🤖 Hello, this is CLEVI.
We attended the 「2025 JBNU Research Fair Week」 event hosted by Jeonbuk National University from January 20 to 22.
This was a meaningful event that showcased the direction of future technologies linked to industrial and national strategies, centred on Jeonbuk National University’s outstanding research achievements, and provided a forum for exchange among industry, academia, research institutions and government.
In particular, CLEVI CEO Lee Hwan-ho was invited as a VIP speaker for the opening ceremony held on the 20th! 👏
Let’s relive the atmosphere of that day!
This is CEO Lee Hwan-ho preparing for his talk before the event began. You can see him considering how to make the talk more informative and easy to understand.
The event began with the opening ceremony.
The event opened in spectacular fashion with a live art performance completed on the spot by a Jeonbuk National University student.
After the first keynote speech, CLEVI CEO Lee Hwan-ho delivered the second keynote speech on the topic, “Why Does AI Need a Physical Medium?” Let’s take a brief look.
How does AI actually understand the world,
create sentences, and even take action?
We’re going to look inside the “brain” of large language models (LLMs), such as ChatGPT and Claude, which we use every day, to see exactly how they understand and generate sentences.
- Assigning name tags (Token ID): AI first assigns every word a unique number, such as “5” or “8,” to manage it.
- Pinning an address on a map (Vector): However, a simple number isn’t enough to convey meaning, so each word is given a “detailed address (coordinate)” consisting of hundreds of numbers.
- Finding neighbours: Words with similar meanings (for example, “the” and “a”) live close together on this map, like neighbours.
So how does AI predict the next word?
Now that we understand the basic data transformation, let’s look at the entire process of how these numbers become an actual sentence. Suppose we have the input “Paris is the city...” here.
Tokenizer (breaking into pieces): As seen earlier, the sentences we enter are converted into numbers (Token IDs) through the tokenizer. "Paris" becomes number 1652, "is" becomes number 5... and so on.
LLM (massive calculator): These bundles of numbers enter a massive calculator called an LLM (the Bloom model shown in the image). Here, it understands the context through the 'vector' operations we learned about earlier. It recognizes, "Ah, we're talking about the city of Paris right now!"
Logits & Softmax (probability calculation): After completing its calculations, the AI assigns scores to tens of thousands of possible word candidates that could come next. It then passes them through a filter called 'softmax' to convert them 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%
Ultimately, the AI selects **'of'**, which has the highest probability, and completes the sentence "Paris is the city of...". It repeats this process tens of thousands of times to create the long answers we see.
In conclusion, for an AI, remembering a word does not mean memorizing the letters. It is more like remembering the 'address' of where that word lives on a massive map of meaning.
LLMs can be extended to vision, robotics, and behavioural learning.
AI that understands meaning, sees and understands, and acts.
Giving AI arms and legs means expanding intelligence from language to perception, and further to action.
AI is no longer merely an area of ideas or imagination; it has entered the stage of implementation and reality.
CLEVI's next step
CLEVI plans to continue expanding various opportunities, including 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 lecture at Jeonbuk National University became a valuable bridge connecting the students and us, we will continue to serve as a bridge between technologists and users, and between academia and industry.