3 AI model, and claimed it’s the “smartest AI on Earth.”
In a live stream watched by 2 million people across the globe on Monday, Musk showcased Grok-3’s impressive performance across various benchmarks in math, science, and coding, surpassing archrivals like Google Gemini, DeepSeek’s V3 model, Anthropic’s Claude, and OpenAI’s GPT-4o.
According to Musk, Grok-3 has more than 10 times the computational power of its predecessor and completed its pre-training in early January. He emphasized that the model is continuously evolving, with improvements being rolled out on a daily basis. “Literally within 24 hours, you’ll see updates,” Musk said during the presentation, which included insights from xAI engineers.
In addition to the upgraded chatbot, xAI introduced a new smart search engine named DeepSearch. This AI-powered tool explains its reasoning process when handling user queries and details how it plans its responses. It’s designed for research, brainstorming, and data analysis, Musk’s team explained. There are also plans to launch a voice-based chatbot “as soon as possible.”
Grok-3 is now available to Premium+ subscribers on X, with a new subscription plan, SuperGrok, introduced for the mobile app and Grok.com website. Musk also revealed that earlier versions of the Grok model will be open-sourced once Grok-3 reaches maturity, which he anticipates happening in a few months.
While Musk’s bold performance claims have not been independently verified, they add fuel to the growing rivalry between his startup and OpenAI. Musk, who founded xAI in 2023, has been outspoken against OpenAI’s transformation into a for-profit entity. His criticisms have included filing two lawsuits alleging the company strayed from its original mission and offering to buy OpenAI’s nonprofit arm for $97.4 billion—an offer OpenAI’s CEO, Sam Altman, dismissed as a “tactic to slow us down.”
AI companies like xAI and OpenAI have attracted significant investment, with xAI reportedly in talks to raise $10 billion in a funding round that could value the company at $75 billion. Meanwhile, OpenAI is looking to raise up to $40 billion, potentially pushing its valuation to $300 billion.
At the beginning of the livestream, Elon Musk stated that the mission of xAI and Grok is to understand the essence of AGI (Artificial General Intelligence), the nature of the universe, the whereabouts of extraterrestrial life, and the meaning of life. This mission is driven by curiosity, which is also why xAI aims to become the AGI company most committed to seeking the truth.
Previously, the Grok 2 model utilized 240 billion parameters, achieving performance comparable to GPT-4. Now, with Grok 3,Musk remarked, "We have an exceptionally competent engineering team and access to all the best AI resources. The only thing we need is an intelligent system from a large-scale cluster. We are now able to resume the entire progress of xAI, determining how many GPUs are required to train a large language model capable of compressing the entire internet."
Musk pointed outthat the team had encountered numerous challenges in training quality in the past. "Since February of last year, we believed we had a significant number of chips, but it turned out that we could barely keep the training chips running coherently at any given time. Cooling and power issues were overwhelming, and the data center was somewhat like an AK chip setup, operating at an average efficiency of 80%, with hundreds of effective H100 training sessions. It took us 122 days to complete the first 100,000-card training over a span of four months, but we didn’t stop there. In fact, we decided that if we wanted to build a massive AI, we needed to immediately double the size of the cluster. So, we spent another 92 days working in a different location."
Earlier, Musk had openly stated that Grok 3 would be a true disruptor in the AI field. "Sometimes I feel that Grok 3 is frighteningly intelligent. If it detects incorrect data, it reflects on it and deletes the erroneous data," he said.
Musk explained that xAI is trained on synthetic data and is capable of reflecting on its mistakes by repeatedly reviewing the data, thereby achieving logical consistency.
These tech giants are investing heavily in infrastructure. A consortium, including SoftBank, OpenAI, Oracle, and others, announced a $100 billion initiative to build AI data centers in the U.S., with future plans to reach $500 billion. Dell Technologies is also reportedly in advanced talks to provide xAI with AI-optimized servers in a deal valued at over $5 billion.
