{"text": "大家,這兩天非常熱門的模型是Jer。那這個模型呢,非常有意思。它不能對話,不能寫代碼,也不能識別圖片。它唯一能做的事情就是幫你去做選擇。那本期視頻就給大家來了解一下這一款熱門的決策型模型Jer。其實當我看到這個模型的時候,我也非常詫異。一款不能對話,也不能寫代碼,也不能識別圖片的模型到底有什麼用呢?當我看了官網的例子和外網的一些例子之後,我發現這款模型真的非常有用。因為其實不管是你在工作還是生活中,非常非常多的時候是需要決策的。如果我們用其他的模型,比如說CHALEJPT,或者是Cloud模型去做這種決策的話,推理的成本是非常高的。反應的時間也不會特別快,相當於你用大炮打了一個蚊子。那麼Jer這個模型呢,就是專門幹決策。那我們先用一個非常通俗的例子來理解一下Jer這個模型的輸入和輸出。那麼我們如果要用Jer這個模型的話,需要提供兩個東西。一個叫背景知識。這是我的理解。背景知識就是你要讓他做決定的話,你肯定要提供一些背景知識給他。然後就是問題,你的問題是什麼?發個Jer。那Jer就會做出選擇。然後你拿到這個選擇之後,再去走其他的流程。所以從這個大的這個流程裡面,我們能知道Jer是一個中間環節。就你可能這是你的一個工作流。那麼Jer只是起到一個中間路由,或者是中間判斷的一個過程。那麼他的選擇呢支持三種選擇。那第一種就是啊,我們可以認為叫做yes or no。我們用一個例子來解決定。比如說,小明的歲月成績考了60分。小明的父母是比較溫和的。但是如果他分數低於及格線的話,也會很生氣。那麼我們的背景知識就是這個。那我們的問題,那第一種回復的方式。問題是,父母會生氣嗎?那發給這個Jer。那Jer就會反過一個這樣的概率。會的概率是10%,那麼不會的概率是90%,那麼你就知道父母是不會生氣的。那這種就是Jer的第一種選擇方式。那麼你要告訴我,你在,要用這個Jer就是大摩寫生說,你要告訴他用的是哪一種。那第二種就是我們應該用的最多的,就是選項。還是一樣。但是問題就變成了,父母最可能如何回應小明這次考試。那麼你給出四個選項。A,表揚小明。B,溫和的鼓勵。三,認真的溝通。D,明顯生氣或嚴厲批評。那發給這個Jer。那Jer就給每個選項一個評分。那麼你從評分裡面能看到,是0.55。那C這個選項是最有可能的。認真溝通學習問題保持冷靜。那麼你就得到了一個最有可能的這個結果。最有可能的結果。那這個是第二種。第二種我覺得是在現實場面中用的最多的。那最後一個就是評分式的。那這種方式比如說,這個問題就變成了,父母對這次考試成績會有多不高興。你看這個問題就變成了一個分數的一個判斷。就是你有多大力度去這個選項。比如說A是基本滿意。沒有明顯不高興。B是略有失望。C是明顯不滿意。D呢非常生氣。那他的這個生氣的力度是在增加的。那你發給J5,J5會給到一個ABCD的相似投票一樣。那麼我們可以看到這個他的回答。B0.6現在得了65票。也就是說。那越有失望。讓態度溫和的概率是最高的。那這個時候。那你就可以根據他的評分的不同。去做下一步的選項。所以通過這三個選項的處理啊。還有這個場景啊。我們大概網上裡面就有一個概念了。那J5其實做最大的事情就是。你給他變成知識。你給他問題。給他選項。或者讓他回答yes or no。或者讓他去打一個分數。他其實最大的作用就是這個。幫你去做決策的參考。那很多人可能到這裡的時候。會覺得為什麼我們不用模型。去做這種決策的參考呢。那其實模型也可以做這種決策的參考。但是他Token的推理成本就高。那他小型速度可能不會很快。那比如說J5他的小型時間是100毫秒左右。成本是100萬Token。只要0.042美元。輸出完全免費。所以呢。這也就決定了J5這個模型。只能是成為你眾多工作流的一個部分。某些很多細分場景的模型的使用。它是不可能取代像Cloud。或JPT這些通用模型的。那麼這裡有一個案例可以讓大家進一步去了解。J5這個模型和其他大模型怎麼配合。比如說。你在你的這個A型工具裡面做了一個模型的路由。簡單的任務用便宜的模型。复杂的任務用好的模型。那怎麼去判斷這個任務的。复杂、簡單。那麼這個就可以交給J5去判斷。那他的判斷成本是非常低的。他判斷完了之後。然後再路由到這個模型裡面去。再接著往下走。所以他就成為你A型的裡面的。某個環節裡面的一部分了。那麼在這裡來我也做了一個。用J5來玩這個。然後用J5去做這種選項的判斷。因為俄羅斯方塊只要控制方向就行了。我們來看一下。那這邊的話會接收俄羅斯方塊的這種做標信息。然後他這邊會返回。應該是怎麼操作。返回的指令裡面是包括了他的操作方向。那麼J5的這個判斷的準確率。也不是說是100%。大概在五六成左右。這個可能需要模型去疊代。然後你用其他的大模型其實也可以玩。但是成本和速度可能是沒有這個J5更快。更省成本。那這個想像速度是非常快的。那官網這邊提供了非常多的案例。比如說像這種模型路由啊。還有這種客戶的這種處理。客戶支持的處理。保險理賠。法律合規。那麼我在外網上也是有一個很多。比如說這種去除流氧器廣告。把整個流氧器的道路信息發送給J5。讓J5去判斷哪個是廣告。並且消除。還有這種郵件欺詐的。就是說把所有的郵件和相關的背景信息發給J5。讓J5去判斷哪些郵件的欺詐。那還有這個非常有意思的是什麼呢?它是用J5做了一個Cloud的插件。來判斷一下Cloud有沒有調用不必要的工具。那節省這個Token從100萬降到86K。這個非常厲害。有空的話就可以去試一下。然後這裡面其實還有很多很多這種場景。也就是你可以想像一下。你在生活工作中這些場景是需要判斷的。那麼你就可以使用J5來去做這個郵船裡面的替換。那麼你如果你想體驗這款模型的話,在OpenRot上還有這個官網上都可以去申請使用。官網上這邊申請的話很快就通過了。而且用起來速度也非常快。以上就是我對J5的一些初步看法了。那麼大家可以試一下自己的場景。真的有很多很多場景可以使用它來去體驗一下。OK,那本視頻就到這。希望這個視頻給你有所幫助。希望這個視頻給你更多。希望這個視頻給你更多。希望這個視頻給你更多。希望這個視頻給你更多。希望這個視頻給你更多。希望這個視頻給你更多。希望這個視頻給你更多。希望這個視頻給你更多。希望這個視頻給你更多。希望這個視頻給你更多。希望這個視頻給你更多。希望這個視頻給你更多。希望這個視頻給你更多。希望這個視頻給你更多。希望這個視頻給你更多。希望這個視頻給你更多。希望這個視頻給你更多。希望這個視頻給你更多。", "segments": [{"id": 0, "seek": 0, "start": 0.0, "end": 28.080000000000002, "text": "大家,這兩天非常熱門的模型是Jer。那這個模型呢,非常有意思。它不能對話,不能寫代碼,也不能識別圖片。它唯一能做的事情就是幫你去做選擇。那本期視頻就給大家來了解一下這一款熱門的決策型模型Jer。其實當我看到這個模型的時候,我也非常詫異。一款不能對話,也不能寫代碼,也不能識別圖片的模型到底有什麼用呢?當我看了官網的例子和外網的一些例子之後,我發現這款模型真的非常有用。", "tokens": [50365, 6868, 11, 2664, 16313, 6135, 14392, 32745, 37232, 1546, 41908, 39823, 1541, 41, 260, 1543, 4184, 6287, 41908, 39823, 6240, 11, 14392, 2412, 16697, 1543, 11284, 28590, 2855, 11103, 11, 28590, 4510, 104, 19105, 16337, 120, 11, 6404, 28590, 43143, 16158, 2523, 244, 16668, 1543, 11284, 17198, 107, 2257, 8225, 10907, 1546, 24675, 5620, 32187, 2166, 6734, 10907, 21625, 13167, 229, 1543, 4184, 8802, 16786, 27333, 29940, 3111, 17798, 6868, 25364, 17278, 8861, 32260, 48798, 32745, 37232, 1546, 33540, 7973, 244, 39823, 41908, 39823, 41, 260, 1543, 14139, 13118, 1654, 18032, 6287, 41908, 39823, 20643, 11, 27761, 14392, 6716, 104, 6904, 108, 1543, 2257, 48798, 28590, 2855, 11103, 11, 6404, 28590, 4510, 104, 19105, 16337, 120, 11, 6404, 28590, 43143, 16158, 2523, 244, 16668, 1546, 41908, 39823, 33883, 2412, 7598, 9254, 6240, 30, 13118, 1654, 4200, 2289, 31929, 26352, 1546, 17797, 7626, 12565, 12022, 26352, 1546, 38515, 17797, 7626, 26362, 11, 1654, 38927, 2664, 48798, 41908, 39823, 8034, 14392, 2412, 9254, 1543, 51769], "temperature": 0, "avg_logprob": -0.1505556451268943, "compression_ratio": 1.5866261398176291, "no_speech_prob": 9.342056642158703e-12}, {"id": 1, "seek": 2808, "start": 28.08, "end": 40.36, "text": "因為其實不管是你在工作還是生活中,非常非常多的時候是需要決策的。