translation hallucination

Author: hwisu

src/content/posts/18.three-internships.md

@@ -14,7 +14,7 @@ My first gig was at a small outfit called *Platfarm*, where I did Android dev wo
 For round two I jumped into server work at Smilegate, one of the biggest game companies in Korea (and globally). They’d carved out a newbie‑friendly program where interns ship a small project end‑to‑end. Building backend services was a blast, and I stayed tight with the friends I met there. The downside? Game studios run on serious overtime—the on‑site sauna was always lit, which told me everything. (Silly as it sounds, that’s when I decided I’d rather chase startups than AAA games.)
 
 **Internship #3; Netmarble (AI team)**
-Next up was Netmarble’s AI team. We prototyped all sorts of stuff to get AI into production. The project that actually made it to live servers was a reinforcement‑learning–based anti‑cheat system. We also tried voice synthesis with Tacotron, but the quality was meh. Looking back, speech synthesis feels like a problem half the startups on Earth are now tackling; makes me wonder if we could’ve pushed it further. Doing an “AI internship” sounded cool, but I learned that if you want real AI work you either (a) grind through grad‑school‑level theory and papers, or (b) own the data pipelines/clean‑up side. I’d sampled the data‑wrangling bit—not my cup of tea—and grad school didn’t call to me.
+Next up was Netmarble’s AI team. We prototyped all sorts of stuff to get AI into production. We tried voice synthesis with Tacotron, but the quality was meh. Looking back, speech synthesis feels like a problem half the startups on Earth are now tackling; makes me wonder if we could’ve pushed it further. Doing an “AI internship” sounded cool, but I learned that if you want real AI work you either (a) grind through grad‑school‑level theory and papers, or (b) own the data pipelines/clean‑up side. I’d sampled the data‑wrangling bit—not my cup of tea—and grad school didn’t call to me.
 
 **What I figured out**
 Hopping across those fields made my tastes crystal clear—kind of like dating. I love the conceptual side: analysing a problem, turning ideas into code, and then iterating to make it faster and cleaner. I like that backend code tends to stick around and matter. Long story short: servers are my thing.