Tomcat
Professional
- Messages
- 2,695
- Reaction score
- 1,072
- Points
- 113
According to the report, this year the world has really made headway in the application and study of artificial intelligence. AI thrives in various fields - education, technical invention and scientific research.
AI IN EDUCATION, SCIENCE AND INVESTMENT
According to the report:
HOW IS THE POWER OF THE AI?
All of this is impressive, but it's important to know that no matter how quickly AI thrives, its progress cannot be compared to what pop culture and media headlines credit it with. In fact, artificial intelligence is still very limited in some important aspects.
This is demonstrated in the “Human Performance Indicators” graph in the named report. Here are the moments when the AI was equal to or superior to humans.
The counting begins in the 1990s, when programs began to beat people in checkers and chess, and ends with very fresh victories for artificial intelligence, when it won matches in the board game of Go and in Dota 2. The graph also includes cases such as identifying skin cancer from photographs at the human level and translation from Chinese into English (although here many experts disagree and prefer human translation).
However, these victories should not be misleading: most of them happen under well-defined rules and fairly simple simulations, such as in video games, which are very conducive to AI training. In general, AI is taught this way: because of its power, it plays as many games per day as a person would play in 180 years, and then wins: this fact underlines how quickly people learn compared to computers.
In addition, the AI that won, for example, in StarCraft II, could not win at chess or checkers, because it would have to be taught again. While artificial intelligence can identify breast tumors in the same way an oncologist can, it cannot do the same with lung cancer (let alone write a prescription or diagnose). In other words: artificial intelligence systems are still quite limited tools that can be used in one area, and not flexible intelligent systems that replace people.
However, this does not negate the fact that AI is sometimes just VERY useful. Despite the limitations, the report shows, machine learning is accelerating and attracting more funding, public interest and helping to solve many problems.
AI IN EDUCATION, SCIENCE AND INVESTMENT
According to the report:
- AI research is advancing rapidly. Between 1998 and 2018, the number of peer-reviewed scientific articles published on the topic of artificial intelligence grew by 300%. Attendance at industry conferences has also grown, with the largest conference, NeurIPS (Neural Information Processing Systems Conference), received 13,500 attendees this year, an 800% increase over 2012.
- Artificial intelligence education is gaining popularity. The number of students taking machine learning courses at universities and online continues to grow. Artificial Intelligence is the most common specialization for computer science graduates in North America. More than 21% of PhDs in computer science specialize in AI, more than double the second most popular area: information security.
- The US continues to be the world leader in AI by most metrics. Although China publishes more articles on artificial intelligence than any country, work done in the United States is having a greater impact, with American authors being cited 40% more than others. The United States invests the most in AI (almost $ 12 billion a year; China is in second place, investing $ 6.8 billion) and registers far more patents than any other country (three times more than, for example, Japan, which ranks second a place).
- AI algorithms learn faster and become cheaper. The report notes that the time it takes to train the algorithm on the popular ImageNet dataset has dropped from three hours in October 2020 to 88 seconds in July 2021. Costs have also dropped from thousands of dollars to double digits.
- AI cars have received more private investment than other areas of AI. About 10% of global private investment ($ 7.7 billion) went to research into driverless vehicles. Medical research and facial recognition (both attracted $ 4.7 billion each) shared the second line in investments, while, for example, the automation of production processes using AI received $ 1 billion, and supply chain management using AI - just over $ 500 million.
HOW IS THE POWER OF THE AI?
All of this is impressive, but it's important to know that no matter how quickly AI thrives, its progress cannot be compared to what pop culture and media headlines credit it with. In fact, artificial intelligence is still very limited in some important aspects.
This is demonstrated in the “Human Performance Indicators” graph in the named report. Here are the moments when the AI was equal to or superior to humans.
The counting begins in the 1990s, when programs began to beat people in checkers and chess, and ends with very fresh victories for artificial intelligence, when it won matches in the board game of Go and in Dota 2. The graph also includes cases such as identifying skin cancer from photographs at the human level and translation from Chinese into English (although here many experts disagree and prefer human translation).
However, these victories should not be misleading: most of them happen under well-defined rules and fairly simple simulations, such as in video games, which are very conducive to AI training. In general, AI is taught this way: because of its power, it plays as many games per day as a person would play in 180 years, and then wins: this fact underlines how quickly people learn compared to computers.
In addition, the AI that won, for example, in StarCraft II, could not win at chess or checkers, because it would have to be taught again. While artificial intelligence can identify breast tumors in the same way an oncologist can, it cannot do the same with lung cancer (let alone write a prescription or diagnose). In other words: artificial intelligence systems are still quite limited tools that can be used in one area, and not flexible intelligent systems that replace people.
However, this does not negate the fact that AI is sometimes just VERY useful. Despite the limitations, the report shows, machine learning is accelerating and attracting more funding, public interest and helping to solve many problems.