如果我們用其他的模型,比如說CHALEJPT,或者是Cloud模型去做這種決策的話,推理的成本是非常高的。", "tokens": [50365, 11471, 14139, 45376, 1541, 41410, 41315, 25583, 49958, 5975, 11, 14392, 14392, 6392, 1546, 14010, 1541, 35748, 33540, 7973, 244, 1546, 1543, 13119, 5884, 9254, 9572, 31309, 41908, 39823, 11, 36757, 4622, 34, 4983, 2634, 41, 47, 51, 11, 31148, 1541, 32787, 41908, 39823, 6734, 10907, 33005, 33540, 7973, 244, 21358, 11, 33597, 13876, 1546, 11336, 8802, 1541, 14392, 12979, 1546, 1543, 50979], "temperature": 0, "avg_logprob": -0.2347959435504416, "compression_ratio": 1.3303030303030303, "no_speech_prob": 1.1893099352566772e-11}, {"id": 2, "seek": 2808, "start": 40.36, "end": 53.16, "text": "反應的時間也不會特別快,相當於你用大炮打了一個蚊子。那麼Jer這個模型呢,就是專門幹決策。那我們先用一個非常通俗的例子來理解一下Jer這個模型的輸入和輸出。", "tokens": [50979, 22138, 20481, 1546, 20788, 6404, 21121, 38446, 10251, 11, 15106, 13118, 19488, 2166, 9254, 3582, 4804, 106, 12467, 2289, 8990, 164, 248, 232, 7626, 1543, 20504, 41, 260, 6287, 41908, 39823, 6240, 11, 5620, 46072, 37232, 29932, 33540, 7973, 244, 1543, 46714, 10108, 9254, 8990, 14392, 19550, 7792, 245, 1546, 17797, 7626, 3763, 13876, 17278, 8861, 41, 260, 6287, 41908, 39823, 1546, 12290, 116, 14028, 12565, 12290, 116, 7781, 1543, 51619], "temperature": 0, "avg_logprob": -0.2347959435504416, "compression_ratio": 1.3303030303030303, "no_speech_prob": 1.1893099352566772e-11}, {"id": 3, "seek": 5316, "start": 53.16, "end": 64.16, "text": "那麼我們如果要用Jer這個模型的話,需要提供兩個東西。一個叫背景知識。這是我的理解。背景知識就是你要讓他做決定的話,你肯定要提供一些背景知識給他。", "tokens": [50365, 20504, 5884, 13119, 4275, 9254, 41, 260, 6287, 41908, 39823, 21358, 11, 35748, 20949, 3254, 249, 34623, 26978, 1543, 8990, 19855, 46329, 50218, 6498, 43143, 1543, 16600, 14200, 13876, 17278, 1543, 46329, 50218, 6498, 43143, 5620, 32085, 21195, 5000, 10907, 33540, 12088, 21358, 11, 2166, 14356, 107, 12088, 4275, 20949, 3254, 249, 38515, 46329, 50218, 6498, 43143, 17798, 5000, 1543, 50915], "temperature": 0, "avg_logprob": -0.1736479785954841, "compression_ratio": 1.383399209486166, "no_speech_prob": 1.8103286231196947e-11}, {"id": 4, "seek": 5316, "start": 64.16, "end": 72.16, "text": "然後就是問題,你的問題是什麼?發個Jer。那Jer就會做出選擇。然後你拿到這個選擇之後,再去走其他的流程。", "tokens": [50915, 10213, 5620, 17197, 11, 18961, 17197, 35851, 30, 14637, 3338, 41, 260, 1543, 4184, 41, 260, 47778, 10907, 7781, 21625, 13167, 229, 1543, 10213, 2166, 24351, 4511, 6287, 21625, 13167, 229, 26362, 11, 8623, 6734, 9575, 9572, 31309, 27854, 29649, 1543, 51315], "temperature": 0, "avg_logprob": -0.1736479785954841, "compression_ratio": 1.383399209486166, "no_speech_prob": 1.8103286231196947e-11}, {"id": 5, "seek": 7216, "start": 72.16, "end": 85.16, "text": "所以從這個大的這個流程裡面,我們能知道Jer是一個中間環節。就你可能這是你的一個工作流。那麼Jer只是起到一個中間路由,或者是中間判斷的一個過程。", "tokens": [50365, 7239, 21068, 6287, 39156, 6287, 27854, 29649, 32399, 11, 5884, 8225, 7758, 41, 260, 1541, 8990, 5975, 11016, 40063, 108, 27694, 1543, 3111, 2166, 16657, 16600, 18961, 8990, 41315, 27854, 1543, 20504, 41, 260, 36859, 9147, 4511, 8990, 5975, 11016, 24658, 23786, 11, 31148, 1541, 5975, 11016, 2437, 97, 4307, 115, 1546, 8990, 8816, 29649, 1543, 51015], "temperature": 0, "avg_logprob": -0.16951107449001737, "compression_ratio": 1.2838983050847457, "no_speech_prob": 1.9642058812774366e-11}, {"id": 6, "seek": 7216, "start": 85.16, "end": 91.16, "text": "那麼他的選擇呢支持三種選擇。那第一種就是啊,我們可以認為叫做yes or no。", "tokens": [51015, 20504, 31309, 21625, 13167, 229, 6240, 35488, 10960, 16516, 21625, 13167, 229, 1543, 4184, 18049, 16516, 5620, 4905, 11, 5884, 6723, 39450, 19855, 10907, 2346, 420, 572, 1543, 51315], "temperature": 0, "avg_logprob": -0.16951107449001737, "compression_ratio": 1.2838983050847457, "no_speech_prob": 1.9642058812774366e-11}, {"id": 7, "seek": 9116, "start": 91.16, "end": 102.16, "text": "我們用一個例子來解決定。比如說,小明的歲月成績考了60分。小明的父母是比較溫和的。但是如果他分數低於及格線的話,也會很生氣。", "tokens": [50365, 5884, 9254, 8990, 17797, 7626, 3763, 17278, 33540, 12088, 1543, 36757, 4622, 11, 7322, 11100, 1546, 26256, 6939, 11336, 21030, 122, 26504, 2289, 4550, 6627, 1543, 7322, 11100, 1546, 34131, 35744, 1541, 25174, 16084, 104, 12565, 1546, 1543, 11189, 13119, 5000, 6627, 30622, 41377, 19488, 25703, 30921, 31699, 21358, 11, 6404, 6236, 4563, 8244, 24090, 1543, 50915], "temperature": 0, "avg_logprob": -0.19104016080815742, "compression_ratio": 1.4426751592356688, "no_speech_prob": 2.2579048006288183e-11}, {"id": 8, "seek": 9116, "start": 102.16, "end": 110.16, "text": "那麼我們的背景知識就是這個。那我們的問題,那第一種回復的方式。問題是,父母會生氣嗎?那發給這個Jer。", "tokens": [50915, 20504, 33641, 46329, 50218, 6498, 43143, 5620, 6287, 1543, 4184, 33641, 17197, 11, 4184, 18049, 16516, 8350, 2172, 102, 1546, 9249, 27584, 1543, 17197, 1541, 11, 34131, 35744, 6236, 8244, 24090, 7434, 30, 4184, 14637, 17798, 6287, 41, 260, 1543, 51315], "temperature": 0, "avg_logprob": -0.19104016080815742, "compression_ratio": 1.4426751592356688, "no_speech_prob": 2.2579048006288183e-11}, {"id": 9, "seek": 9116, "start": 110.16, "end": 118.16, "text": "那Jer就會反過一個這樣的概率。