Welcome to read the Open Source Daily produced by the OSCHINA editorial department, which is updated every day.
# 2024.10.25
Today's News
Arc browser is "useless", the developer announced to stop development
Last week, Arc browser developer The Browser Company released a video saying that it is developing a new browser, but it is not a successor to Arc. However, it is more "simple and easy to use" than Arc and aims to attract a wider user group.
Josh Miller, CEO of The Browser Company, said in the video that although Arc has developed rapidly, its complexity and uniqueness, as well as its "over-pursuit of innovation", have resulted in high learning costs, making it difficult to reach a wider market and unlikely to become a mainstream product. Although the company will not abandon the Arc browser, it will mainly provide stability updates and fix bugs in the future, and shift its development efforts to the new browser.
The core goal of the new browser is to become a "more proactive, more powerful, and more AI-based" platform - it can be called an "Internet computer" or "iPhone of web browsers", which can automatically handle the tedious operations of users between different web pages and applications. For example, copying and pasting data between office platforms, looking up order numbers on e-commerce platforms and pasting them into customer service emails, etc.
It is said that the new browser will be released early next year, adopting a more traditional horizontal tab design to reduce the cost of user switching. The focus of the initial version is to provide a smooth basic user experience, and then gradually introduce more innovative features.
Apple will release several new Mac products equipped with M4 processors this week
Apple will release new M4 Mac series products this week, including MacBook Pro, iMac and Mac mini, as follows:
- 24-inch iMac
- Two versions of the Mac mini
- New 16-inch MacBook Pro
- New 14-inch MacBook Pro, including entry-level and high-end versions
Among them, the iMac will be equipped with an entry-level M4 chip, the Mac mini will have basic and M4 Pro versions to choose from, and the MacBook Pro will be equipped with a high-end M4 processor.
In addition to the chip upgrade, the new MacBook Pro also upgrades the starting memory to 16GB and adds an additional Thunderbolt port, increasing the number of ports from two to three.
Apple releases AI cloud architecture white paper and some code
Apple recently officially released a cloud AI model called Private Cloud Compute (PCC) and invited all security and privacy researchers to access and study it.
Apple has now made available a PCC Security Guide that includes comprehensive technical details about the components of PCC and how they work together to provide breakthrough privacy protections for AI processing in the cloud. Apple says this will help other researchers better understand PCC.
PCC technology can meet the computationally intensive requests of Apple Intelligence while providing breakthrough privacy and security protections, and is one of the foundations for Apple Intelligence to be realized. Apple claims that Private Cloud Comput is the most advanced and secure AI computing architecture in the cloud.
Google is developing an AI agent for Chrome, codenamed "Project Jarvis"
According to an exclusive report from The Information , Google is developing an AI agent optimized specifically for the Chrome browser, code-named "Project Jarvis", which will perform tasks for users, including "collecting research information, purchasing products or booking flights."
Project Jarvis will be powered by Gemini 2.0 and will only work on web browsers, the report said. The tool will help people "automate everyday web tasks" by capturing screenshots and interpreting them, then clicking buttons or entering text.
In its current state, Project Jarvis will take a few seconds to think when performing different operations. It is said that Google may announce "Project Jarvis" in December this year, but the specific date is still uncertain.
Today's Observation
Social Observation
The gap between algorithm theory and practice
If we believe that algorithm theory is purely a mathematical theory (game), this debate can be stopped and it is not impossible to develop on our own. However, if we admit that algorithms are ultimately for solving problems and that theory still has to explain reality or guide practice, then we will think about whether algorithm theory over the years has a large trajectory dependency component (after all, Big O existed before the invention of electronic computers), which has led to a lack of awareness of reality and a growing disconnect.
The main research objects of algorithm theory and practice are different. Theoretical considerations of extremely simple algorithms only involve definition issues, while actual calculations on computers are programs and examples. Is there a more specific algorithm theory that can explain why many industrial examples are easy, and also provide some algorithm ideas that can really be useful in practice?
- Weibo Cai Shaowei
Open source speech recognition model running locally: Moonshine
Moonshine, an open-source speech recognition model that runs locally without an Internet connection, has received 1.4k stars on GitHub in less than a week. It is so popular among users because of its three features: smaller, faster, and more accurate.
