The real purpose of Google’s acquisition of DeepMind is actually like this.

Just entering the new year, the Alpha dog, which became famous last year, made another big news. On the major Go game platforms, the upgraded version of the Alpha Dog anonymously picked up dozens of top players from China, Japan and Korea, and none of the 60 battles were defeated. At the moment when the identity was revealed, the world of chess and even the entire technology circle was lost.

The Alpha dog, the correct name should actually be AlphaGO, a Go artificial intelligence program developed by DeepMind, a British artificial intelligence research laboratory that was acquired by Google (microblogging) several years ago.

DeepMind has now become a global sign in the global AI space, and this brand effect will help Google attract the top AI talent, even in this competition.

DeepMind is currently based in Google's new office park near London's King's Cross. From the outside, we can't see that there are even the world's most technologically influential companies that Google and Facebook are vying to buy.

Of course, we all know the final result. Google successfully acquired DeepMind in January 2014 for £400 million (approximately $660 million). But why did Google buy the British AI company? You know, Google has already been in the industry leader in machine learning and AI, and what value will it bring to Google for Deepmind?

Before October 2015, Google’s huge advertising revenue was enough to allow the “ambition project” such as DeepMind to generate revenue without any worries. Subsequently, Google made a total adjustment to the company's structure, founded the Alphabet "Umbrella Company" (UmbrellaCompany), and separated its search, YouTube, and other network subsidiaries from the R&D investment department. The most important change in this structure is that the balance sheets of its various business units began to be calculated separately, and they could no longer rely on Google’s cash cow to sit back and relax.

However, Google's acquisition of DeepMind is not only a matter of economics, but the true value of DeepMind is far more profound than we think.

The most direct benefit of DeepMind for Google is that the company has given Google and Alphabet a strategic advantage in the battle for AI with other technology companies around the world. At the same time, this brand effect helped Google attract the top AI talent and let competitors including Facebook, Microsoft and Amazon lose their talent advantage. To date, DeepMind's London office has recruited 400 computer scientists and neuroscientists, and the team is expected to expand to around 1,000.

Since its acquisition, DeepMind has been featured on the cover of Nature Weekly, a well-known academic journal. Today, replicas of the two cover images are attached to the wall of DeepMind's lobby. In March 2016, "Alpha Dog" defeated World Go Champion Li Shishi and became the headline news of a large number of media.

In addition to talent absorption and public attention, the strategic significance of DeepMind is not limited to this. DeepMind CEO and co-founder Demis Hassabis described DeepMind as a new type of research organization that combines the long-term perspective of academics with the energy and concentration of technology start-ups.

In 2010, Hasseles and Mustafa Suleyman and Shane Legg founded the company. Among them, Legg and Hasabis are neuroscience researchers at University College London, while Suleiman is a small child of Hasabis.

Hasabis believes that the long-term goal of DeepMind is to "solve the problem of solving intelligence." This requires the company to continuously develop versatile and even general-purpose artificial intelligence products that are as broad and efficient as humans.

In this regard, DeepMind is gaining some obvious advantages by Google's acquisition. First, DeepMind can call Google's powerful computing resources. Second, Google’s financial resources are sufficient to support DeepMind’s high R&D spending. Obviously, if a company with a financially simple company acquires DeepMind, the former is likely to demand that it be profitable as soon as possible, but the addition of Google's Hassels can focus more on research. Moreover, because the DeepMind office location and Google's headquarters in Mountain View maintain a "safe distance", this is enough to give him more practical control over DeepMind.

If DeepMind can successfully create a generic AI product, this will have a significant positive impact on Alphabet. For now, DeepMind's research agenda shows that it is not a company that operates in a business model because its development cycle is quite long. Hasabis said that his roadmap for DeepMind planning spans 20 years and hopes that DeepMind will invent new AI algorithms inspired by the way human brains work.

Of course, even if DeepMind has not developed an artificial intelligence product that is equivalent to or exceeds the human level, the learning software that the company has developed will also contribute to the advancement of other Alphabet business. For example, in July of this year, Google announced that learning software developed by DeepMind has found ways to save power for cooling Google's data centers, saving about 40%. The learning software compresses the data center's operational logs and optimizes the process in analog form.

According to the news, DeepMind has begun to use its own AI research and development results to solve practical problems for the company. Suleiman revealed that some of DeepMind's R&D projects can be used for energy infrastructure management, medical systems and clean water improvements, and DeepMind is also beginning to benefit.

In November of this year, DeepMind signed a five-year contract with the medical institution Royal Free London, whose main task is to handle the medical records of 1.7 million patients. Earlier this year, DeepMind also gained access to two databases in other London hospitals. DeepMind used AI software to analyze about 1 million retinal scan reports to successfully identify early signs of degenerative eye disease, or to allow AI software to distinguish between healthy and cancerous tissues through head and neck cancer images.

Skilled programmers and powerful computing power are important to the development of the AI ​​business, but obtaining real-world data is also essential. For example, when we want to use AI and machine learning technology to improve different systems such as hospitals, power grids, and factories, we need to use a lot of actual operational data.

Fortunately, Alphabet has massive data for its AI products to mine. However, in some specific areas of exploration, DeepMind may need to find ways to dig deeper. Take the project that it recently participated in to study lip language, the key to the success of the project is that the research team has a huge database.

According to the news, the research team of this project is mainly from Oxford University, led by computer vision research expert Andrew Zisserman. The BBC provided hundreds of thousands of hours of news and video information to the research team, and without the data, the research team could not allow the AI ​​system to start training in lip language.

However, Hasabis does not seem to want to overemphasize the importance of data acquisition for the future of DeepMind. He simply said, "It is enough to let human engineers create simulation models to solve problems, and then deploy AI learning tools in these models."

However, this is not the way most machine learning systems work today. Take today's world-famous "Alpha Dog", which first learns millions of chess strategies from the 160,000 human-playing Go game, and then continues to improve its self-game. Therefore, if DeepMind needs to collect a large amount of personal information, then the company needs to consider how to deal with consumer concerns about corporate access to personal data.

If DeepMind can solve these problems, then this subsidiary will become one of Alphabet's most important assets, namely an algorithm factory and an AI talent pool. Moreover, the data ownership rights handled by DeepMind still belong to the original owner, but the software developed through data learning belongs to Alphabet, and this method will basically solve the privacy problem that is of great concern to the outside world.

Although DeepMind may not be able to generate large revenues in the future by using AI programs to solve complex problems, the useful information obtained by DeepMind AI software by analyzing the data is enough for Google to invest the huge amount of money for the original bid. .

HDI PCB

Electronic design is also trying to reduce its size while continuously improving the performance of the whole machine. From mobile phones to smart small portable products, small is always the same pursuit. High-density integration (HDI) technology enables end-product design to be more compact while meeting higher standards of electronic performance and efficiency. HDI is currently widely used in mobile phones, digital (camera) cameras, MP3, MP4, notebook computers, automotive electronics and other digital products, among which mobile phones are the most widely used. HDI boards are generally manufactured using a build-up method. The more times the layers are stacked, the higher the technical grade of the board. Ordinary HDI boards are basically one-time laminate, high-order HDI uses two or more layers of technology, and adopts advanced PCB technology such as stacking holes, plating holes, and laser direct drilling. High-end HDI boards are mainly used in 3G mobile phones, advanced digital video cameras, IC carrier boards, etc.

Pcb Hdi,Hdi Board,Hdi Pcb Design,Hdi Pcb Technology

Chuangying Electronics Co.,Ltd , https://www.cwpcb.com