Then he specifically mentioned the machine learning software TensorFlow that Google developed and used internally. In 2015, Google opened the source code, and since then the software has been downloaded more than 15 million. Pichai said: 'We make TensorFlow as much as possible for everyone. Use AI. '
In the big tech companies that are fiercely competing in the development of AI technology, this open preaching has become a standard behavior. Facebook, Amazon and Microsoft, like Google, have released their own engineers' open source software for machine learning. All of these companies, including Apple, known to some extent for its confidentiality, encourages their AI researchers to announce their latest ideas and help companies recruit the best teachers and graduates from universities.
At the same time, these companies that advocate AI openness are also claiming ownership of AI technology and applications. In recent years, applications for AI-related patents, especially machine learning patents, have grown rapidly. So far, technology companies have not used these patents to sue opponents or threaten litigation. But if AI Patents become the arms of enterprises, and the current opening of AI research and ideas may end, which may hinder AI research.
According to a study published this month by the US Bureau of Economic Research, applications submitted by the United States related to machine learning (currently the technology that drives AI development) are growing rapidly. One of the authors of the report, Stanford University researcher Michael Weber said: We see a major outbreak in patent application activity on AI and machine learning, and this geometric growth will continue.'
In 2010, only 145 patents in the United States involved machine learning, but it grew to 594 by 2016, and this number is not all, because the USPTO only publishes applications 18 months after the registration of patents. The number of patent applications has increased from 94 in 2010 to 485 in 2016.
Google itself has this trend. The USPTO database shows that in 2010 Google only applied for a patent involving machine learning or neural networks, but in 2016 Google and other subsidiaries of Alphabet submitted 99 patents. Facebook in 2016 Submitted 55 patents related to machine learning or neural networks, 0 in 2010. IBM received 1,400 AI-related patents in 2017.
It is not surprising that AI patent applications have grown substantially. In 2012, neural networks suddenly became a research hotspot for technology companies, because this technology can greatly improve voice and picture recognition capabilities. But the action of locking technology emphasizes openness when these companies openly discuss AI strategies. In sharp contrast. The growth of AI patent applications is reminiscent of the fierce battle around smartphone patents.
NBER documents show that Apple and Samsung have at least 50 lawsuits on smartphone technology and design, and Apple and Google have about 20 lawsuits. Former Independent Researcher SRI's General Counsel, Stanford University Lecturer Richard Abra Mson said that the more patent applications filed in a particular field, the greater the likelihood of litigation. 'If you give everyone a gun, you can be sure that the probability of a shooting incident will rise.'
The AI lawsuit may damage the open process that technology giants want to promote. Abramson said that 25 years ago, patent litigation mainly occurred between companies using certain technologies in products. Now many lawsuits come from 'patent rogues' companies, they hold Patents are not for use but for claims. But there is no indication that leading AIs are prepared to use their AI patents to sue.
A spokesperson for Google and DeepMind said that their patents were defensive, not for fighting with other companies. Google spokesperson also mentioned that patent applications filed by companies accounted for only a small percentage of recent applications. Facebook spokesperson said that Patent applications should not be construed as involving current or future plans. IBM Chief Patent Adviser Manny Schechter said that the company's patents only reflect investments in basic research.
But these statements leave room for future policy changes. Given the recent history of technology patent wars, some researchers are still concerned that AI patent accumulation may be used to hinder progress. The value and scope of evaluating patents is complex, even if experts explain Cautious. But Oxford University's AI researcher Miles Brendagg said that some of the applications submitted by Google and other companies seem to describe the basic technology with a wide range of applications.
Georgia Institute of Technology professor Mark ⋅ Riddle said that patents on algorithms and machine learning for other basic technologies made him feel bad. Although these patents have not caused problems, the legal ownership of relatively abstract concepts is not suitable for machine learning. Such an exciting development process.
Not all applications submitted for AI ideas and techniques can be approved. Since the US Supreme Court ruled in 2014 that it was not enough to implement a patent on a computer to become a patent, software patents have become more difficult to obtain. Last year USPTO significantly increased AI patents. The number of reviewers is expected to phase out more applications. But there seems to be no major change in which AI ideas can be patented. Patent attorney Chaucer Holovachak said: 'The company that submitted a large number of patents in this field is A large part of economic life. '
This means they can lobby and hire lawyers, and allow parliamentarians and courts to support their practices, which seems to make all AI technologies widely recognized as patents. What makes tech companies happy is that USPTO Director Andrei Iancu has said He is thinking about the AI patent issue. In April of this year, he told the Senate Judiciary Committee that he believes that the recent court ruling has made chaos of what algorithms can become patents.
Ianku believes that algorithms including AI algorithms are usually available, he said: 'We must ensure that our policies, including intellectual property, are very focused on stimulating this innovation.'