news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

A week's work ended with a cry of grief: Microsoft blue screen!

Starting from July 18th local time and July 19th Beijing time, some Microsoft applications and services experienced access delays, incomplete functionality, or were inaccessible.

From ordinary workers to airlines, banks, exchanges, schools, etc., all may be affected.

I left work early because my computer screen was blue. When I was walking on the street, I saw billboards and the big screen in front of the mall were blue. I took a walk to the supermarket to buy groceries and found that the supermarket cash register system was also blue. At this time, a friend sent me a message saying that the airport was blue, the flight was cancelled, and I couldn't come back tomorrow.

This scene is a bit too cyber.

At present, the "culprit" of Microsoft's large-scale blue screen has been found out - CrowdStrike, a US network security service provider, due to an update error. Microsoft also said that the root problem has been solved, but the residual impact has not been completely eliminated.

But humans are clearly still in shock: If Microsoft hadn’t suddenly fallen ill, who would have realized how much humans rely on it?

As the winner of the computer operating system war, Microsoft has penetrated every aspect of people's lives and has become a basic tool for human society. Today, a new business war is underway around the world, the war of artificial intelligence (hereinafter referred to as AI). AI will be the next "infrastructure", which has almost become a consensus at this stage.

Just a few hours before the Microsoft blue screen incident, OpenAI released a small model, GPT-4o mini, and the price dropped significantly. Although there was no press conference and no fanfare, the news still excited the entire technology community. A small model that is strong, small, and cheap enough means that application developers are blessed. From another perspective, OpenAI's move will make its model more influential in the application layer.

Coincidentally, on the same day, Mistral AI and NVIDIA also released a small model, Mistral NeMo. The battle of models in Silicon Valley has evolved from big to small and is becoming more and more grounded.

Forbes gave this example: “An emergency heart or brain surgery scheduled based on an AI-driven system. If the underlying technology fails, such as a machine not rebooting properly, due to a patch update from CrowdStrike or another company in the software supply chain, the consequences could be fatal.”

This may be one of the positive implications of the Microsoft blue screen incident. Humans who are eager to enter the AI ​​era need to realize the dangers of "dependence".OpenAI, which is working tirelessly to create the next generation of infrastructure, needs to prevent blue screen attacks even more.

A

How much chaos did Microsoft's blue screen incident cause?

Elon Musk, the world's richest man, famous entrepreneur and well-known Internet celebrity, called it "the biggest mistake in IT history" on the social platform X (formerly Twitter).

The workers were quite happy, just thinking about how to finish work on Friday and start a happy weekend, when the computer suddenly broke down, a blue screen appeared and it kept restarting. What can I do? I have to leave work early, but I am helpless.

Oh, by the way, don’t forget to take a photo before leaving get off work. Happy International Blue Screen Day to everyone!

Of course, there are also many workers and students who are not happy. After all, sometimes at a critical moment, a blue screen does not mean taking a rest early, but it means you have the intention of reincarnating early. Especially when the file is not saved...

Others were unhappy because almost all the computers in the company were blue, except for their own. Thank you Microsoft, you have confirmed your status as the "chosen one".

The small blue screen in the office is nothing. What is more shocking is that after you walk out of the office, you will find that the big screen in front of the shopping mall, the large and small screens at the airport, the supermarket cash register system, and even the huge electronic billboards on the street are all "blue", and the sad expression in the corner hits the heart directly.

In fact, the Microsoft blue screen incident caused a series of "disasters".

Airports in many places including the United States, Germany, and Australia are almost paralyzed. In the United States alone, more than 2,000 flights are grounded (Delta Airlines and United Airlines have grounded all flights), and more than 5,300 flights are delayed.A large number of passengers were stranded, and the airport was a "blue ocean" from the small screen at the check-in counter to the huge guide screen.

The London Stock Exchange was unable to publish news during the Asian trading hours. The Australian Broadcasting Corporation encountered a failure and the British Sky News was unable to broadcast live. Banks were also affected, including banks in South Africa, Israel, New Zealand and other countries. Users of at least two banks in the UK were unable to log in to their accounts.

What is most worrying is that some emergency hotlines and medical institutions have also been affected. 911 call centers in many places in the United States are not working properly; most general practitioner clinics in the UK are unable to operate due to problems with patient appointments and medical record systems, and the Israeli Ministry of Health said that more than a dozen hospitals have been affected. Fortunately, most hospitals around the world still have more or less manual, paper-based work as a foundation, so work can still be carried out and emergency departments can operate as usual.

Of course, many businesses will also be affected.

Tesla factories in the United States were also affected. Some workers at Tesla's super factories in Austin, Texas and Nevada left work early because equipment in some factories began to report errors.

This made Tesla CEO Musk very angry, and he even chased Microsoft CEO Satya Nadella on X (formerly Twitter) to express his dissatisfaction, saying that "this has brought a serious impact on the automotive supply chain."

After discovering that the cybersecurity company CrowdStrike was the culprit behind the incident, Musk said he had deleted CrowdStrike components from all systems. He then added: “Unfortunately, many of our suppliers and logistics companies are using it.”