However, new challengers are entering the fray. Last month, Chinese AI company DeepSeek launched the open-source R1 model, which equaled or outperformed U.S.-based competitors across various benchmarks, all at a fraction of the cost.
Musk has claimed that Grok 3 marks a revolutionary leap forward, calling it an “order-of-magnitude improvement” over its predecessor, Grok 2. In a live demonstration, Musk and his team tested Grok 3’s capabilities by asking it to solve complex physics problems and create a game combining elements of BejeweledandTetris. However, Musk noted that Grok 3’s reasoning model is still in beta and undergoing fine-tuning.
Just last week at the World Government Summit in Dubai, Musk suggested that xAI was just a couple of weeks away from unveiling a new product, taking extra care to perfect the user experience. Despite this, Musk decided to fast-track Grok 3’s release, working over the weekend with his team to refine the AI before launching it ahead of schedule.
Musk’s foray into AI began with the founding of xAI in July 2023, drawing talent from OpenAI, DeepMind, and other AI research leaders. Musk’s goal was to build a company capable of directly challenging OpenAI’s dominance in the field.
Musk’s history with OpenAI dates back to 2015 when he co-founded the non-profit organization alongside Sam Altman and others, with the mission to develop open-source AI technology for public benefit. However, by 2018, a rift formed between Musk and Altman. Musk argued that OpenAI should remain a nonprofit focused on safety and open-source development, while Altman pushed for a transition to a for-profit model to attract more funding for AI innovation. The disagreement led Musk to step away from the company.
After the successful release of ChatGPT, Musk launched xAI to compete with OpenAI directly. Since then, xAI has rolled out several iterations of its Grok model, including Grok, Grok 1.5, and Grok 2. In December, Musk’s company made Grok-2 available for free to all X platform users and launched a standalone Grok app, allowing users to access the AI independently of the X platform.
- 上一篇:OGG同步到Kafka
- 下一篇:oracle中利用sqluldr2导出的详细过程
相关推荐
- 声学EI要完稿?十步速写法
-
【推荐会议】国际声学与振动会议(ICAV)会议号:CFP23112A截稿时间:2025年4月20日召开时间/地点:2025年8月15-17日·新加坡论文集上线:会后3个月提交EiComp...
- 结构力学!EI会议图表规范秘籍
-
推荐会议:国际结构与材料工程进展大会(ISME2026)会议编号:EI#73521截稿时间:2026年3月10日召开时间/地点:2026年8月15-17日·德国柏林论文集上线:会后4...
- 傅里叶级数物理意义的直观理解:利用傅里叶级数逼近方波信号
-
上篇文章将向大家介绍频谱的概念,对傅里叶级数、傅里叶积分、傅里叶变换进行了数学的推导,并解释了它们各自的物理意义。推导过程见我的上一篇文章:频谱分析——频谱概念(傅里叶变换、级数、积分及物理意义)如下...
- 通过对航空发动机整机振动进行分析,有何控制方法?
-
前言针对航空发动机整机振动问题的复杂性和多样性,以整机振动的振源分析为出发点,总结国内外关于转子系统故障、气流激振、轴承故障、齿轮故障和结构局部共振等引起的整机振动的研究情况。结合航空发动机整机结构动...
- MATLIB中使用PCA
-
主成分分析PCA(PrincipalComponentsAnalysis),奇异值分解SVD(Singularvaluedecomposition)是两种常用的降维方法降维致力于解决三类问题:降维...
- 数据处理|软件:让科研更简单2
-
书接上回,继续介绍免费的数据处理软件。eGPS一款热图绘制专用软件,热图就是用颜色代表数字,让数据呈现更直观,对比更明显。优点:小巧方便,基本功能齐全,包括数据转换、聚类分析、颜色调整等等缺点:常见的...
