“In point of fact, if a person and a girl date, it would either end in romance or a fight. ‘Zeta’ is that form of platform. It’s going to be probably the most immersive chatbot that goes one step beyond the prevailing persona chatbots.”
‘Zeta’ is a synthetic intelligence (AI) ‘story platform’ launched by Scatter Lab in April. It’s a platform that lets you create stories in real time by chatting with AI, going beyond the corporate’s signature emotional conversation.
Specifically, users can create AI characters as they want and develop the story within the direction they need. That is so-called ‘hyper-personalized content.’
Hwang Seong-gu, CTO of Scatter Lab, said that the largest difference between the prevailing ‘Nutty’, which serves Iru Da-na and Kang Da-on, is the sense of immersion.
“When talking to Iruda in Nutty, the user talks as ‘myself,’ but in Zeta, the user becomes a personality within the story,” he explained. “The user can directly give instructions for the situation and immerse themselves more deeply.”
So, conversations with Zeta chatbots are inclined to either get sweet or sour. After all, in the event you coax the chatbot like an individual, things will improve again.
Scatter Lab began its service before famous overseas firms resembling Character.AI and Toki within the Persona chatbot field. With the launch of Zeta, it has once more turn out to be one step ahead of them when it comes to technology.
We added a wide range of ideas which are far more ‘service-like’ when it comes to operation in addition to technology. A representative example is the directive system.
For instance, other chatbots even have situational descriptions resembling ‘sigh’ and ‘shout’, but Zeta is the primary to permit users to directly give instructions. Ji-su Jeong, the product owner (PO) answerable for planning, explained, “The straightforward idea of pressing ‘*’ on the keyboard and entering instructions gives a distinct form of fun to the event of the situation and allows users to immerse themselves more deeply.”
The chatbot’s hallucination phenomenon can be being utilized as a creative element called ‘unpredictable situation’. Since people make mistakes while talking, this example is more humane. CTO Hwang said, “So long as the story’s plausibility is maintained, the chatbot’s unexpected answers can result in creative story development,” adding, “I quite welcome the hallucination phenomenon.”
The introduction of a rating system has made it more fun. PO Jeong said, “We’re revealing the rankings so as of the quantity of real-time conversations to motivate creators,” and “We’re planning a revenue model that may provide incentives to popular creators in the long run.”
It’s explained that whenever you complete missions resembling becoming friendly with popular characters, the quantity of dialogue increases significantly.
For that reason, as of June, there are over 400,000 chatbots created in Zeta, and the variety of users has exceeded 160,000.
Scatter Lab also has top-notch LLM technology. Last November, it developed its own small language model (sLM) ‘PingPong-1’. On the time of Iruda’s relaunch at the tip of 2022, it introduced a multimodal function that had not yet been applied to services overseas, and even read photos.
Through this, the service is provided to users freed from charge. It’s a business model that can’t be attempted because of operating cost issues in places that utilize third-party models. CTO Hwang also emphasized that “technological capabilities are Scatter Lab’s best competitive edge.”
Zeta has also been launched in Japan. Japanese users are less averse to talking to virtual characters, and there are numerous popular IPs resembling light novels and animations, so it’s seen as a market with high potential for achievement.
CTO Hwang said, “We’re watching to see if the speed of return of service users stays constant,” and “In comparison with Korea, we’re seeing an 80% performance.” If the goal figure is achieved, we plan to actively perform marketing.
For this purpose, we also developed a Japanese-specific SLM. “It is because with a purpose to provide sufficient immersion, it’s essential to learn in regards to the culture in addition to simply translate the language,” he explained.
In this manner, Scatter Lab is currently specializing in Zeta service and operation. With a purpose to increase user immersion, the background screen of the chat window is changing depending on the situation, and voice and multimodal functions are being added.
CTO Hwang said, “We’re also developing a function that enables AI to suggest unexpected situations or provide random situational instructions to creatively develop the story,” and “We plan to strengthen the function of profile creation to suit various user preferences.”
As a representative consumer service company, it also has a goal of becoming one in all the ‘most used services by the younger generation’ beyond the keyword of AI chatbot.
In actual fact, 90% of all users are of their 10s and 20s. Specifically, those of their 20s account for 60% of all users.
CTO Hwang Sung-gu said, “The typical every day usage time of users is about 2 hours,” adding, “This can be a significant figure when put next to the typical YouTube viewing time of Koreans, which is 1 hour.”
He continued, “Our goal is to be in the highest 10 services that individuals of their teens and twenties spend probably the most time on by the tip of the 12 months,” and emphasized, “We’ll grow into a number one service within the AI field that may compete with YouTube and Naver Webtoon.”
Reporter Park Soo-bin sbin08@aitimes.com