會的概率是10%,那麼不會的概率是90%,那麼你就知道父母是不會生氣的。", "tokens": [51315, 4184, 41, 260, 47778, 22138, 8816, 8990, 39425, 26181, 44866, 1543, 6236, 1546, 26181, 44866, 1541, 3279, 8923, 20504, 21121, 1546, 26181, 44866, 1541, 7771, 8923, 20504, 41045, 7758, 34131, 35744, 1541, 21121, 8244, 24090, 1546, 1543, 51715], "temperature": 0, "avg_logprob": -0.19104016080815742, "compression_ratio": 1.4426751592356688, "no_speech_prob": 2.2579048006288183e-11}, {"id": 10, "seek": 11816, "start": 118.16, "end": 126.16, "text": "那這種就是Jer的第一種選擇方式。那麼你要告訴我,你在,要用這個Jer就是大摩寫生說,你要告訴他用的是哪一種。", "tokens": [50365, 4184, 33005, 5620, 41, 260, 1546, 18049, 16516, 21625, 13167, 229, 9249, 27584, 1543, 20504, 32085, 31080, 1654, 11, 41410, 11, 4275, 9254, 6287, 41, 260, 5620, 3582, 34783, 102, 4510, 104, 8244, 4622, 11, 32085, 31080, 5000, 9254, 24620, 17028, 2257, 16516, 1543, 50765], "temperature": 0, "avg_logprob": -0.20400628348849467, "compression_ratio": 1.4407894736842106, "no_speech_prob": 2.5715417642513394e-11}, {"id": 11, "seek": 11816, "start": 126.16, "end": 135.16, "text": "那第二種就是我們應該用的最多的,就是選項。還是一樣。但是問題就變成了,父母最可能如何回應小明這次考試。", "tokens": [50765, 4184, 19693, 16516, 5620, 5884, 26087, 9254, 1546, 8661, 6392, 1546, 11, 5620, 21625, 8313, 227, 1543, 25583, 22977, 1543, 11189, 17197, 3111, 26327, 11336, 2289, 11, 34131, 35744, 8661, 16657, 43526, 8350, 20481, 7322, 11100, 2664, 9487, 26504, 22099, 1543, 51215], "temperature": 0, "avg_logprob": -0.20400628348849467, "compression_ratio": 1.4407894736842106, "no_speech_prob": 2.5715417642513394e-11}, {"id": 12, "seek": 11816, "start": 135.16, "end": 146.16, "text": "那麼你給出四個選項。A,表揚小明。B,溫和的鼓勵。三,認真的溝通。D,明顯生氣或嚴厲批評。那發給這個Jer。", "tokens": [51215, 20504, 2166, 17798, 7781, 19425, 3338, 21625, 8313, 227, 1543, 32, 11, 17571, 11673, 248, 7322, 11100, 1543, 33, 11, 16084, 104, 12565, 1546, 41111, 241, 7978, 113, 1543, 10960, 11, 22041, 8034, 16084, 251, 19550, 1543, 35, 11, 11100, 46605, 8244, 24090, 19780, 29593, 112, 20644, 3416, 117, 39115, 1543, 4184, 14637, 17798, 6287, 41, 260, 1543, 51765], "temperature": 0, "avg_logprob": -0.20400628348849467, "compression_ratio": 1.4407894736842106, "no_speech_prob": 2.5715417642513394e-11}, {"id": 13, "seek": 14616, "start": 146.16, "end": 154.16, "text": "那Jer就給每個選項一個評分。那麼你從評分裡面能看到,是0.55。那C這個選項是最有可能的。", "tokens": [50365, 4184, 41, 260, 3111, 17798, 23664, 3338, 21625, 8313, 227, 8990, 39115, 6627, 1543, 20504, 2166, 21068, 39115, 6627, 32399, 8225, 18032, 11, 1541, 15, 13, 13622, 1543, 4184, 34, 6287, 21625, 8313, 227, 1541, 8661, 2412, 16657, 1546, 1543, 50765], "temperature": 0, "avg_logprob": -0.23762037370588396, "compression_ratio": 1.5016501650165017, "no_speech_prob": 1.8761547870838058e-11}, {"id": 14, "seek": 14616, "start": 154.16, "end": 160.16, "text": "認真溝通學習問題保持冷靜。那麼你就得到了一個最有可能的這個結果。", "tokens": [50765, 22041, 6303, 16084, 251, 19550, 21372, 34025, 17197, 24302, 17694, 32499, 5363, 250, 1543, 20504, 41045, 5916, 21381, 8990, 8661, 2412, 16657, 1546, 6287, 35181, 1543, 51065], "temperature": 0, "avg_logprob": -0.23762037370588396, "compression_ratio": 1.5016501650165017, "no_speech_prob": 1.8761547870838058e-11}, {"id": 15, "seek": 14616, "start": 160.16, "end": 164.16, "text": "最有可能的結果。那這個是第二種。第二種我覺得是在現實場面中用的最多的。", "tokens": [51065, 8661, 2412, 16657, 1546, 35181, 1543, 4184, 41427, 19693, 16516, 1543, 19693, 16516, 17689, 1541, 3581, 9581, 10376, 21446, 8833, 5975, 9254, 1546, 8661, 6392, 1546, 1543, 51265], "temperature": 0, "avg_logprob": -0.23762037370588396, "compression_ratio": 1.5016501650165017, "no_speech_prob": 1.8761547870838058e-11}, {"id": 16, "seek": 14616, "start": 164.16, "end": 173.16, "text": "那最後一個就是評分式的。那這種方式比如說,這個問題就變成了,父母對這次考試成績會有多不高興。", "tokens": [51265, 4184, 20578, 8990, 5620, 39115, 6627, 27584, 1546, 1543, 4184, 33005, 9249, 27584, 36757, 4622, 11, 6287, 17197, 3111, 26327, 11336, 2289, 11, 34131, 35744, 2855, 2664, 9487, 26504, 22099, 11336, 21030, 122, 6236, 2412, 6392, 1960, 12979, 44089, 1543, 51715], "temperature": 0, "avg_logprob": -0.23762037370588396, "compression_ratio": 1.5016501650165017, "no_speech_prob": 1.8761547870838058e-11}, {"id": 17, "seek": 17316, "start": 173.16, "end": 179.16, "text": "你看這個問題就變成了一個分數的一個判斷。