The number of posts on each topic tells the story. Over the past ten years, it has doubled from 162,444 in 2010 to 334,497 in 2021. The most popular AI categories reported by scientists and others recognized design, technology, and algorithms.

Additionally, there will be 30 times more patent applications related to AI innovation in 2021 than in 2015. Although most patents filed in 2021 will be in China, most patents are licensed in the United States. United. The number of users participating in the GitHub open source AI software library has also increased from 2015 to 2021. This library contains data written by computer code used in commercial applications and products. Tensor Flow is the most popular, based on OpenCV (with meta AI).

A computer that measures images and understands speech

In particular, among the many ways AI can be realized, last year's community research focused on the application of AI to computer vision, a sub-line that teaches machines to understand images and videos to improve image distribution, product knowledge, and art. I do it. Position and movement of human joints, face detection (with or without a face).

The most popular file for training AI models for image sharing is called ImageNet. Some researchers first share their additional data sample before posting it on ImageNet. However, the model makes the mistake of misidentifying 1 of the 10 intermediate images. The best performing models come from the Google Brain team. In addition to identifying images and faces, AI can create fake images that are almost indistinguishable from reality, and to prevent this, scientists have worked on extensive research on file-based files like Face Forensics++.

Natural language processing, a small field studied since the 1950s, is gradually increasing in English comprehension, communication, sense of purpose, sense of meaning, speaking and writing skills, and translation . For easy reading comprehension, AI may be more efficient than humans, but when speech works hard, like when the meaning of a word needs to be translated, humans still have the courage. Artificial intelligence experts, on the other hand, worry that inaccuracies will affect large-scale models drawn from a mixed bag of information.

Big Tech and AI

Tech companies like Amazon, Netflix, Spotify, and YouTube are improving the AI   used in their deals. One is the fact of AI's role in promoting learning. This allows them to react and perform well in virtual games such as Chess and Go. Support learning can also teach weak vehicle functions, such as line changes, or data patterns can be used to predict future events.

As AI seems to do better what we want it to do, tuition fees have been reduced by 60% since 2018. Meanwhile, the system used for 6 minutes to train in 2018 now only takes 13 seconds. If you include the cost of hardware, image distribution training will cost less than $5 in 2021, but will cost more than $1,000 in 2017.

Lots of jobs and students

More AI applications in industry means more needs for AI training and operations. In the United States by 2021, California, Texas, New York, and Virginia will have the highest demand for AI operations. Artificial intelligence and learning technologies have been at the heart of the concerns of doctoral students in computer science for the past ten years. Private investment in AI will reach historic highs in 2021, totaling $93.5 billion (double the 2020 figure). According to the report, AI companies in "data management, processing and cloud" are making the most money in 2021, trailing the company's focus on "healthcare and medicine" and financial technology (fintech for short).

How does the government use it and control it?

In fiscal year 2021, U.S. agencies spent $1.53 billion on non-defense AI research and development, or 2.7 times the 2018 budget. Ideally, the Department of Defense has invested in 500 AI research and development by 2021, about 6% more than it spent last year. The second use of AI is in prototyping technology and programs against weapons of mass destruction.

Finally, the report examines international, federal, and state laws affecting AI (looking for key terms such as intelligence, technology, vehicle drivers, or algorithmic bias). The report, which looks at 25 countries around the world, found that 55 AI laws were combined between 2016 and 2021. Last year, Spain, the UK and the US all joined. our AI policy.