The environment surrounding AI is changing rapidly due to the impact of the spread of the new coronavirus infection from 2020 and the growing attention to DX.
Such a situation is also called VUCA. VUCA is a coined word that takes the initials of Volatility, Uncertainty, Complexity, and Ambiguity, and refers to situations in which it is difficult for society and businesses to predict the future.
In this situation, various companies are under pressure to reform their organizations, such as work styles and business models.
- 2021 when the context of “AI in DX” has strengthened
- A year of digitization led by companies with the rise of DX
- Growing importance of DX human resources | Government and council support is the key
- Significant progress has been made in the development of natural language processing models in the Japanese-speaking world
- Focus on corporate sustainability | At the same time, AI governance is important
2021 when the context of “AI in DX” has strengthened
It can be said that AI in 2021 was a year in which it was used by companies as “AI in DX”. I will explain the background and points.
A year of digitization led by companies with the rise of DX
AI technology has been on the rise since the early 2010s, and in the latter half of the 2010s, the need for social utilization increased. However, when focusing on the environment for introducing AI, it was noticeable that the infrastructure necessary for AI utilization, which corresponds to Layer A and Layer B in the diagram below, was not in place, such as underdeveloped software infrastructure and inconsistent data. .
Until 2020, while the introduction of AI was attracting attention in the media, there were many cases in which companies applied AI to limited issues in some departments, and it can be said that local introduction was progressing.
However, due to the spread of the new coronavirus infection from 2020 onwards, an increasing number of companies are developing environments for remote work. In addition, many companies have emerged that digitally replace operations and processes in Layer A by utilizing RPA.
In addition, with the establishment of the Digital Agency in September 2021 and the digitization of the entire country, DX is being rapidly promoted not only by companies but also by the government.
In particular, local governments are using chatbots and AI-OCR to reduce the burden of responding to the increasing number of residents due to the spread of the new coronavirus. As shown in the figure below, the introduction of AI tools in local governments is progressing year by year, mainly for character recognition and voice recogniti
FY2020 Research on AI utilization in local governments (local government AI joint development promotion project) “AI utilization and introduction guidebook in local governments” from page 12
In this way, it can be said that 2021 was a year in which society’s attention was focused not only on AI technology, but also on software in general. Along with this, an increasing number of companies are developing vertical SaaS-style services that solve niche issues in the AI field.
For example, PKSHA Technology revealed its transformation into an “AI SaaS ” company. Focusing on the expanding SaaS market, the company is accelerating investments to expand its SaaS business.
“AI SaaS ” defined by PKSHA Technology is an advanced SaaS that has human perceptual abilities whose functions have been expanded by AI, and uses neural networks to enable complex internal processing.
The level of automation and sophistication of operations is higher than that of general SaaS, and the use of “AI SaaS ” is currently increasing in the area of customer contact points, expanding the customer base. In the future, the company aims to grow by providing “AI SaaS ” products to a wide variety of industries . (*Refer to PKSHA Technology “Business Explanation Material” (December 27, 2021))
OzawaAI has become an entity that is included in DX, and its position as “AI in DX” has advanced. I feel that the number of companies that are also turning overall reforms into services has increased.
Growing importance of DX human resources | Government and council support is the key
And with the flow of DX, there is a big change in the human resources area.
Until now, the need for so-called “AI human resources” who can build machine learning models has been called out, but this year, attention has been focused on “DX human resources”, which is the upper level, and many companies are working to develop DX human resources. It began.
For example, the “ Growth X AI Edition ” provided by Growth X, which supports corporate marketing human resource development, provides a curriculum that allows even humanities to learn the basics of AI in an easy-to-understand manner, supporting the development of digital human resources. In addition to classroom lectures, programs such as AI planning workshops are available within the chat-type app, and it is possible to develop human resources with practical skills who can utilize knowledge and plan on the actual site.
The importance of not only “human resources related to AI” as in the past, but also “human resources who can advance the digitalization of companies and increase their competitive advantage” is increasing.
In November 2021, Prime Minister Kishida announced the creation of a 400 billion yen policy package over three years to develop digital human resources. The government’s backup will further strengthen the education of digital human resources.
In addition, the establishment of the “Digital Literacy Council” by the Japan Deep Learning Association, the Data Scientist Association, and the IPA is one of the major trends in 2021, when the morale of DX human resource development has increased. The Digital Literacy Council defines the basic skills, knowledge, and mindset for using IT, data science, and AI as “Di-Lite,” and conducts educational environment development and enlightenment activities aimed at developing digital human resources. I’m going to continue.