就是你有多大力度去這個選項。", "tokens": [50365, 16529, 6287, 17197, 3111, 26327, 11336, 2289, 8990, 6627, 30622, 1546, 8990, 2437, 97, 4307, 115, 1543, 5620, 43320, 6392, 3582, 13486, 13127, 6734, 6287, 21625, 8313, 227, 1543, 50665], "temperature": 0, "avg_logprob": -0.1756838739445779, "compression_ratio": 1.3142857142857143, "no_speech_prob": 1.7976882135672945e-11}, {"id": 18, "seek": 17316, "start": 179.16, "end": 187.16, "text": "比如說A是基本滿意。沒有明顯不高興。B是略有失望。C是明顯不滿意。D呢非常生氣。", "tokens": [50665, 36757, 4622, 32, 1541, 37946, 32197, 9042, 1543, 6963, 11100, 46605, 1960, 12979, 44089, 1543, 33, 1541, 6904, 98, 2412, 28564, 22694, 1543, 34, 1541, 11100, 46605, 1960, 32197, 9042, 1543, 35, 6240, 14392, 8244, 24090, 1543, 51065], "temperature": 0, "avg_logprob": -0.1756838739445779, "compression_ratio": 1.3142857142857143, "no_speech_prob": 1.7976882135672945e-11}, {"id": 19, "seek": 17316, "start": 187.16, "end": 194.16, "text": "那他的這個生氣的力度是在增加的。那你發給J5,J5會給到一個ABCD的相似投票一樣。", "tokens": [51065, 4184, 31309, 6287, 8244, 24090, 1546, 13486, 13127, 1541, 3581, 24228, 252, 9990, 1546, 1543, 4184, 2166, 14637, 17798, 41, 20, 11, 41, 20, 6236, 17798, 4511, 8990, 32, 7869, 35, 1546, 15106, 7384, 120, 37094, 41925, 22977, 1543, 51415], "temperature": 0, "avg_logprob": -0.1756838739445779, "compression_ratio": 1.3142857142857143, "no_speech_prob": 1.7976882135672945e-11}, {"id": 20, "seek": 19416, "start": 194.16, "end": 200.16, "text": "那麼我們可以看到這個他的回答。B0.6現在得了65票。也就是說。", "tokens": [50365, 20504, 5884, 6723, 18032, 6287, 31309, 8350, 28223, 1543, 33, 15, 13, 21, 12648, 5916, 2289, 16824, 41925, 1543, 6404, 26222, 1543, 50665], "temperature": 0, "avg_logprob": -0.3173091961787297, "compression_ratio": 0.8842105263157894, "no_speech_prob": 1.5973561015569793e-11}, {"id": 21, "seek": 20016, "start": 200.16, "end": 204.16, "text": "那越有失望。讓態度溫和的概率是最高的。那這個時候。", "tokens": [50365, 4184, 25761, 2412, 28564, 22694, 1543, 21195, 45248, 13127, 16084, 104, 12565, 1546, 26181, 44866, 1541, 8661, 12979, 1546, 1543, 4184, 6287, 14010, 1543, 50565], "temperature": 0, "avg_logprob": -0.22145891189575195, "compression_ratio": 1.2608695652173914, "no_speech_prob": 1.956961849514105e-11}, {"id": 22, "seek": 20016, "start": 204.16, "end": 209.16, "text": "那你就可以根據他的評分的不同。去做下一步的選項。", "tokens": [50565, 4184, 2166, 36781, 31337, 36841, 31309, 39115, 6627, 1546, 47123, 1543, 6734, 10907, 45271, 31429, 1546, 21625, 8313, 227, 1543, 50815], "temperature": 0, "avg_logprob": -0.22145891189575195, "compression_ratio": 1.2608695652173914, "no_speech_prob": 1.956961849514105e-11}, {"id": 23, "seek": 20016, "start": 209.16, "end": 216.16, "text": "所以通過這三個選項的處理啊。還有這個場景啊。我們大概網上裡面就有一個概念了。", "tokens": [50815, 7239, 19550, 8816, 2664, 10960, 3338, 21625, 8313, 227, 1546, 39289, 13876, 4905, 1543, 15569, 6287, 21446, 50218, 4905, 1543, 5884, 32044, 26352, 5708, 32399, 3111, 2412, 8990, 26181, 33679, 2289, 1543, 51165], "temperature": 0, "avg_logprob": -0.22145891189575195, "compression_ratio": 1.2608695652173914, "no_speech_prob": 1.956961849514105e-11}, {"id": 24, "seek": 21616, "start": 216.16, "end": 225.16, "text": "那J5其實做最大的事情就是。你給他變成知識。你給他問題。給他選項。或者讓他回答yes or no。", "tokens": [50365, 4184, 41, 20, 14139, 10907, 8661, 39156, 24675, 5620, 1543, 2166, 17798, 5000, 26327, 11336, 6498, 43143, 1543, 2166, 17798, 5000, 17197, 1543, 17798, 5000, 21625, 8313, 227, 1543, 31148, 21195, 5000, 8350, 28223, 2346, 420, 572, 1543, 50815], "temperature": 0, "avg_logprob": -0.1790689468383789, "compression_ratio": 1.4523809523809523, "no_speech_prob": 1.761305858105633e-11}, {"id": 25, "seek": 21616, "start": 225.16, "end": 231.16, "text": "或者讓他去打一個分數。他其實最大的作用就是這個。幫你去做決策的參考。", "tokens": [50815, 31148, 21195, 5000, 6734, 12467, 8990, 6627, 30622, 1543, 5000, 14139, 8661, 39156, 11914, 9254, 5620, 6287, 1543, 32187, 2166, 6734, 10907, 33540, 7973, 244, 1546, 47375, 26504, 1543, 51115], "temperature": 0, "avg_logprob": -0.1790689468383789, "compression_ratio": 1.4523809523809523, "no_speech_prob": 1.761305858105633e-11}, {"id": 26, "seek": 21616, "start": 231.16, "end": 237.16, "text": "那很多人可能到這裡的時候。會覺得為什麼我們不用模型。去做這種決策的參考呢。", "tokens": [51115, 4184, 20778, 4035, 16657, 4511, 18771, 20643, 1543, 6236, 11274, 18987, 5884, 24384, 41908, 39823, 1543, 6734, 10907, 33005, 33540, 7973, 244, 1546, 47375, 26504, 6240, 1543, 51415], "temperature": 0, "avg_logprob": -0.1790689468383789, "compression_ratio": 1.4523809523809523, "no_speech_prob": 1.761305858105633e-11}, {"id": 27, "seek": 21616, "start": 237.16, "end": 242.16, "text": "那其實模型也可以做這種決策的參考。