🤖 Hello, this is Clevi.

From January 20th to 22nd, I attended the 「2025 JBNU Research Fair Week」 event hosted by Jeonbuk National University.

This event was a meaningful occasion that presented the direction of future technologies linked to industry and national strategies, centered on Jeonbuk National University's outstanding research achievements, and provided a venue for exchange among academia, industry, research, and government.

In particular, Clevi's CEO, Hwanho Lee, was invited as a VIP speaker at the opening ceremony held on the 20th! 👏

Let's relive the atmosphere of that day once again!

Before the start, CEO Hwanho Lee is preparing for his lecture, contemplating how to make it more substantial and easier to understand.

The event began and the opening ceremony was held.

First, the event was splendidly opened with a live art performance completed on the spot by a Jeonbuk National University student.

After the first keynote lecture, the second keynote lecture titled “Why Does AI Need a Physical Medium?” was given by our CEO Hwanho Lee of Clevi. Shall we take a brief look?

How exactly does AI understand the world,

form sentences, and even take actions?

We will take a look inside the 'brain' of large language models (LLMs) like ChatGPT or Claude, which we use every day, to see how they understand and generate sentences.

Tagging names (Token ID): AI first assigns a unique number like '5' or '8' to every word.

Marking addresses on a map (Vector): But since simple numbers alone can't convey meaning, each word is given a 'detailed address (coordinate)' consisting of hundreds of numbers.

Finding neighbors: Words with similar meanings (e.g., 'the', 'a') are clustered together like 'neighbors' on this map.

So how does AI guess the next word?

Now that we've understood basic data transformation, let's look at the entire process of how these numbers are turned into actual sentences. Let's suppose the input is "Paris is the city...".

Tokenizer (splitting): The sentence we input is converted into numbers (Token IDs) through a tokenizer as seen earlier. "Paris" becomes 1652, "is" becomes 5, and so on.

LLM (giant calculator): These bundles of numbers go into the LLM (the Bloom model in the picture), a giant calculator. Here, through the 'vector' operations we learned earlier, it understands the context. It realizes, 'Ah, we're talking about the city of Paris!'

Logits & Softmax (probability calculation): After finishing the calculations, the AI scores thousands of possible next word candidates. Then, through a filter called 'softmax', it converts these into **'probabilities'** that are easier for us to understand.

Final selection: Look at the graph on the right.

of → 33.7%

that → 15.3%

where → 9.4%

In the end, AI chooses the word with the highest probability, **'of'**, and completes the sentence "Paris is the city of...". It repeats this process tens of thousands of times to generate the long answers we see.

In conclusion, for AI, remembering a word is not about memorizing letters, but about remembering the 'address' of where that word lives on a vast map of meaning.

LLMs can be expanded to vision, robotics, and behavioral learning.

AI that understands meaning, sees and understands, and acts.

Giving arms and legs to AI means expanding intelligence from language to perception, and further to action.

AI is no longer just an idea or imagination, but has entered the stage of implementation and reality.

This lecture was more than just a technical introduction for Clevi.

It was because we were able to have in-depth conversations with students on site about how AI technology is connected to the real world and how it can actually be utilized.

Clevi plans to continue expanding opportunities such as on- and off-campus workshops, hands-on projects, and experience booths so that more people can directly experience how AI technology works in the real world.

Just as this Jeonbuk National University lecture became a valuable bridge connecting students and us, we will continue to serve as a bridge between engineers and users, academia and industry.

2025 JBNU Research Fair Week - Introducing the Future with Clevi's AI