For Everyone to Use... Naver's 'No Code AI' Experiment

[Asia Economy Reporter Kang Nahum] Naver has officially begun experimenting with ‘No Code AI’ based on its self-developed ultra-large-scale artificial intelligence (AI) HyperCLOVA. The plan is to expand the versatility of HyperCLOVA so that anyone can utilize AI technology.

◆ No Code AI is already becoming a reality = According to Naver on the 26th, the company started an internal closed beta test (CBT) of ‘HyperCLOVA Studio’ on the 20th. HyperCLOVA Studio is a development environment and tool that allows users to generate text using AI even without any knowledge of AI or coding.

It is based on HyperCLOVA, an ultra-large-scale AI developed by Naver, the first domestic company to do so. HyperCLOVA was developed with 204 billion (204B) parameters. The greater the number of parameters, which indicate the size of the AI model, the more problems the AI can solve.

Using HyperCLOVA Studio, various language-related tasks such as text generation, summarization, classification, transformation, and conversation are possible.

HyperCLOVA Studio consists of ‘Studio’ and ‘Explorer.’ In Studio, users first select the AI engine to use, enter a prompt in the input box, and then press a button to output the result.

For example, if you want to create an AI that generates titles for a promotional event, you input the purpose such as “Create a clever promotional event title based on keywords” along with an example pair of the keyword ‘yeojanikeehudeujipeop’ (women’s Nike hood zip-up) and the title ‘A zip-up perfect for the ambiguous transitional season.’ Then, when only the keyword is provided, the AI automatically generates similar titles. You can also set the length, diversity, and type of the results as desired, and export the results as code to create a test app.

Explorer serves as a kind of community space where users can view, utilize, or reference examples created by others.

Naver plans to refine the system during the CBT period and then release it publicly. By making HyperCLOVA Studio accessible to everyone, it is expected that ultra-large-scale AI technology will become even closer to our daily lives.

◆ HyperCLOVA permeating everyday life = Naver is expanding the application areas of HyperCLOVA by also utilizing it in its search services. A representative example is the integration of HyperCLOVA’s deep learning technology into the voice recognition AI engine on the 8th.

Traditionally, AI deep learning has been conducted through ‘supervised learning,’ which trains data paired with labels that describe the data. For example, an image of a cat is trained together with the label ‘cat.’ This ‘data labeling’ (the task of inputting training data necessary for AI creation) is done manually by humans, which requires significant time and cost to build training data.

On the other hand, ‘self-supervised learning’ is a cutting-edge deep learning technique that can learn from data itself without labels, greatly improving learning efficiency. Naver applied the self-supervised learning technique to its AI voice recognition engine ‘NEST (Neural End-to-end Speech Transcriber),’ increasing voice recognition accuracy by about 30% compared to before.

In addition, Naver plans to apply HyperCLOVA to shopping review summarization and is expanding the application of ultra-large-scale AI to internal technologies and services such as Clova Forecast, which predicts logistics demand.

Seong Nak-ho, Chief Leader of Naver Clova CIC, said, “Our long-term goal is to develop HyperCLOVA as a core tool for ‘AI for everyone.’ We will continue to lead innovation in ultra-large-scale AI technology centered on HyperCLOVA and build a differentiated ecosystem.”