但是他Token的推理成本就高。", "tokens": [51415, 4184, 14139, 41908, 39823, 47935, 10907, 33005, 33540, 7973, 244, 1546, 47375, 26504, 1543, 11189, 5000, 51, 8406, 1546, 33597, 13876, 11336, 8802, 3111, 12979, 1543, 51665], "temperature": 0, "avg_logprob": -0.1790689468383789, "compression_ratio": 1.4523809523809523, "no_speech_prob": 1.761305858105633e-11}, {"id": 28, "seek": 24216, "start": 242.16, "end": 247.16, "text": "那他小型速度可能不會很快。那比如說J5他的小型時間是100毫秒左右。", "tokens": [50365, 4184, 5000, 7322, 39823, 31217, 13127, 16657, 21121, 4563, 10251, 1543, 4184, 36757, 4622, 41, 20, 31309, 7322, 39823, 20788, 1541, 6879, 7256, 104, 20956, 29457, 1543, 50615], "temperature": 0, "avg_logprob": -0.15341171026229858, "compression_ratio": 1.3454038997214484, "no_speech_prob": 1.742281319216943e-11}, {"id": 29, "seek": 24216, "start": 247.16, "end": 252.16, "text": "成本是100萬Token。只要0.042美元。輸出完全免費。所以呢。", "tokens": [50615, 11336, 8802, 1541, 6879, 14099, 51, 8406, 1543, 14003, 4275, 15, 13, 14565, 17, 39433, 1543, 12290, 116, 7781, 37100, 2347, 235, 50071, 1543, 40685, 1543, 50865], "temperature": 0, "avg_logprob": -0.15341171026229858, "compression_ratio": 1.3454038997214484, "no_speech_prob": 1.742281319216943e-11}, {"id": 30, "seek": 24216, "start": 252.16, "end": 258.15999999999997, "text": "這也就決定了J5這個模型。只能是成為你眾多工作流的一個部分。", "tokens": [50865, 2664, 6404, 3111, 33540, 12088, 2289, 41, 20, 6287, 41908, 39823, 1543, 14003, 8225, 1541, 11336, 6344, 2166, 29595, 6392, 41315, 27854, 1546, 8990, 32174, 1543, 51165], "temperature": 0, "avg_logprob": -0.15341171026229858, "compression_ratio": 1.3454038997214484, "no_speech_prob": 1.742281319216943e-11}, {"id": 31, "seek": 24216, "start": 258.15999999999997, "end": 264.15999999999997, "text": "某些很多細分場景的模型的使用。它是不可能取代像Cloud。", "tokens": [51165, 17238, 238, 13824, 20778, 41588, 6627, 21446, 50218, 1546, 41908, 39823, 1546, 22982, 9254, 1543, 11284, 1541, 1960, 16657, 29436, 19105, 12760, 32787, 1543, 51465], "temperature": 0, "avg_logprob": -0.15341171026229858, "compression_ratio": 1.3454038997214484, "no_speech_prob": 1.742281319216943e-11}, {"id": 32, "seek": 24216, "start": 264.15999999999997, "end": 269.15999999999997, "text": "或JPT這些通用模型的。那麼這裡有一個案例可以讓大家進一步去了解。", "tokens": [51465, 19780, 41, 47, 51, 29869, 19550, 9254, 41908, 39823, 1546, 1543, 20504, 18771, 2412, 8990, 28899, 17797, 6723, 21195, 6868, 18214, 2257, 31429, 45190, 17278, 1543, 51715], "temperature": 0, "avg_logprob": -0.15341171026229858, "compression_ratio": 1.3454038997214484, "no_speech_prob": 1.742281319216943e-11}, {"id": 33, "seek": 24216, "start": 269.15999999999997, "end": 272.15999999999997, "text": "J5這個模型和其他大模型怎麼配合。比如說。", "tokens": [51715, 41, 20, 6287, 41908, 39823, 12565, 9572, 5000, 3582, 41908, 39823, 11614, 38846, 14245, 1543, 36757, 4622, 1543, 51865], "temperature": 0, "avg_logprob": -0.15341171026229858, "compression_ratio": 1.3454038997214484, "no_speech_prob": 1.742281319216943e-11}, {"id": 34, "seek": 27216, "start": 272.16, "end": 275.16, "text": "你在你的這個A型工具裡面做了一個模型的路由。", "tokens": [50365, 41410, 18961, 6287, 32, 39823, 23323, 39806, 32399, 10907, 2289, 8990, 41908, 39823, 1546, 24658, 23786, 1543, 50515], "temperature": 0, "avg_logprob": -0.18371238264926645, "compression_ratio": 1.6085526315789473, "no_speech_prob": 1.9878957854824186e-11}, {"id": 35, "seek": 27216, "start": 275.16, "end": 279.16, "text": "簡單的任務用便宜的模型。复杂的任務用好的模型。", "tokens": [50515, 31995, 1546, 26443, 39517, 9254, 27364, 2415, 250, 1546, 41908, 39823, 1543, 1787, 235, 4422, 224, 1546, 26443, 39517, 9254, 20715, 41908, 39823, 1543, 50715], "temperature": 0, "avg_logprob": -0.18371238264926645, "compression_ratio": 1.6085526315789473, "no_speech_prob": 1.9878957854824186e-11}, {"id": 36, "seek": 27216, "start": 279.16, "end": 282.16, "text": "那怎麼去判斷這個任務的。复杂、簡單。", "tokens": [50715, 4184, 11614, 6734, 2437, 97, 4307, 115, 6287, 26443, 39517, 1546, 1543, 1787, 235, 4422, 224, 1231, 31995, 1543, 50865], "temperature": 0, "avg_logprob": -0.18371238264926645, "compression_ratio": 1.6085526315789473, "no_speech_prob": 1.9878957854824186e-11}, {"id": 37, "seek": 27216, "start": 282.16, "end": 284.16, "text": "那麼這個就可以交給J5去判斷。", "tokens": [50865, 20504, 6287, 36781, 28455, 17798, 41, 20, 6734, 2437, 97, 4307, 115, 1543, 50965], "temperature": 0, "avg_logprob": -0.18371238264926645, "compression_ratio": 1.6085526315789473, "no_speech_prob": 1.9878957854824186e-11}, {"id": 38, "seek": 27216, "start": 284.16, "end": 286.16, "text": "那他的判斷成本是非常低的。", "tokens": [50965, 4184, 31309, 2437, 97, 4307, 115, 11336, 8802, 1541, 14392, 41377, 1546, 1543, 51065], "temperature": 0, "avg_logprob": -0.18371238264926645, "compression_ratio": 1.6085526315789473, "no_speech_prob": 1.9878957854824186e-11}, {"id": 39, "seek": 27216, "start": 286.16, "end": 290.16, "text": "他判斷完了之後。