Kim Yoo-cheol, head of the AI ​​X unit at LG AI Research Institute, emphasized the establishment of a ‘use case’ as a key factor for domestic corporations’ competitiveness within the AI ​​era. For instance this, we’ve divided the long run of generative AI into five stages.

Director Kim Yoo-cheol held a session on the theme of ‘Responsible AI and the role of corporations in the brand new AI era’ on the ‘fifth International Smart Technology Conference’ held at COEX in Seoul on the twenty ninth, and said, “Until last yr, in the event you go to a lecture or seminar, I used to be busy explaining what ‘super-giant AI’ is, but now you possibly can skip that part,” he said, first conveying the atmosphere about AI, which has increased significantly in recent times.

“So now, most individuals ask ‘what should I do, and what should our company do?’, and explained this division by way of ‘supply and demand’.

■ Countermeasures from the attitude of supply to directly develop super-large AI

He cited OpenAI’s ChatGPT for instance, saying that the present mainstream of generative AI is “responding to multiple tasks with one large language model (LLM) model with very large parameters.” Nevertheless, it is dear in comparison with its excellent performance and security can be identified, and for that reason, small-scale LLM, or sLLM, has appeared.

In actual fact, the recent AI ecosystem has been revitalized around sLLM, and it’s predicted that the competition for leadership within the ecosystem between Big Tech’s LLM through cloud and sLLM based on open source will likely be intensified over the following few years.

On this atmosphere, domestic AI corporations have explained that thus far, two foremost ways to distinguish themselves from these global corporations have been established. The primary was to create a Korean-style AI that speaks Korean well, and the second was to create a ‘use case’ that works in the sector before global corporations do.

The LG AI Research Center said it was the just one in Korea to say a use case since last yr. He said that the atmosphere began to vary regularly from January of this yr after the looks of ChatGPT.

Because the background for the ‘Korean AI wave’ to develop into a ‘use case wave’, he said, “The success equation of the domestic industry prior to now has develop into a great precedent.” The domestic industry quickly introduced global core technologies and applied them to the domestic market, and based on them, advanced technologies or secured competitiveness, after which challenged the worldwide market. It’s the logic that it has grown to the extent of dominating the worldwide market in some areas with ‘light and fast application technology’ slightly than ‘heavy and huge core technology’.

As well as, two reasons for objecting to the proposal to create Korean AI first and use it to enter the worldwide market were presented. To start with, he explained that the present development of AI may be very fast, so it will not be a situation to sell Korean wells steadily. Here, “AI, unlike other fields, is nice at developing core technologies in addition to other things,” he said.

Due to this fact, he emphasized that domestic supply corporations should give attention to making ‘specific cases for every industry’. It was also identified that since generative AI provided by global big techs is concentrated on versatility, foreign corporations also need separate fine-tuning to introduce AI.

It’s the strategy of the LG AI Research Center that domestic corporations should create specialized cases for every industry and enter the worldwide market, and so they are literally progressing that way, he said. Specifically, it was revealed that manufacturing, bio, finance, medical, media, and public sectors, that are the most important axes of the domestic industry, can be possible to enter overseas markets when successful cases arise in these fields.

■ Response plan from the attitude of demand using super-large AI

Director Kim Yoo-cheol predicted that expectations and concerns about AI would proceed to coexist. It’s an evaluation that, as much because the expectation that AI will expand and innovate the business, there’ll proceed to be concerns that AI will sooner or later encroach on my work and business.

Due to this fact, to begin with, it will be significant to acknowledge the direction of development of generative AI, and suggested five steps within the direction of evolution of generative AI.

The primary is the degree of making something that didn’t exist before with the ‘create, summarize, classify’ function and stimulating human inspiration based on this. The second step is that it became possible to create long texts naturally, which corresponds to the current.

Steps 3 and 4 to follow are that AI will give you the option to predict the long run, followed by criticism and hypotheses, and eventually make decisions at the extent of human experts.

The last step is ‘agentization’. In other words, that is the stage where AI judges as a person entity and even executes it directly. “Existing artificial intelligence (AI) was at the extent of suggesting, ‘What would you want me to do?’ It can go from listing out the crucial tasks to really putting them into motion by combining them with robots or other tools. That is the long run of generative AI.”

“As AI technology continues to advance like this, can everyone really make good use of this high-level AI? It can be different too.”

In other words, it’s predicted that in the event you would not have good data and would not have a number of funds, you’ll use a general model (general model), and vice versa, you’ll use an optimized model.

Director Kim said that the use case is a very powerful so as to not fall into this ‘AI wealthy and poor’.

“You should have a problem-solving ability that differentiates you from others. You should select use cases akin to ‘Only I can solve this problem’ and ‘I’m the very best expert’ and systematically acquire data to resolve the issue.”

“That way, even when AI develops to the agent level, it’ll not be eroded, and ‘this company will likely be evaluated as the very best in a selected case and can give you the option to advance globally.”

Reporter Juyoung Lee juyoung09@aitimes.com