Musk’s imitation account thoughtfully posted an AI-generated picture, in which Musk stood in front of a burning CrowdStrike server, expressing the anger of the person involved in the incident.

B

The "July 19 Microsoft Blue Screen Incident" is destined to go down in history, and CrowdStrike has also become a household name from being well-known in the industry.

The cause of the disaster was so minor it seemed almost absurd: CrowdStrike sent an update to Microsoft.

More specifically, CrowdStrike has a platform called Falcon, which is the company's flagship product. Before the incident, CrowdStrike released a sensor configuration update for the Windows system, which is part of the Falcon platform's protection mechanism. Such updates are not whim, but come several times a day.

However, this configuration update triggered a logic error, causing system crashes and blue screens on affected systems.

Shortly after the incident, CrowdStrike CEO George Kurtz sent several X messages, indicating that they were doing their best to fix the problem, and reassured everyone to stay calm, and repeatedly emphasized that the cause of the incident was not a cyber attack. At 9 a.m. Beijing time on July 20, Kurtz released a technical statement on the incident. During the period, he also appeared in a media interview to explain to the public the cause and response to the incident, looking a little haggard.

However, a sincere attitude can hardly make up for huge mistakes. As of the close of U.S. stocks on July 19, CrowdStrike's stock price closed down 11%.

You know, CrowdStrike has long been famous in the industry. The company was founded in 2011 and provides online security solutions, focusing on providing a cloud-based endpoint protection platform. Some of the largest cloud service providers, including Microsoft and Amazon AWS, are its customers, and its customers also include major global banks, healthcare and energy companies.

CrowdStrike's main platform, Falcon, is also quite interesting. It uses AI and machine learning technology to detect, prevent and respond to cyber threats.In other words, this is an industry leader in AI-driven security solutions.

This incident is bound to deal a blow to CrowdSrike's path and highlights the important issue of the limitations and risks of AI in cybersecurity. While AI promises to enhance detection capabilities and speed up response, we need more favorable supervision and verification processes to mitigate the risk of unexpected failures.

Can we really rely on AI to manage our most sensitive security needs, or are we placing too much trust in an imperfect tool?

C

What’s interesting is that a cybersecurity company that uses AI as an important selling point made a big mistake, while on the other hand, a star company in the AI ​​field is launching a new model.

In the early morning of July 19th, Beijing time, just a few hours before the Microsoft blue screen incident, OpenAI quietly launched the GPT-4o mini model. The release was very low-key, but the news spread immediately because it was a very different model for OpenAI.

GPT-4o mini, as its name suggests, is small in size, powerful in function and attractive in price.

OpenAI did not disclose its specific parameter scale, but called it the company's most economical and cost-effective small model. GPT-4o mini's knowledge was updated to October last year. In terms of the types of languages ​​it supports, it is comparable to GPT-4o, with a context window of 128k and an MMLU score of 82, which is quite competitive.

Most importantly, GPT-4o mini only costs 15 cents per million token inputs and 60 cents per million outputs. This is 96% to 97% cheaper than GPT-4o and 60% to 70% cheaper than GPT-3.5 Turbo. OpenAI CEO Sam Altman compared it to the "best" model in the world in 2022, text davinci-003 (GPT-3), which is far less powerful than GPT-4o mini, but costs 100 times more.

The release of small models and the entry into a price war highlight OpenAI’s determination to expand its applications.

The product manager of OpenAI's API platform even said that GPT-4o mini truly realized OpenAI's mission and allowed people to have more extensive access to AI.

He revealed that OpenAI made this move because they noticed that developers were increasingly eager to use smaller models.

Small models have become the new anchor for Silicon Valley giants and star startups. On the same day, Mitral AI teamed up with NVIDIA to launch a small model, Mistral Nemo, with a parameter size of only 12B, surpassing Meta's Llama-3 8B in performance. The founder of Hugging Face said: "This is the week of small models."

More small models have been released earlier, such as Google's Gemini Flash, Anthropic's Claude Haike, etc. Even Apple's AI, which has been long awaited, did not take the "big cup" route, but used a small model on the device side, focusing on fine-tuning the user experience.

In the blog post about GPT-4o mini on OpenAI’s official website, it says:We envision a future where models will be seamlessly integrated into every app and every website.GPT-4o mini is paving the way for developers to build and scale powerful AI applications more efficiently and cost-effectively. The future of AI is becoming more accessible, reliable, and embedded in our everyday digital experiences, and we’re excited to continue leading the way.”

"Seamlessly integrated into every application" and "embedded into daily digital experience" are what Microsoft's operating system has done. From super-large models with increasing parameters to small models that are flexible, easy to use, and inexpensive, AI manufacturers want to become the next "Microsoft".

However, there is still no definite path to ensure the security of AI models and applications. AI security is discussed every day, but there is no final conclusion. Meanwhile, AI is still moving forward.

Perhaps the Microsoft blue screen incident is a wake-up call for the transition from the old to the new. If AI is really "embedded in daily digital experience" and "seamlessly integrated into every application" as OpenAI and others enjoy, the disaster caused by the "blue screen" is left to your imagination.