AI Briefing#kakao
The latest AI and developer news about #kakao, with the original source and a short summary.
The latest AI and developer news about #kakao, with the original source and a short summary.
Emotional expressiveness improved through DPO training and actor recording data, surpassing GPT-4o on Korean image-based benchmarks
By consuming 200 million tokens per day and breaking down role boundaries, the typical development timeline of several months was reduced to 2 weeks.
This introduces a case study of how Kakao manages its corporate attack surface through **YEYE**.
The article introduces an exam design strategy that ensures discrimination and blocks AI's 'click-to-solve' approach through Human-in-the-loop collaboration and dataset personalization.
98% of participants in the AI Mileage Program experienced reduced lead times, and Kakao is driving productivity innovation across all job functions beyond development.
Over 1,000 crew members from Kakao and 12 group companies participated to share AI utilization experiences and technical insights.
Eleven Kakao affiliates participated, showcasing 33 booths covering AI agents, infrastructure operations, and fintech.
At the AI Talk Lounge of if(kakao)25 Krew Day, opportunities and concerns in the AI era are discussed.
Seven CTOs from the Kakao community discuss the opportunities and challenges Kakao faces in the AI era.
This is a panel talk session held at if(kakao)25 Krew Day, capturing the experiences and concerns of FE developers.
At the panel talk hosted by Kakao CTO Jung Kyu-gyu, Shin Jung-gyu emphasized that 40% of coding is currently replaced by AI and that developers must transform into orchestrators managing AI agents.
Kakao has released an FAQ for AI agent competition participants, announcing bonus points for utilizing the Kanana 1.5 model or registering and using tools on playMCP.
This introduces a review of the 1K: Vibe Coding Exhibition, where 99 MVPs were developed in just 1 hour using AI.
Kakao and the Korean Institute of Information Scientists and Engineers (KIISE) are hosting an AI agent competition with a total prize pool of 30 million KRW, offering bonus points for using the Kanana model or playMCP.
98% of developers participating in the pilot reported time savings, and 89% perceived an improvement in quality.
At if(kakao)25, Kakao unveiled a technology evolution roadmap toward Agentic AI, including Kanana-2, which improves inference throughput by applying MLA and MoE, and the omni model Kanana-o, which delivers performance comparable to GPT-4o.
Kakao has established an Orchestrator-based HA standard to replace MHA through R&D efforts since 2023.
It presented concrete prompt examples, such as assigning a Devil's Advocate persona, deriving NFRs from competitor reviews, and automating Definition of Ready (DoR) checklists.
Completed the construction of a 22-page dashboard in one day, which previously took 1-2 weeks, thereby accelerating the speed of analysis sharing.
In a Kakao survey, 98.9% reported improved productivity, while Stack Overflow traffic has decreased by nearly 50% from its peak.
Kakao has developed an AI QA agent that autonomously executes natural language test cases by leveraging LLMs and accessibility APIs on its DeviceFarm infrastructure.
AI handles primary filtering to reduce operators' psychological burden and improve judgment consistency.
The LLM provides regulatory reasons and summaries in JSON format, reducing monitoring review time.
To reduce the burden of manual review, Kakao introduced the Swin Transformer and developed a cognitive psychology-based labeling tool to improve operational efficiency.