- Smaller: Moonshine can be deployed on local devices, even wearable devices, and can run offline.
- Faster: Moonshine is 5 times faster than Whisper when processing audio less than 10 seconds.
- More accurate: On the OpenASR dataset, Moonshine has a lower average word error rate than Whisper.
- Weibo Quantum Bit
Microsoft open-sources pure vision-based GUI agent: OmniParser
Microsoft has open-sourced a purely visual GUI agent: OmniParser. It can accurately identify interactive icons on the interface and understand the semantics of each element in the screenshot to achieve more automated interface interaction scenarios, such as automated testing and automated operations. Mainly through its powerful screen parsing technology, it converts the user interface into structured elements, thereby improving the visual recognition capabilities of multimodal models such as GPT-4V.
GitHub：https://github.com/microsoft/OmniParser
- WeiboGitHubDaily
The birth of the "open source concept" can be traced back to the 1970s
The birth of the "open source concept" can be traced back to the 1970s. MIT received a donated printer, and a programmer named Richard Stallman in the school found that due to the confidentiality agreement, he could not access and modify the source code of the printer, resulting in the inability to add even a simple function such as "alarm after paper jam". As a result, Stallman deeply realized the importance of software freedom. He decided to create a new operating system, an operating system where all users can freely view, modify and share its source code. In this way, the GNU project was born and became the cornerstone of the open source movement. And Linux can have the influence it has today, but without the buff of open source, it would just be an excellent personal project.
Dialogue with Academician Mei Hong: The current large language model technology path cannot lead to AGI
Key points:
1. The current large language model technology path cannot lead to AGI, and may reach the ceiling in a few years.
2. AI cannot replace humans. It lacks creativity and can only be said to be an auxiliary tool.
3. From the perspective of science and technology ethics, intelligence is a unique feature of humans, especially cognitive ability. Why should we create something to replace our cognition?
4. Machines do not have so-called consciousness or emergence. In essence, they organize and organize existing content based on probability statistics. In this sense, the large language model can be regarded as a "knowledge base" compressed from existing "corpus".
Media Watch
Accelerate the forging of industrial software competitiveness
Industrial software drives the optimization of manufacturing management processes, changes in production models, and the improvement of total factor productivity, and is an important tool for accelerating the construction of a modern industrial system. my country is the world's largest industrial country and has built a relatively complete industrial chain and supply chain system, but there are still prominent shortcomings and weaknesses in the field of industrial software. The support of industrial software for the digitalization, networking, and intelligent transformation of the manufacturing industry needs to be strengthened.
-Economic Daily
A one-time reward of 1 million yuan! Wuhan releases the country's first open source system construction plan
Wuhan will implement open source technology foundation projects, open source project crowd innovation projects, open source enterprise cultivation projects, open source platform source introduction projects, open source data opening projects, and open source talent gathering projects. Wuhan has set a goal to strive to break through more than 10 open source software technologies, incubate more than 300 open source projects, gather more than 200 open source innovation companies, build more than 10 high-level open source communities and open source public service platforms, build a digital infrastructure benchmark city, gather 1 million open source developer ecosystems, and initially form an open source system with leading technology, emerging projects, concentrated enterprises, active communities, and abundant talents by 2027.
-Greater Wuhan
Liu Qingfeng said that AI applications have entered the deep waters of scenario-based applications
Looking to the future, Liu Qingfeng believes that there are five key words for the development of the artificial intelligence industry: standing tall and firm, autonomous and controllable, combining general and specialized, end-to-end cloud linkage, and software and hardware integration. Liu Qingfeng said: "Adhere to these five key points, our artificial intelligence will have a great future, it will truly liberate productivity, unleash imagination, change the industrial structure and competition landscape, bring about changes in the scientific research paradigm, and bring changes to all walks of life."
-Times Weekly
“Pure-blooded” Hongmeng explodes! Fund managers’ latest analysis
The fund managers interviewed believe that Huawei's launch of its self-developed native Hongmeng operating system has raised investors' expectations for the future development of the Hongmeng ecosystem, thereby driving the strength of related concept stocks. At the same time, Huawei's hardware sales performance such as mobile phones, cars and tablets have been strong, laying a solid foundation for the widespread adoption of the Hongmeng operating system. However, due to the obvious short-term gains, there is also pressure for adjustment in the future market. Investors should pay attention to the risks and not blindly follow the trend.
