If you look at the direction of Naver AI, you can feel the potential of AI Network
Naver had a time to share the present and future of Naver AI technology yesterday. ‘HyperClova’, an ultra-large AI model comparable to Elon Musk’s Open AI, was unveiled here.
Until now, the representative of the ultra-large AI language model was GPT-3, but now it is possible to use ultra-large AI models that surpass GPT-3 language models in Korean.
Naver’s artificial intelligence technology research organization is called ‘Clova’. Naver has developed various AI technologies such as natural language processing, voice recognition, voice synthesis, and OCR.
Naver’s capabilities, which started as a search engine and turned into Big Tech, have been well shown not only by artificial intelligence services such as Clova Note but also by artificial intelligence devices such as Clova Clock and Clover Lamp.
Naver is now able to collaborate with SoftBank due to the merger of Yahoo Japan and Line. At a time when we are expecting another synergy with Japan’s Big Tech, we can see that Naver’s artificial intelligence service is forming a super gap.
The trend of artificial intelligence is now well revealed by one word, ‘Big BIG’. The existing artificial intelligence development method took a long time and a lot of resources to get the results separately. Also, it was hard to be used for another model.
Now it’s a trend to integrate through one big model. Big models mean that one large model can be applied to a variety of areas. Platforming and expanding the model becomes possible. The fact that one big model can be applied to a variety of services, such as shopping recommendations and customer service, means that it will be able to take less time and use less resources. A new era of artificial intelligence is opening as a ‘big model’.
Considering the development of semiconductor technology, it is easy to understand. Every time the integration level of semiconductors increased, a new service could come out. In artificial intelligence, the number of ‘parameters’ is the same as the integration of semiconductors, so increasing parameters increases the likelihood of solving new problems that artificial intelligence can solve. Global technology giants are accelerating investment with the disruptive innovations that large artificial intelligence models will bring, and this direction eventually represents an era of bigger models and faster learning.
Naver also participated in this competition recently. To this end, Naver has invested in supercomputers, data, and artificial intelligence experts. First of all, Naver introduced a supercomputer with more than 700 petaflops performance last year. It’s a supercomputer that can calculate around 70 times per second.
Naver is a search engine used by most people in Korea use. There will be a lot of data that comes up every day by day. As a result, it was able to build on 6,500 times more Korean data than GPT-3 made by Elon Musk’s OpenAI.
It is a super-large artificial intelligence language model that has learned almost all types of Korean, which is 3000 times more accessible than the existing Naver language model, information worth 9 years, news worth 50 years. This was possible thanks to ‘unsupervised learning’ in which artificial intelligence learns and trains data on its own.
Naver’s ‘HyperClova’ Artificial Intelligence Service, which combines state-of-the-art artificial intelligence technology but can be enjoyed without knowing the details of the technology! It seems perfect! But it seems to feel that artificial intelligence is what Naver can’t do ‘on its own.’
Naver said Seoul National University’s AI graduate school and KAIST AI graduate school, as well as global R&D cooperation, will continue to create an ‘artificial intelligence ecosystem’ together.
AI Network has also been presenting various open-source artificial intelligence models through its own artificial intelligence platform, ‘Ainize’. ‘AI Writers’ as well as various open chatbots, resume writing models, Kant models, etc. can be implemented directly in Ainize.
As you can see in today’s article, AI Network is solving the problem of ‘computing resources’ that even super-large companies like Naver feel limited.
The ‘resource’ problem of artificial intelligence development, which can become polarized as artificial intelligence models move toward ‘BIG’, can be properly solved in AI Network at a time when the need for ‘open resources’ emerges in ‘open sources’. If there is an AI network that everyone implements together, global artificial intelligence developers who ‘can’t easily use supercomputers’ can gather together to create bigger resources.
It was time to check the direction of AI network through Naver once again. AI Network have already collaborated with Line and Line Blockchain and have attracted token investment from Line’s blockchain subsidiary, Unblock. It was also a time to look forward to future cooperation with Naver, which is leading in both artificial intelligence and blockchain.
If you want to replay today’s conference, you can check it out at the link below.