The most read article on AINOW in 2021 was the following article on the theme of “DX human resources”. This article explains DX human resources by industry, skill, and mindset. Please take a look.
Significant progress has been made in the development of natural language processing models in the Japanese-speaking world
In terms of technological trends in the AI field, 2021 was a year in which technological progress advanced in the area of natural language processing, such as large-scale language models (general-purpose language models).
“ GPT-3 ” announced by OpenAI, an American non-profit organization, in 2020 attracted a lot of attention in the English-speaking world due to its advanced performance, but the current GPT-3 has a Japanese generation ability compared to English. inferior.
On the other hand, it was 2021 when the development of large-scale language models progressed in the Japanese-speaking world.
For example, it was noted that rinna Co., Ltd. released a pre-learning model “RoBERTa”, which is an improved version of “GPT-2” and Google’s “BERT”, as open source software in August 2021.
“GPT-2″ is a language model that predicts the next word by considering the words before the word you want to predict. On the other hand, “BERT” can predict not only the word before the word you want to predict, but also the word after it, and the published “RoBERTa” has been reported to have higher performance than BERT. It can be applied to various tasks such as sentence generation, sentence classification, and question answering.
We are also rapidly developing a large-scale general-purpose language model, “HyperCLOVA,” which is currently being developed by LINE Corporation.
This general-purpose language model will use over 175 billion parameters and over 10 billion pages of Japanese data as training data. If we compare this to a newspaper, this is a huge amount of information, equivalent to about 2,760 years of information on the newspaper.
The realization of this gigantic language model will enable the use of LINE services, such as the development of new dialogue AI and the improvement of search service quality, as well as the creation of documents, creative activities, and natural dialogue similar to those performed by humans.
ELYZA Co., Ltd., an AI startup from the University of Tokyo’s Yutaka Matsuo Laboratory, has successfully developed a Japanese AI engine “ELYZA Brain” that utilizes a general-purpose language model since BERT. is realized.
AINOW collaborates with Stockmark Co., Ltd. to publish articles that illustrate natural language processing. Please look at them all together.
OzawaWe cannot communicate without text. Without textbooks, it is impossible to accumulate knowledge or collaborate with others in carrying out work.
As the development of a general-purpose language model progresses in the Japanese-speaking world, it may be possible to automate the sending and receiving of emails, the creation of proposals, and the adjustment of schedules. This increases the possibility of achieving fundamental operational efficiency. On the other hand, there are still hurdles that natural language processing technology must overcome, such as the ambiguity associated with the interpretation of human language. I would like to pay more attention to future movements and disseminate information.
Focus on corporate sustainability | At the same time, AI governance is important
In 2021, society will pay more attention to the SDGs, and at the same time, the environment surrounding companies will change significantly. A particularly important change is that companies are no longer just pursuing sales, but are now emphasizing the value they provide to stakeholders, and the sustainability of corporate management is now being evaluated.
In relation to the SDGs, ESG investment is becoming popular, especially among institutional investors. ESG investment has placed more emphasis on corporate initiatives for the environment, society, and governance.
Among these, the most important item related to AI is G (governance).
Currently, machine learning technology, which is widely used, learns huge amounts of data and autonomously learns its trends, enabling highly accurate image recognition, voice recognition, and prediction. On the other hand, the vast amount of data used to train machine learning models is human-generated big data. Bias (prejudice) is reflected even unconsciously in human-generated data.
As a result, as machine learning models become more popular, it is pointed out that AI may learn about human discrimination and may lead to danger due to erroneous judgments.
Until now, the AI field has focused on how to generate profits in business, but 2021 is also characterized by increased attention on how to promote fair and ethical AI utilization.
As DX permeates and the development of layers A and B progresses, it is predicted that the use of AI will further advance. At this time, it will be necessary to discuss issues such as the transparency of AI and how to assign responsibility in the event of a problem.
In 2022, the importance of discussions on AI and governance will continue to grow.
Publishes interviews with Arisa Ema of the University of Tokyo, who is engaged in international discussions on AI governance, and Masaya Mori, executive officer of the Deloitte Tohmatsu Group. I would love for you to read it.
In short, 2021 can be said to have been a year in which the AI boom calmed down. At the same time, AI is included in the broad concept of DX, and the importance of looking at AI in corporate digital strategies has increased.
When limited to the AI field, it can be said that not only the evolution of existing machine learning technology such as natural language processing technology, but also the technological evolution of next-generation AI is required. In 2021, AINOW has been actively collaborating with the Japanese Society for Artificial Intelligence, disseminating an AI Map that maps research areas in the AI field, and tying up with the journal of the Japanese Society for Artificial Intelligence.