'In AI development, companies are ahead of academic societies,' 'AI helps the environment but also harms it,' 'The world's best new scientist is AI,' and 'The number of incidents related to AI abuse is increasing rapidly.' Stanford University publishes 'AI Index Report 2023' to report
AI-related services such as ChatGPT and Stable Diffusion have risen from 2022 to 2023, but AI has not only the advantage of helping human life, but also the disadvantage of humans using AI to harm others. has also been frequently pointed out. The AI Research Institute at Stanford University has released a report that collects and analyzes a large amount of data on AI.
AI Index Report 2023 – Artificial Intelligence Index
The important points in this report are roughly as follows.
・ 01: Companies are ahead of academic societies
・ 02: Cannot be measured with conventional benchmark software
03: AI is both useful and harmful to the environment
04:Is the world's best new scientist AI?
・ 05: Incidents related to misuse of AI surged
・ 06: Demand for professional skills related to AI rises in almost all industrial fields in the United States
・07: Private investment in AI decreased year-on-year for the first time in the past 10 years
・ 08: While the percentage of companies adopting AI has peaked, companies that have adopted AI continue to dominate
・ 09: Policy makers' interest in AI is rising
10: Which country is the most positive for AI products and services?
◆ 01: Companies are ahead of academic societies
Until 2014, most of the major machine learning models were published by academic societies, but since then the industry has taken the lead. In 2022, there will be 32 important machine learning models published by industry, compared to just 3 by academia. Building state-of-the-art AI systems requires a large amount of data, computers, and funds, and more of these resources are held by companies than by non-profits or academia.
The total number of AI publications will more than double from 2010 to 2021.
GPT-2, announced in 2019, is said to be the first large-scale language model, but it has 1.5 billion parameters and costs an estimated $50,000 (about 6.6 million yen) to train.
◆ 02: Can not be measured with conventional benchmark software
Many of the AI models have had good results in benchmark tests, but the improvement in scores compared to the previous year seems to be slight. This indicates that many benchmark software have reached saturation, and the emergence of new benchmark software that measures the accuracy of models used in AI is required.
03: AI is both useful and harmful to the environment
Research suggests that AI systems may have a serious impact on the environment. For example, when BLOOM, one of the AI models, is trained, it emits 25 times as much carbon dioxide as a one-way trip from New York to San Francisco, or across the US from the East Coast to the West Coast. That's what I'm talking about.
The figure below is a graph showing the amount of carbon dioxide (unit: tons) emitted by each model. While there are concerns about the environmental impact of models, there are also examples of how AI can be used to optimize energy use, so we plan to improve existing systems while considering the disadvantages of AI is required.
04: Is the world's best new scientist AI?
AI is driving scientific progress at a rapid pace and has brought many benefits to the scientific community, such as AI-assisted humans successfully generating new antibodies. Microsoft is developing AI 'BioGPT' specialized in the biomedical field, and it is expected that the field will continue to develop.
Microsoft Research develops AI ``BioGPT'' specialized in the biomedical field, capable of answering questions with accuracy comparable to human experts-GIGAZINE
◆ 05: Incidents related to misuse of AI surged
According to the AIAAIC database that tracks incidents related to the ethical use of AI, the number of incidents involving AI has increased 26-fold from 2012 to 2021. In 2022, Ukrainian President Volodymyr Zelensky's deepfake video showing his intention to surrender and the case of an American prison using AI call monitoring technology for inmates are attracting attention.
There are also concerns about ethical issues such as bias in the results output by AI models. Models that show fair results are not necessarily bias-free, and models that perform well in fairness have also been found to produce gender-biased results.
◆ 06: Demand for professional skills related to AI rises in almost all industrial fields in the United States
As many services using AI are created, more and more companies are introducing AI as a supplement to improve their daily work efficiency. According to a study by Stanford University, it seems that the number of AI-related job openings has risen in almost every industry sector for which there is formal data. Only the agriculture and hunting sectors are down by 0.02%, but at least in the United States, there is a growing trend to seek workers with AI-related skills.
07: Private investment in AI decreased year-on-year for the first time in the past 10 years
Global private investment in AI in 2022 totaled $91.9 billion (approximately 12 trillion yen), down 26.7% from 2021. The total number of AI-related fundraising events and the number of AI companies starting to raise money have similarly declined. However, there is no doubt that the amount of investment in AI has increased significantly over the past decade, and by 2022 private investment in AI will be 18 times higher than in 2013.
In the amount of investment in AI in each country, the United States has a large lead with 47.36 billion dollars (about 6.22 trillion yen), followed by China with 13.41 billion dollars (about 1.76 trillion yen). Among the major countries, Japan is slightly lower at 720 million dollars (about 94.5 billion yen).
The number of Japanese private companies receiving new investment in 2022 was 32.
The field with the most AI-related investment in 2022 is medical/healthcare, followed by data management/processing/cloud, and finance in third place.
◆ 08: While the percentage of companies adopting AI has peaked, companies that have adopted AI will continue to dominate
According to McKinsey's annual survey results, the percentage of companies adopting AI in 2022 has more than doubled since 2017, but has stagnated between 50% and 60% in recent years. . However, organizations that have adopted AI report realizing meaningful cost savings and revenue increases.
09: Increasing interest of policy makers in AI
When Stanford University analyzed the legislative records of 127 countries, the number of AI-related bills enacted as law was only 1 in 2016, but increased to 37 in 2022. . An analysis of parliamentary records on AI from the same 81 countries found that references to AI in legislative proceedings around the world have increased nearly 6.5 times since 2016. Looking at the seven years since 2016, the number of times AI was mentioned was 1092 times in the UK, and 511 times in Japan.
Looking at the number of laws enacted between 2016 and 2022, the United States led the way with 2, Portugal with 13, Spain with 10, and Japan with 3.
We also know that lawsuits related to AI are on the rise, with a total of 110 lawsuits in the United States alone in 2022. This number is about seven times higher than in 2016.
◆ 10: Which country is the most positive about AI products and services?
In a 2022 survey that asked about impressions of products and services that use AI, it was found that the country with the highest percentage of people who answered that ``the advantages outweigh the disadvantages'' was China. Saudi Arabia and India followed, followed by Japan at 42%. The United States, which has the highest investment in AI, accounted for 35%, indicating that few people had a good impression of AI.
Digging into each question, 41% of Japanese answered that they 'understand well what AI is' and that 'products and services using AI will make their lives easier.' 52% responded. Other answers are below.
・Products and services using AI will greatly change my daily life in the next 3-5 years: 53%
・Products and services using AI have greatly changed my daily life in the past 3-5 years: 30%
・I understand what kinds of products and services use AI: 32%
・I trust companies that use AI as much as other companies: 39%
・ I feel uneasy about products and services that use AI: 20%
Related Posts:
in Software, Posted by log1p_kr