Mai Binciken Bayanan Bayani na AI

Free.ai AI Detector yana bincika rubutu don ya yi amfani da hannayensu na harshe na halittar AI - ƙananan burstiness, parallel triads, canji-baƙo, AI-say magana kamar "leverage" / "utilize" / "in today's fast-paced world" - kuma yana mayar da 0-100 na yiwuwar sakamako, per-phrase confidence, da kuma jerin alamun alamun (AI-mai nuna ko mai nuna mutum). Zaɓi wani irin rubutun pill (gaske / jami'a / blog / imel / mai zane / alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun alamun

Stanford HAI (2023), Vanderbilt (2023), da kuma Jami'ar Maryland (2024) bincike duka nuna ƙarya-positive kashi na 5-15% a kan mutum rubutu da ƙarya-negative kashi sama da 30% a kan zamani LLMs (GPT-4, Claude 3/4, Gemini). OpenAI dakatar da su kansu detector a 2023 da kuma ambaton "ƙasa darajar na gaskiya". Turnitin kansu takardun gargadi dasu AI score kada ya kasance kawai tushe ga wani academic-integrity hukunci. Mu kayan aiki ba wani rufewa - shi ne wani pattern-matching heuristic, ba ƙasa gaskiya.

Yi amfani da shi a cikin hanyoyi uku: (1) Ka duba takardun da ka rubuta kaɗai kafin ka wallafa — idan rubutunka ya yi nasara a kan 85% AI, to, ka tabbata ka yi amfani da ChatGPT ka sake rubuta. (2) Wani alamun daga cikin wasu lokacin da kake duba aikawa — ba hukunci ba. (3) Bincike a cikin rubutun-sha'anin gudu. Yi amfani da shi a cikin hanyoyi da ba a yarda da su ba: zargin dalibai, neman aiki, nuna imel a matsayin "AI mai shakka". Kada ka yi amfani da wani daga cikin waɗannan da wani mai gano (da aka biya ko kyauta) — yawan kuskure yana da girma.

GPTZero shine $ 10–24 / mo. Originality.ai shine $ 14.95 / mo ko $ 0.01 / credit. Copyleaks shine $ 10.99 / mo. Turnitin ya haɗa gano AI cikin shirin $ 3 + / takardar jami'a. Winston AI shine $ 18 / mo. Dukkansu suna wallafa ikirarin aminci na aminci (95%, 99% +) wanda bincike na ilimi na musamman ke musanta. Free.ai kyauta ne kuma muna gaskiya: babu mai gano ya lashe 85% na daidaito akan rubutun mutum / AI mai haɗawa; mafi yawansu sun kasa kasa da 75% akan fitarwa LLM ta zamani. Yi amfani damu a matsayin alamun, kada ku kasance tushen kawai don yanke shawara.

Muna da niyyar yin amfani da alamun alamun alamun GPT-3.5, GPT-4, Claude 3/4, Gemini, Llama, da Mistral. Nau'ukan da ke samar da sauti mai yawa (da aka yi kira da umarnin tsarin ganganci, zafin halitta, ko kuma an daidaita su a kan rubutun mutum) suna da sakamako mafi ƙasƙanci har ma idan sun kasance AI. Nau'ukan da aka yi wa mutum ta hanyar kowane kayan aikin mutum (gami da namu a /text/humanizer/) suna da sakamako mafi ƙasƙanci.

Ba a da yawa daga cikin kalmomi ~100 don a iya auna yawan canzawa (alamar da ke bambanta daidaito na AI daga bambancin halittar mutum). Wani imel mai kalmomi biyu zai iya zama AI ko kuma amsa mai sauri daga mutum — babu wata hanya da za a iya gayawa. Ka sanya kalmomi 200–300 aƙalla don a sami alamun da za a iya karantawa.

@ label

Best accuracy in English (most linguistic research is English). Decent results on Spanish, French, German, Portuguese, Italian. Chinese, Japanese, Arabic work but with higher error rates because AI-tell patterns translate imperfectly. For low-resource languages, results are essentially random.

Wannan shi ne yanayin da mai gano ya fi yawan kuskure. Stanford ta nuna cewa marubuta na Ingilishi da ba na gida ba ana nuna su da alamar AI 61% na lokaci saboda rubutunsu na dabi'a ne mai saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin-saurin

An aika rubutu zuwa GPUs na Free.ai da ke da kansu, an yi nazari, kuma an rufe shi. Ba mu yi nazari kan shigarwar mai amfani ba. Muna yin rijista kawai da lokacin aikawa + farashin alamun shiga don biyan kuɗi — ba a taba rubutun asali ba. Tarihin samarwa (masu amfani da suka shiga) yana adana wani gani na farko na haruffa 200; zaka iya share shi ko da yaushe daga /account/?tab=history.

Ba - nazarin farko kyauta ne, ba imel ba, ba katin bashi ba. Yi rajista don 10,000 free signup tokens + tarihin halitta, mafi girman maɓallan yau da kullun, da damar shiga cikin dukkan 400 + Free.ai kayan aiki.

Na'am. POST zuwa /v1/chat/ tare da tsarin gano tambaya kuma ka rarraba amsawar JSON a cikin mai amfaninka. Idan kana gina wani abu wanda ke amfani da gano AI, KA KARANTA AKADEMIC CAVEATS a sama — kada ka gina halaye da suke yin zargi bisa ga fitarwa na gano kawai. Full snippets at /api/.