On July 10, at the "Google Solve with AI" event, In his speech, Jeff Dean, a senior researcher at Google and head of Google AI, pointed out that machine learning is the best way to explain AI. Machine learning can automatically collect information through the accumulation of data, build a powerful database, and find a pattern that can use mathematical operations for reasoning. "But how to use the trained functions is another proposition."
Jeff Dean mentioned that machine learning has already performed very well in some fields, watching, listening, speaking, and understanding , At these four points, all can penetrateThe function of machine learning is more convenient for our lives. "As a kind of computer science, machine learning believes that through programming, computers can become smart through learning, and it may be easier than programming them to be smart."
Currently, data security and user privacy are A widely discussed issue. In his speech, Jeff Dean emphasized that the principles of Google AI are currently based on a set of ethics, "in our research, applications and services, to guide our development and use of artificial intelligence." Based on this set of principles, Jeff Dean revealed that Over 100 projects were reviewed and thousands of Google employees were trained on the fairness of machine learning.
Jeff Dean mentioned a form of machine learning Federated Learning (Federated Learning), which is Google AI's privacy, securityResearch on integrity and fairness can make applications and services work better for everyone without the need to collect raw data from devices.
In addition, in exploring how artificial intelligence can better benefit society, Google launched the "AI for Social Good" project in 2018. Jeff Dean mentioned that this project has two core concerns: research and engineering; and helping others solve important problems.
In terms of scientific research and engineering, Google is exploring how its research on machine learning will have a positive impact on society, humanitarian causes, and environmental issues. ?
In terms of helping others solve important problems, Google provides related tools, such as TensorFlow; and training, such as a crash course on machine learning, which has 15,000 in Asia this yearPeople sign up to help people solve problems.
"We believe that artificial intelligence can help solve the most difficult social and environmental problems of our time-such as medical care, disaster prediction, environmental protection, agriculture or cultural protection." Jeff Dean said.
In an interview with the media, Jeff Dean answered the patent dispute some time ago. It is reported that Google’s Dropout patent has been authorized. The main content of the patent relates to the method of training neural networks. The industry is worried that this patent will become a lethal weapon in the AI field in the future.
In this regard, Jeff Dean said that applying for a patent is to avoid unnecessary trouble, "If someone or a company wants to use our ideas or concepts, Google will communicate, but we will not treat them as Make a weapon or profit from it."