然後再路由到這個模型裡面去。", "tokens": [51065, 5000, 2437, 97, 4307, 115, 41665, 26362, 1543, 10213, 8623, 24658, 23786, 4511, 6287, 41908, 39823, 32399, 6734, 1543, 51265], "temperature": 0, "avg_logprob": -0.18371238264926645, "compression_ratio": 1.6085526315789473, "no_speech_prob": 1.9878957854824186e-11}, {"id": 40, "seek": 27216, "start": 290.16, "end": 293.16, "text": "再接著往下走。所以他就成為你A型的裡面的。", "tokens": [51265, 8623, 14468, 19382, 29510, 4438, 9575, 1543, 7239, 5000, 3111, 11336, 6344, 2166, 32, 39823, 1546, 32399, 1546, 1543, 51415], "temperature": 0, "avg_logprob": -0.18371238264926645, "compression_ratio": 1.6085526315789473, "no_speech_prob": 1.9878957854824186e-11}, {"id": 41, "seek": 27216, "start": 293.16, "end": 295.16, "text": "某個環節裡面的一部分了。", "tokens": [51415, 17238, 238, 3338, 40063, 108, 27694, 32399, 1546, 2257, 32174, 2289, 1543, 51515], "temperature": 0, "avg_logprob": -0.18371238264926645, "compression_ratio": 1.6085526315789473, "no_speech_prob": 1.9878957854824186e-11}, {"id": 42, "seek": 27216, "start": 295.16, "end": 299.16, "text": "那麼在這裡來我也做了一個。用J5來玩這個。", "tokens": [51515, 20504, 3581, 18771, 3763, 27761, 10907, 2289, 8990, 1543, 9254, 41, 20, 3763, 19912, 6287, 1543, 51715], "temperature": 0, "avg_logprob": -0.18371238264926645, "compression_ratio": 1.6085526315789473, "no_speech_prob": 1.9878957854824186e-11}, {"id": 43, "seek": 29916, "start": 299.16, "end": 302.16, "text": "然後用J5去做這種選項的判斷。", "tokens": [50365, 10213, 9254, 41, 20, 6734, 10907, 33005, 21625, 8313, 227, 1546, 2437, 97, 4307, 115, 1543, 50515], "temperature": 0, "avg_logprob": -0.16848373413085938, "compression_ratio": 1.3291666666666666, "no_speech_prob": 2.4122894654032834e-11}, {"id": 44, "seek": 29916, "start": 302.16, "end": 305.16, "text": "因為俄羅斯方塊只要控制方向就行了。", "tokens": [50515, 11471, 7792, 226, 36004, 30758, 9249, 26268, 14003, 4275, 48707, 25491, 9249, 24282, 3111, 46606, 1543, 50665], "temperature": 0, "avg_logprob": -0.16848373413085938, "compression_ratio": 1.3291666666666666, "no_speech_prob": 2.4122894654032834e-11}, {"id": 45, "seek": 29916, "start": 305.16, "end": 306.16, "text": "我們來看一下。", "tokens": [50665, 5884, 3763, 28324, 1543, 50715], "temperature": 0, "avg_logprob": -0.16848373413085938, "compression_ratio": 1.3291666666666666, "no_speech_prob": 2.4122894654032834e-11}, {"id": 46, "seek": 29916, "start": 306.16, "end": 310.16, "text": "那這邊的話會接收俄羅斯方塊的這種做標信息。", "tokens": [50715, 4184, 22821, 21358, 6236, 14468, 18681, 7792, 226, 36004, 30758, 9249, 26268, 1546, 33005, 10907, 46299, 17665, 26460, 1543, 50915], "temperature": 0, "avg_logprob": -0.16848373413085938, "compression_ratio": 1.3291666666666666, "no_speech_prob": 2.4122894654032834e-11}, {"id": 47, "seek": 29916, "start": 310.16, "end": 311.16, "text": "然後他這邊會返回。", "tokens": [50915, 10213, 5000, 22821, 6236, 3316, 242, 8350, 1543, 50965], "temperature": 0, "avg_logprob": -0.16848373413085938, "compression_ratio": 1.3291666666666666, "no_speech_prob": 2.4122894654032834e-11}, {"id": 48, "seek": 29916, "start": 311.16, "end": 313.16, "text": "應該是怎麼操作。", "tokens": [50965, 26087, 1541, 11614, 13167, 235, 11914, 1543, 51065], "temperature": 0, "avg_logprob": -0.16848373413085938, "compression_ratio": 1.3291666666666666, "no_speech_prob": 2.4122894654032834e-11}, {"id": 49, "seek": 29916, "start": 313.16, "end": 316.16, "text": "返回的指令裡面是包括了他的操作方向。", "tokens": [51065, 3316, 242, 8350, 1546, 25922, 49061, 32399, 1541, 41828, 2289, 31309, 13167, 235, 11914, 9249, 24282, 1543, 51215], "temperature": 0, "avg_logprob": -0.16848373413085938, "compression_ratio": 1.3291666666666666, "no_speech_prob": 2.4122894654032834e-11}, {"id": 50, "seek": 29916, "start": 316.16, "end": 318.16, "text": "那麼J5的這個判斷的準確率。", "tokens": [51215, 20504, 41, 20, 1546, 6287, 2437, 97, 4307, 115, 1546, 28220, 24293, 44866, 1543, 51315], "temperature": 0, "avg_logprob": -0.16848373413085938, "compression_ratio": 1.3291666666666666, "no_speech_prob": 2.4122894654032834e-11}, {"id": 51, "seek": 31816, "start": 318.16, "end": 321.16, "text": "也不是說是100%。