US Stanford University AI report “The stage of emerging technology has passed”
Last year’s investment of 100 trillion won… Laws related to achieving low cost and high efficiency
are on the rise… Ethics as a new challenge
From emerging technology to mature technology.Stanford University Human-Centered Artificial Intelligence Institute (HAI) in the US recently announced the '2022 AI Index' report that "2021 was the year in which artificial intelligence took off as a mature technology after passing the stage of emerging technology." did.
This means that AI technology has permeated industry and life in general. In the 2012 image recognition contest, a new deep learning method called 'convolutional neural network' appeared and created an artificial intelligence boom 10 years after the announcement of AI's incorporation into the mainstream. Earlier , in a report from the research institute in 2019, the research institute announced that the performance of artificial intelligence technology was doubling every 3.4 months with acceleration starting in 2012. This is seven times faster than Moore's Law, which states that the density of computer chips doubles every two years.
"We are no longer talking about artificial intelligence in reasoning, but its positive and negative impacts on the real world," said Jack Clark, co-chair of the report's steering committee, in a press release.According to the report, which was released for the fifth time this year, artificial intelligence is fully integrated into the mainstream economy, and is seen throughout funding, research and deployment. There are several reasons for this judgment in the report.
First of all, investment in artificial intelligence has surged.In 2021, private investment in artificial intelligence more than doubled from a year earlier to $93.5 billion (about 113 trillion won). This is the largest increase since it doubled in 2013-2014.However, the target of investment showed a biased pattern. The number of newly invested companies decreased from 762 in 2020 to 746 last year. This is the third consecutive year of decline since 2019. As a result, the average size of investment also increased, and the number of cases of raising more than $500 million nearly quadrupled from 4 in 2020 to 15 last year.This is largely due to the brisk mergers and acquisitions. Paypal, an online payment platform, acquired Paydy, a BNPL service, for $2.7 billion, and Microsoft acquired Nuance, a voice recognition technology company, for $16 billion.By sector, data management and processing and cloud technology companies received the most investment, followed by healthcare and fintech companies. By country, the United States was the overwhelming majority. It is more than twice that of China, which is in second place.
Second, the low-cost and high-efficiency of artificial intelligence is making remarkable achievements.The report found that since 2018, the cost of training image classification has decreased by 63.6%, while the learning time has improved by 94.4%. In the case of ImageNet, a leading image learning tool, training costs fell to $4.6 in 2021. That's only 0.4% of the $1112 cost to achieve similar performance in 2017. This means that the cost of artificial intelligence training has fallen to a low level in four years.
In the 2020 report of OpenAI, an artificial intelligence development company founded by Elon Musk, ImageNet showed that the amount of learning required for training doubles every 16 months. The language model Retro, recently developed by DeepMind, a subsidiary of Google Alphabet, is 25 times more efficient than other language models.There is also a clear decline in the price of robot arms, which has been in great demand since Corona 19. In five years, the price has dropped to a quarter of the level. According to the report, the median price of a robotic arm fell 74% in five years, from $50,000 in 2016 to $12,845 in 2021.There are active research achievements at the foundation of this trend. According to the report, the number of AI patent applications in 2021 is 30 times higher than in 2015. The average annual growth rate is 77%.
Third, the ethical issue of artificial intelligence is emerging as a major concern.It is a phenomenon caused by the growing impact on industry and life. At the heart of ethical issues are fairness and bias.Research on the transparency and fairness of artificial intelligence has increased fivefold in the past four years. Researchers affiliated with companies presented 71% more research on the fairness of artificial intelligence at academic conferences and workshops than a year ago.However, the report notes that "the trend is encouraging, but companies are trying to limit internal research looking at their technology critically." For example, the AI ethics team of a large corporation such as Meta (Facebook) is like a 'teethed tiger'. The report reports that Google disbanded just a week after creating an AI advisory board in 2019, and that IBM, which has been emphasizing fairness, once secretly trained facial recognition and racial classification models with New York police to build a video surveillance system. Examples of what has been done.In fact, efforts to solve ethical issues in the AI industry are still not meeting expectations. According to a McKinsey survey, a consulting firm, 29% and 41% of business respondents perceive 'fairness and fairness' and 'explainability' as risks when adopting AI, but 19% each take steps to reduce these risks , only 27%.
As ethical issues arose, laws related to artificial intelligence are being enacted one after another.The number of AI-related laws passed by parliaments of 25 countries has surged from one in 2016 to 18 in 2021. Spain, the United Kingdom and the United States passed the most relevant bills, three each.In the United States, the number of artificial intelligence-related bills proposed by lawmakers exploded from just one in 2015 to 130 in 2021. However, only 2% of the bills were passed.Along with the report, the institute also announced the 'Global AI Vibrancy Tool', which shows the competitiveness of AI in each country around the world.Based on 23 indicators, the United States took first place in the AI vitality leaderboard, which weights R&D and the economy. It was followed by China, India, and the United Kingdom, followed by Korea in sixth place after Canada. Korea has been moving up and down from 4th to 8th every year since 2017, when the rankings started.