SNSで拡散する画像がAI生成かを見分けるツールは多いが、各社は正解率の測定方法を公開していないため比較が難しい。筆者は小規模な自作検証として、本物の写真50枚とAI生成画像50枚の計100枚を用意し、複数の判定エンジンにかけた。結果、Sightengineは99%、Hiveは95%の正解率を示し、どちらも本物の写真をAIと誤判定しなかった。一方でサンプル数が少ないエンジンもあり、単純な順位付けはできないと筆者は強調している。さらに、正解率の内訳では本物をAIと誤認する誤りの方が実害が大きい点や、判定しきい値の選び方で結果が大きく変わる点が指摘された。
This is an AI-generated summary. ShortSingh links to the original source for the complete article.
Written with AI assistance. Every number comes from my own runs; if you spot an error, tell me and I'll fix it. I had Qwen3-TTS 1.7B read two long narrations, one in Korean and one in English, and scored every sentence chunk against the voice sample using ECAPA speaker embeddings. In these two runs, English held steady and Korean wandered more: the English narration had a median similarity of 0.861 and no chunk below 0.75, while the Korean one had a median of 0.806, with 10 of its 51 chunks under 0.75. The practical lesson is short.
GitHub has logged a new incident on its public status page, as referenced by a Hacker News post. The item links directly to the incident report hosted on githubstatus.com. The post drew modest attention on Hacker News, with 8 points and one comment at the time of capture. The source material provides no details on the affected services, the duration of the disruption, or its root cause. Readers seeking specifics should consult the linked GitHub status incident page.
A developer building SalesData.in, a domain-intelligence platform, explains the difficulties of identifying the technologies behind a website. Detection is complicated because sites may conceal fingerprints, serve assets through CDNs, customize installations, or run several technologies at once. Relying on one signal risks false positives and false negatives, so the approach combines evidence such as HTML patterns, asset paths, HTTP headers, DNS records, and SSL data. Engineering challenges include handling incomplete responses, keeping records consistent across rescans, validating detection rules against real sites, and distinguishing missing signals from failed scans. The author notes that technology detection is only one part of broader domain analysis.
A developer's devlog describes a Blueprint-based sprint and stamina system built in Unreal Engine 5 and tied to the player's health and thirst. Hydration drains at a steady rate while walking, and sprinting uses stamina while speeding up that loss. When stamina runs out, the added sprint drain stops, but the normal drain continues. Reaching zero hydration causes the player to start losing health, so resource management becomes part of movement. Drinks restore hydration, medkits recover health, and the UI highlights critical levels as resources drop.
Discussion (0)
Log in to join the discussion and vote.
Log in