大概在五六成左右。", "tokens": [50365, 6404, 7296, 4622, 1541, 6879, 4, 1543, 32044, 3581, 21001, 30566, 11336, 29457, 1543, 50515], "temperature": 0, "avg_logprob": -0.2111447731160229, "compression_ratio": 1.4123076923076923, "no_speech_prob": 2.243719966765756e-11}, {"id": 52, "seek": 31816, "start": 321.16, "end": 323.16, "text": "這個可能需要模型去疊代。", "tokens": [50515, 6287, 16657, 35748, 41908, 39823, 6734, 18140, 232, 19105, 1543, 50615], "temperature": 0, "avg_logprob": -0.2111447731160229, "compression_ratio": 1.4123076923076923, "no_speech_prob": 2.243719966765756e-11}, {"id": 53, "seek": 31816, "start": 323.16, "end": 326.16, "text": "然後你用其他的大模型其實也可以玩。", "tokens": [50615, 10213, 2166, 9254, 9572, 31309, 3582, 41908, 39823, 14139, 47935, 19912, 1543, 50765], "temperature": 0, "avg_logprob": -0.2111447731160229, "compression_ratio": 1.4123076923076923, "no_speech_prob": 2.243719966765756e-11}, {"id": 54, "seek": 31816, "start": 326.16, "end": 330.16, "text": "但是成本和速度可能是沒有這個J5更快。", "tokens": [50765, 11189, 11336, 8802, 12565, 31217, 13127, 16657, 1541, 6963, 6287, 41, 20, 19002, 10251, 1543, 50965], "temperature": 0, "avg_logprob": -0.2111447731160229, "compression_ratio": 1.4123076923076923, "no_speech_prob": 2.243719966765756e-11}, {"id": 55, "seek": 31816, "start": 330.16, "end": 331.16, "text": "更省成本。", "tokens": [50965, 19002, 2862, 223, 11336, 8802, 1543, 51015], "temperature": 0, "avg_logprob": -0.2111447731160229, "compression_ratio": 1.4123076923076923, "no_speech_prob": 2.243719966765756e-11}, {"id": 56, "seek": 31816, "start": 331.16, "end": 333.16, "text": "那這個想像速度是非常快的。", "tokens": [51015, 4184, 6287, 7093, 12760, 31217, 13127, 1541, 14392, 10251, 1546, 1543, 51115], "temperature": 0, "avg_logprob": -0.2111447731160229, "compression_ratio": 1.4123076923076923, "no_speech_prob": 2.243719966765756e-11}, {"id": 57, "seek": 31816, "start": 333.16, "end": 335.16, "text": "那官網這邊提供了非常多的案例。", "tokens": [51115, 4184, 31929, 26352, 22821, 20949, 3254, 249, 2289, 14392, 6392, 1546, 28899, 17797, 1543, 51215], "temperature": 0, "avg_logprob": -0.2111447731160229, "compression_ratio": 1.4123076923076923, "no_speech_prob": 2.243719966765756e-11}, {"id": 58, "seek": 31816, "start": 335.16, "end": 337.16, "text": "比如說像這種模型路由啊。", "tokens": [51215, 36757, 4622, 12760, 33005, 41908, 39823, 24658, 23786, 4905, 1543, 51315], "temperature": 0, "avg_logprob": -0.2111447731160229, "compression_ratio": 1.4123076923076923, "no_speech_prob": 2.243719966765756e-11}, {"id": 59, "seek": 31816, "start": 337.16, "end": 339.16, "text": "還有這種客戶的這種處理。", "tokens": [51315, 15569, 33005, 32316, 1486, 114, 1546, 33005, 39289, 13876, 1543, 51415], "temperature": 0, "avg_logprob": -0.2111447731160229, "compression_ratio": 1.4123076923076923, "no_speech_prob": 2.243719966765756e-11}, {"id": 60, "seek": 31816, "start": 339.16, "end": 341.16, "text": "客戶支持的處理。", "tokens": [51415, 32316, 1486, 114, 35488, 1546, 39289, 13876, 1543, 51515], "temperature": 0, "avg_logprob": -0.2111447731160229, "compression_ratio": 1.4123076923076923, "no_speech_prob": 2.243719966765756e-11}, {"id": 61, "seek": 31816, "start": 341.16, "end": 342.16, "text": "保險理賠。", "tokens": [51515, 24302, 49753, 13876, 13352, 254, 1543, 51565], "temperature": 0, "avg_logprob": -0.2111447731160229, "compression_ratio": 1.4123076923076923, "no_speech_prob": 2.243719966765756e-11}, {"id": 62, "seek": 31816, "start": 342.16, "end": 343.16, "text": "法律合規。", "tokens": [51565, 11148, 2172, 233, 14245, 2888, 237, 1543, 51615], "temperature": 0, "avg_logprob": -0.2111447731160229, "compression_ratio": 1.4123076923076923, "no_speech_prob": 2.243719966765756e-11}, {"id": 63, "seek": 31816, "start": 343.16, "end": 345.16, "text": "那麼我在外網上也是有一個很多。", "tokens": [51615, 20504, 40528, 12022, 26352, 5708, 22021, 2412, 8990, 20778, 1543, 51715], "temperature": 0, "avg_logprob": -0.2111447731160229, "compression_ratio": 1.4123076923076923, "no_speech_prob": 2.243719966765756e-11}, {"id": 64, "seek": 34516, "start": 345.16, "end": 347.16, "text": "比如說這種去除流氧器廣告。", "tokens": [50365, 36757, 4622, 33005, 6734, 32999, 27854, 6610, 100, 34386, 17527, 96, 16846, 1543, 50465], "temperature": 0, "avg_logprob": -0.13146500408968187, "compression_ratio": 1.450310559006211, "no_speech_prob": 1.9990285468618474e-11}, {"id": 65, "seek": 34516, "start": 347.16, "end": 350.16, "text": "把整個流氧器的道路信息發送給J5。", "tokens": [50465, 16075, 27662, 3338, 27854, 6610, 100, 34386, 1546, 6025, 24658, 17665, 26460, 14637, 29309, 17798, 41, 20, 1543, 50615], "temperature": 0, "avg_logprob": -0.13146500408968187, "compression_ratio": 