Hello, everyone! 😊
It's time to break down difficult IT news into easy and fun comics, week 34.
This week, there was an incident that made us rethink "Is a conversation with AI a secret?". From the news of new AI models to stories that will affect our writing in the future, we've got it all covered. Would you like to take a look slowly in comic form? 👇
(Just scroll through the images.)
1️⃣ Is the AI landscape shaking again? 'Groq 4.6' released suddenly
Elon Musk's AI company has released a new model, 'Groq 4.6'. It was only 5 weeks since the previous version was released, so the release speed is really fast. Its performance is ranked just below the top tier (OpenAI·Anthropic). The real strength is its 'cost-effectiveness'. It provides similar performance at half the price. However, there are also criticisms that it "confidently gives wrong answers" sometimes.
2️⃣ "Long conversation stopped, everything disappeared" — OpenAI Codex announces major update
'Codex', an AI coding tool for developers, has announced a major update. In the past, long tasks would often stop in the middle, which was frustrating. This update aims to improve it so that the conversation can continue without interruption. It's about making AI work longer and more autonomously. Since it is still in the announcement stage, we need to wait for the official release for details.
3️⃣ "Is a conversation with AI a secret?" — A case where a threatening conversation led to a report
In the United States, a man repeatedly threatened his ex-girlfriend through a chatbot. The AI company reported this to law enforcement (FBI), leading to his arrest. This was because it wasn't just emotional expression but appeared to be a concrete plan. What this incident tells us is clear: Conversations with AI are not entirely private and anonymous, and real threats towards others can be detected and reported. It's something to think about when using AI.
4️⃣ "Can you tell if it was written by AI?" Claude introduces text watermarking (Technology & Economy)
Anthropic's AI 'Claude' has started embedding invisible 'hidden markers (watermarks)' in the text it creates. They say that even if you copy and paste it, the mark will follow. This is a measure taken in response to the new EU AI law which states "AI-generated content must be labeled". Google and OpenAI are also participating. There are expectations that this will help identify misinformation or 'AI spam articles', but there are also concerns that "even my slightly edited text could be branded as AI". Note that this watermark is only a signal of the "possibility of AI involvement" and does not prove who wrote it.
I hope you enjoyed this week's issue. See you next week with more useful news! 🙌
