Analysis: Apple in the face of artificial intelligence can be self-evolution or revolution?

In 2011, Apple co-founder Steve Jobs (Steve Jobs) announced the iPhone launched voice assistant Siri, Apple became the first to Artificial Intelligence (AI) on the number of Millions of consumers in the pocket of the company. However, eight years later, the technology giant is still struggling in the field of artificial intelligence. Analysts say that Apple's ability to create great artificial intelligence products for Apple's development in the next decade is essential, like the iPhone in the past 10 years to help Apple made such a brilliant achievement.

But the tech giant faces a daunting challenge because the essence of artificial intelligence is pushing Apple into its 'comfort zone' in its sleek design hardware and service space. Artificial intelligence programs require data collection and in-depth mining, but this is strictly adhered to Apple's privacy policy and its location does not collect user information.

In addition, Apple's long-standing preference for the company has become less popular in the eyes of potential star employees. These new employees are from the top computer science departments in the United States, who are attracted by companies that are willing to publish their research results. 'Artificial intelligence is not Apple's DNA,' said Steve Munster, a venture capitalist and Apple analyst. 'They understand that every company will become an artificial intelligence company in the future, but they are in the A difficult situation.

In the 2017 annual developer conference, Apple is trying to build 'AI engine' image. It launched a new independent smart speaker HomePod, and vigorously promote the Apple product of artificial intelligence applications. During the two and a half hours of demonstration, Apple executives repeated the 'machine learning' term with about 6,000 developers present.

The so-called 'machine learning' is also a branch of artificial intelligence, it describes a very high-speed, complex computer data analysis and statistical modeling form. Siri can now use machine learning to predict morning commuter time, or browse travel news on Apple's Safari browser, and then suggest suggestions such as booking a hotel room.

In addition, Siri can use the machine to learn to talk to you, and through the price of $ 349 home automation device HomePod to help you organize music. It will automatically organize your photos into albums, such as '2 Anniversary' albums, and you do not need to provide any information about the background of the photo. Apple has even introduced a new software toolkit, CoreML, that can handle the vast amount of data collected in machine learning applications faster. (A executives ridicule, Apple artificial intelligence processor speed is six times that of Google).

But while Apple announced the news, other technology companies have released similar innovations and have invested billions of dollars in the booming artificial intelligence arms race. Many companies bet on artificial intelligence, such as intelligent software that can be used to chat with people like you in the future, or to identify real-world objects in computer vision, which can be used to support full-function unmanned vehicles.

For Apple, it is hard to get ahead of the field, and in this area, the core of Apple's secrecy culture has also been cut. 'This is a major shift for Apple,' said Daniel Gross, an Apple executive who specializes in artificial intelligence research. 'Apple's focus is on creating great products rather than publishing papers.'

10 years after the launch of the iPhone, Apple is looking for another can replace the iPhone's heavy product. IPhone sales to promote Apple to become the world's highest market value of the company, but iPhone sales still account for more than half of Apple's total revenue, 2016 total revenue of 215.6 billion US dollars But in 2016, Apple's iconic smartphone purchases fell for the first time, indicating that the high-end smartphone market was finally saturated.

'With the first generation of iPhone and second-generation iPhone iterations, compared to last year's iPhone and this year's iPhone,' said Benedict Evans, a partner at Andreessen Horowitz, a Silicon Valley venture capital firm, and mobile expert Benedict Evans. The difference is much smaller. 'Thus, the S-curve (the industry term that describes the growth of the innovation index) tends to be flat. Then you need to have a whole new revolutionary technology, but not yet. '

Apple did not respond to reporters' requests for multiple interviews and comments.

Apple plans to launch HomePod later this year, and Amazon and Google launched their automated devices Echo and Google Home a few years ago, sending consumers' voice queries back to California-based servers for analysis. Google, Tesla, Uber and others have been driving autopilot cars on public roads for years, but Apple has just got a test permit two months ago. Siri is still based on web search before this year's developer conference, and occasionally tells a pre-set joke.

Now, Apple is catching up. In October 2016, Apple hired Russ Salakhutdinov, a professor at Carnegie Mellon University, who was known as 'deep learning' or 'unsupervised learning' Of the field of artificial intelligence, which is the complex branch of machine learning. In this process, the computer is trained to replicate the neurons of the brain to identify objects or sounds.

Sarah Hutinov is a disciple of Geoffrey Hinton, the world's top researcher in this field. Sarah Huddinov also works at Carnegie Mellon University and Apple, while Sindon is working for both Google and the University of Toronto. Richard Zemel, a professor of machine learning at the University of Toronto and director of the Victorian Institute of Computer Research at the University of Toronto, says networking with superstars in academia not only helps improve the product, but also becomes a key recruitment tool The 'In the past, you will not see anyone wearing an apple badge at the meeting, and now you will see it.'

Last December, Apple presented and published the first artificial intelligence academic paper at an industry conference. Sarah Hutdinov said in an interview that another paper has been accepted by the famous computer vision conference and will be released in July this year. He also said he was not authorized to discuss any details of his work at Apple. Zemor said in an interview with Bloomberg two years ago that Apple was doing a good job in secrecy, and that such secrecy measures prevented them from benefiting from the significant progress in the field, and now Apple is 'making some changes' The But he added: 'Apple will need to do more work.'

Elite University researchers said in an interview, Apple is still not the first choice for computer science graduates, Google, Facebook and Amazon is still far ahead, but Apple's ranking is rising. Experts say Apple's entry into the AI market is also slower than peers because it does not want to dig deeper into data mining operations like rivals such as Google and Facebook. And when you have to collect and analyze the data, Apple has spent a lot of resources to build more privacy protection.

Unlike Google and Facebook, Apple thinks it should limit the collection of user information. Google and Facebook are primarily advertising agencies that need to collect large amounts of data to match users. At the previous developer conference, Apple executives boasted that they did not build a user profile. Apple CEO Tim Cook (Tim Cook) to Apple as a anti-Google company.

But the researchers said that if you are building artificial intelligence, then the stand against data collection has become a big problem. For example, home appliances must collect and analyze people's conversations to improve the ability of devices to talk to humans. To make Siri smarter, it needs to collect and analyze data from other applications, such as your calendar, restaurant appointments, and now browsing history.

Last year, as Apple began to embrace artificial intelligence on the iPhone, the company launched a large privacy program. The project uses an academic concept called 'Differentiated Privacy' and applies it to artificial intelligence applications on the iPhone. 'Differentiated privacy' works by inserting noise or bad information into good data to confuse outsiders who want to learn about personal information.

For example, in order for Apple to combine your dog's photos into separate albums, it needs to collect many photos of the dog. Apple collects the images, but they encrypt the data and mix the data with other data, so that if someone tries to restore the original data set, they do not know what is bound to a single user. Compared with other privacy methods, such as the use of mathematical formulas to present the user's anonymous state, this technology is considered more powerful.

Gross said Apple's focus on privacy may allow companies to slow down the pace of developing certain products, but it will allow consumers to get the trust. 'Apple is working hard on hard science and technology to protect your privacy, and I think Google and Facebook will have to launch similar products to show their commitment to protecting privacy,' he said. '(Source: AI Research Institute)

Internet Research Papers