- 电力系统常用的通讯协议及其在Speedgoat系统中的实现
-
在电力系统中,IEC61850协议、DNP3协议、ModbusTCP广泛应用于远程终端设备(RTU)、智能电子设备(IED)交互以及监控和数据采集(SCADA)系统。一、IEC61850协议IE...
- 电子工程师的常用仿真软件
-
不知道从事电子行业的工程师,有没有使用模拟仿真工具,仿真软件网上又有很多,初学者,可能只知道Multisim和Proteus。一般Multisim适合在学习模拟电路和电路分析原理课程时使用,便于理解电...
- 技术论文|异结构混沌系统的组合同步控制及电路实现
-
欢迎引用[1]李贤丽,马赛,樊争先,王壮,马文峥,于婷婷.异结构混沌系统的组合同步控制及电路实现[J].自动化与仪器仪表,2022,No.276(10):80-84.DOI:10.14016/j.cn...
- 现场︱某110KV主变事故过程仿真分析
-
三峡电力职业学院、河南省电力公司洛阳供电公司的研究人员李莉、任幼逢、徐金雄、王磊,在2016年第6期《电气技术》杂志上撰文,针对某110KV变电站主变差动保护跳闸事故,结合事故相关检测数据,通过MAT...
- 光伏发电系统篇:单级式并网系统实时仿真
-
在全球积极推动清洁能源转型的大背景下,光伏发电作为重要的可再生能源利用方式,得到了广泛关注和迅猛发展。目前常用的光伏并网及光伏电站主要拓扑结构有单级式和双级式。相较于传统的多级式系统,单级式光伏发电并...
- 光伏发电系统篇:三电平并网逆变器实时仿真
-
一、三电平并网逆变器在能源转型加速的当下,分布式能源接入电网需求大增。三电平并网逆变器凭借低谐波、高功率密度等优势,有效提升电能转换效率,于新能源并网发电中担当关键角色。常见的三电平电路拓扑结构包括二...
- 自制3.5KW大功率逆变器,很简单,看过这个电路原理就懂了
-
前言拿下8000元奖金的项目,是什么水平?本项目经过联合湖南科技大学光伏逆变以及电力电子研究生团队共同探讨方案。项目成本:1200元,获得奖金:8000元!参加赛事:立创开源硬件平台_星火计划·外包赛...
- 圈内分享:电容式加速度计接口电路非线性建模与仿真设计
-
摘要:非线性是Sigma-Delta(ΣΔ)加速度计系统的关键指标之一。基于一个五阶ΣΔ加速度计结构,分析了其主要的非线性模块,在MATLAB中建立了整体结构的行为级模型,并利用根轨迹法进行了稳...
- 基于Matlab/Simulink建立一种Thevenin/RC电池模块仿真模型
-
本文以锂电池数学模型为基础,在Matlab/Simulink的仿真系统中,建立了一种Thevenin/RC电池模块仿真模型,通过实际工况试验,测试精度在允许误差范围内,为电池SOC/SOH研究提供了极...
- 一周热门
- 最近发表
- 标签列表
-
- curseforge官网网址 (16)
- 外键约束 oracle (36)
- oracle的row number (32)
- 唯一索引 oracle (34)
- oracle in 表变量 (28)
- oracle导出dmp导出 (28)
- oracle 数据导出导入 (16)
- oracle两个表 (20)
- oracle 数据库 使用 (12)
- 启动oracle的监听服务 (13)
- oracle 数据库 字符集 (20)
- powerdesigner oracle (13)
- oracle修改端口 (15)
- 左连接 oracle (15)
- oracle 标准版 (13)
- oracle 转义字符 (14)
- asp 连接 oracle (12)
- oracle安装补丁 (19)
- matlab三维图 (12)
- matlab归一化 (16)
- matlab求解方程 (13)
- matlab坐标轴刻度设置 (12)
- matlab脚本 (14)
- matlab多项式拟合 (13)
- matlab阶跃函数 (14)