-China Fund News
Zhao Bin of Agora: Generative AI drives four major changes in the IT industry. AI+RTE becomes an important theme for the next decade
As generative AI rapidly iterates and evolves, more companies are beginning to worry about how to better commercialize AI. However, the $600 billion problem is the most pressing issue as there is currently a huge gap between the huge investment in AI and the actual output.
-China Business News
Investors flee big models
At present, the Hundred Model War is shifting in a new direction, and the players who previously competed in technology must face the difficulties of landing commercialization. Recently, there have been rumors in the AI circle that due to the cooling of financing, valuations are constantly rising. Large model companies are beginning to struggle, and some of the six tigers of large models are beginning to abandon large model research and development and turn to applications. Romantic technology encounters the reality of survival problems, but in fact, according to data, many AI companies have declared bankruptcy both at home and abroad. There are also some with strong R&D capabilities, but they eventually fell due to the difficulty of no follow-up funds. After all, large models are an extremely money-burning track.
-Rongzhong Finance
Today's Recommendation
Open Source Projects
flameshot-org/flameshot
https://github.com/flameshot-org/flameshot
Flameshot is a powerful yet easy-to-use screenshot software for Windows, Linux, and macOS.
Daily Blog
FlinkSQL temporary join development
In real-time development, temporary join is often used when two streams are joined to obtain the attributes of the target at the corresponding time. In the real-time iteration of traffic upgrade, the author needs to accurately match the traffic log with the corresponding product attributes at the browsing time. There are many pitfalls in the development process of using temporary join. Some experience is accumulated in this article for reference and communication.
Open Source Voice
User Viewpoint
JetBrains officially announces two heavyweight IDEs are free for non-commercial use
- Viewpoint 1: WebStorm consumes too much memory. I have a genuine key and gave it up, and used VSCode again.
- Point 2: How funny! Is Vim any better?
- Viewpoint 3: Yes, and there are many
- Viewpoint 4: If Rust IDE is not free, the potential market and users may be snatched away by VS Code in the future, which will allow VS Code to expand its market share again and further occupy Java IDE in the future. At this time, JetBrain will be completely finished. In addition, why choose to open Rider and WebStorm? This is consistent with what I said above. Donet is the main battlefield of Microsoft VSCode and VS, and the mainstream of front-end development is also VSCode. Therefore, it must grab the territory from you and expand its influence so that the market share will not be pulled away by VSCode. Of course, PyCharm Golang CLion will never be free, for the reasons mentioned above.
- Viewpoint 5: Isn’t it because many people use VSCODE to write front-end software?
- Viewpoint 6: I also use vscode for backend, c++, and rust
- Point 7: GoLand doesn’t even have a community version
- Viewpoint 8: It is difficult to achieve pure non-commercial use. It is estimated that mandatory data collection can easily determine whether it is commercial use, which lays the groundwork for future profiteering.
- Viewpoint 9: The reason seems to be that the opponent is too strong
- Opinion 10: It would be nice if CLion was free
- Viewpoint 11: Ideas are all-inclusive
- Point 12: Isn’t vim the best?
- Viewpoint 13: It’s good. If students don’t make trouble, I will continue as usual.
- Point 14: Even with my low-spec computer, vscode is stuck and can’t run.
- Viewpoint 15: Can I get a refund if I paid?
- Point 16: Better than VS
The "phenomenal" open source game that once swept the world - "2048" has been released for 10 years, and the original author has just released a special commemorative update
- Viewpoint 1: A 19-year-old genius programmer announced that he would never make money from the phenomenal games he developed. An 18-year-old me said: If I sleep until noon every day, can I save money on breakfast?
- Viewpoint 2: Even if you don’t open source, this simple game can be made by anyone with a little development experience.
- Viewpoint 3: So what matters is creativity, not experience
- Viewpoint 4: Mobile games were really new at that time
- Viewpoint 5: There is still a need for fusion
- Point 6: It’s ridiculous that game play is not protected by copyright
- Viewpoint 7: I look forward to playing 2048 Special Edition in 2048
- Point 8: Where is my favorite pixel bird? The game that tortured me for three days and couldn't exceed 100
---END---
