2021, which was difficult for everyone due to the endless pandemic, is coming to an end. However, in the non-face-to-face era, the development and adoption of artificial intelligence technology has become more active than ever before. AI technology is being actively used in various fields as tremendous changes have occurred, such as telecommuting becoming commonplace and the adoption of remote technology increasing.
This content concludes 2021 and summarizes the major artificial intelligence news for this year. It was selected based on the importance and frequency of artificial intelligence-related articles introduced through various media in 2021. Due to volume, I will introduce major news related to artificial intelligence and natural language processing in 2021 in 2 parts, including this article.
January
AI chatbot Iruda temporarily suspends service
Image source
The AI chatbot 'Iruda' terminated the service due to hate speech and controversy over personal information leakage. It caused social controversy and destroyed not only databases used for learning, but also deep learning conversation models. However, this raised awareness among all, and discussions on establishing AI ethics began in earnest not only among related companies but also at the national level.
February
Naver Announces AI Code of Ethics
Image source
Naver announced the “AI Code of Ethics” along with the SNU AI Policy Initiative (SAPI). Following Kakao's announcement of the “Algorithm Ethics Charter” in 2018, Naver also established principles that must be followed in the development and use of AI. Specifically, it contains a total of 5 provisions: AI development for people, respect for diversity, balance between rational explanation and convenience, service design that takes safety into consideration, and privacy protection and information security.
Incidentally, Twigfarm also drafted the “Language Processing Engine LETR Ethical Principles” in August. This is the result of considering ethical principles appropriate to social responsibility as a company specializing in artificial intelligence and natural language processing. However, it is still in its infancy, and we plan to gather various internal and external opinions and continue to make improvements, so please check the link below for details.
Artificial intelligence ethics: 03. Human-centered AI and LETR ethical principles
Jeffrey Hinton presents GLOM (GLOM) model paper
Image source
Last February, Jeffrey Hinton, the godfather of deep learning, said,How to represent partial-whole hierarchies in a neural network (How to representpart-whole hierarchies in a neural network)k)“I published a paper called Iran. This paper proposes GLOM (GLOM), an evolved deep learning model. Glom is a model of our brain's analogical reasoning, and it is expected that it will go beyond the limitations of previous ML and recognize it in a much more human-like way.
March
National Institute of Korean Language releases Korean materials for learning artificial intelligence for the second time
Image source
The National Institute of Korean Language “Everyone's Corpus” for 8 types of learning data materials needed to improve the Korean language processing ability of artificial intelligence (https://corpus.korean.go.krIt was released in). This is an additional material that will be released following the release of 1.8 billion verses of 13 types of Korean corpus for artificial intelligence learning in August of last year. This material can be used by anyone who has completed an online agreement in “Everyone's Corpus” and is approved.
May
'Google I/O 2021' announces AI conversation model 'LaMDA (LaMDA)'
Image source
At the annual developer event “Google I/O 2021,” Google presented “Lambda (LaMDA),” an AI conversation model that understands language and context. Lambda is an abbreviation of “Language Model for Dialogue Applications (Language Model for Dialogue Applications),” and it is an AI language model that enables natural human-like conversations even with unanswered questions by improving functionality. Google CEO Sunda Pichai introduced Lambda, saying, “You can have a natural conversation without learning predefined answers.”
Naver Unveils Massive AI 'HyperClova (HyperClova) '
Image source
Naver is the first domestic company to unveil “HyperClova (HyperClova),” a super-scale (Hyper scale) artificial intelligence (AI). The GPT-3 model, which is based on Korean rather than the traditional English language, has appeared. After that, Naver plans to expand its language model and continue to develop it to understand images and images.
Big Science project, an alternative to existing large-scale AI language models
Image source
Researchers from around the world are coming together to build an open source LLM BigScience (BigScience) Projecthas started It focused on the limitations and side effects of large language models (LLM) such as lambdas, such as lack of language diversity and learning harmful language patterns. Also, the LLM being built through this project does not aim for a commercial language model, but will be used as a common resource for the scientific community
References
[1] https://www.technologyreview.kr/geoffrey-hinton-glom-godfather-ai-neural-networks/
[2] https://www.technologyreview.com/2020/11/03/1011616/ai-godfather-geoffrey-hinton-deep-learning-will-do-everything/
[3] https://arxiv.org/abs/2102.12627
[4] https://youtu.be/cllFzkvrYmE
[5] https://www.technologyreview.kr/ai-large-language-models-bigscience-project/
[6] https://www.technologyreview.com/2021/05/20/1025135/ai-large-language-models-bigscience-project/
[7] https://www.theverge.com/2021/5/18/22435419/google-io-2021-event-recap-biggest-announcements-pixel-android-12-wear-os-workspace
[8] https://byline.network/2021/05/19-150/