1.450310559006211, "no_speech_prob": 1.9990285468618474e-11}, {"id": 66, "seek": 34516, "start": 350.16, "end": 351.16, "text": "讓J5去判斷哪個是廣告。", "tokens": [50615, 21195, 41, 20, 6734, 2437, 97, 4307, 115, 17028, 3338, 1541, 17527, 96, 16846, 1543, 50665], "temperature": 0, "avg_logprob": -0.13146500408968187, "compression_ratio": 1.450310559006211, "no_speech_prob": 1.9990285468618474e-11}, {"id": 67, "seek": 34516, "start": 351.16, "end": 352.16, "text": "並且消除。", "tokens": [50665, 35934, 20334, 28837, 32999, 1543, 50715], "temperature": 0, "avg_logprob": -0.13146500408968187, "compression_ratio": 1.450310559006211, "no_speech_prob": 1.9990285468618474e-11}, {"id": 68, "seek": 34516, "start": 352.16, "end": 354.16, "text": "還有這種郵件欺詐的。", "tokens": [50715, 15569, 33005, 11178, 113, 20485, 5988, 118, 6716, 238, 1546, 1543, 50815], "temperature": 0, "avg_logprob": -0.13146500408968187, "compression_ratio": 1.450310559006211, "no_speech_prob": 1.9990285468618474e-11}, {"id": 69, "seek": 34516, "start": 354.16, "end": 357.16, "text": "就是說把所有的郵件和相關的背景信息發給J5。", "tokens": [50815, 26222, 16075, 39300, 1546, 11178, 113, 20485, 12565, 15106, 14899, 1546, 46329, 50218, 17665, 26460, 14637, 17798, 41, 20, 1543, 50965], "temperature": 0, "avg_logprob": -0.13146500408968187, "compression_ratio": 1.450310559006211, "no_speech_prob": 1.9990285468618474e-11}, {"id": 70, "seek": 34516, "start": 357.16, "end": 360.16, "text": "讓J5去判斷哪些郵件的欺詐。", "tokens": [50965, 21195, 41, 20, 6734, 2437, 97, 4307, 115, 17028, 13824, 11178, 113, 20485, 1546, 5988, 118, 6716, 238, 1543, 51115], "temperature": 0, "avg_logprob": -0.13146500408968187, "compression_ratio": 1.450310559006211, "no_speech_prob": 1.9990285468618474e-11}, {"id": 71, "seek": 34516, "start": 360.16, "end": 362.16, "text": "那還有這個非常有意思的是什麼呢?", "tokens": [51115, 4184, 15569, 6287, 14392, 2412, 16697, 24620, 7598, 6240, 30, 51215], "temperature": 0, "avg_logprob": -0.13146500408968187, "compression_ratio": 1.450310559006211, "no_speech_prob": 1.9990285468618474e-11}, {"id": 72, "seek": 34516, "start": 362.16, "end": 365.16, "text": "它是用J5做了一個Cloud的插件。", "tokens": [51215, 11284, 1541, 9254, 41, 20, 10907, 2289, 8990, 32787, 1546, 11673, 240, 20485, 1543, 51365], "temperature": 0, "avg_logprob": -0.13146500408968187, "compression_ratio": 1.450310559006211, "no_speech_prob": 1.9990285468618474e-11}, {"id": 73, "seek": 34516, "start": 365.16, "end": 368.16, "text": "來判斷一下Cloud有沒有調用不必要的工具。", "tokens": [51365, 3763, 2437, 97, 4307, 115, 8861, 32787, 28350, 32943, 9254, 1960, 28531, 4275, 1546, 23323, 39806, 1543, 51515], "temperature": 0, "avg_logprob": -0.13146500408968187, "compression_ratio": 1.450310559006211, "no_speech_prob": 1.9990285468618474e-11}, {"id": 74, "seek": 34516, "start": 368.16, "end": 371.16, "text": "那節省這個Token從100萬降到86K。", "tokens": [51515, 4184, 27694, 2862, 223, 6287, 51, 8406, 21068, 6879, 14099, 47421, 4511, 22193, 42, 1543, 51665], "temperature": 0, "avg_logprob": -0.13146500408968187, "compression_ratio": 1.450310559006211, "no_speech_prob": 1.9990285468618474e-11}, {"id": 75, "seek": 34516, "start": 371.16, "end": 372.16, "text": "這個非常厲害。", "tokens": [51665, 6287, 14392, 20939, 1543, 51715], "temperature": 0, "avg_logprob": -0.13146500408968187, "compression_ratio": 1.450310559006211, "no_speech_prob": 1.9990285468618474e-11}, {"id": 76, "seek": 37216, "start": 372.16, "end": 375.16, "text": "有空的話就可以去試一下。", "tokens": [50365, 2412, 23322, 21358, 36781, 6734, 22099, 8861, 1543, 50515], "temperature": 0, "avg_logprob": -0.14696755466690983, "compression_ratio": 1.4623955431754874, "no_speech_prob": 2.2355646847604937e-11}, {"id": 77, "seek": 37216, "start": 375.16, "end": 377.16, "text": "然後這裡面其實還有很多很多這種場景。", "tokens": [50515, 10213, 18771, 8833, 14139, 15569, 20778, 20778, 33005, 21446, 50218, 1543, 50615], "temperature": 0, "avg_logprob": -0.14696755466690983, "compression_ratio": 1.4623955431754874, "no_speech_prob": 2.2355646847604937e-11}, {"id": 78, "seek": 37216, "start": 377.16, "end": 378.16, "text": "也就是你可以想像一下。", "tokens": [50615, 6404, 5620, 42766, 7093, 12760, 8861, 1543, 50665], "temperature": 0, "avg_logprob": -0.14696755466690983, "compression_ratio": 1.4623955431754874, "no_speech_prob": 2.2355646847604937e-11}, {"id": 79, "seek": 37216, "start": 378.16, "end": 381.16, "text": "你在生活工作中這些場景是需要判斷的。", "tokens": [50665, 41410, 49958, 41315, 5975, 29869, 21446, 50218, 1541, 35748, 2437, 97, 4307, 115, 1546, 1543, 50815], "temperature": 0, "avg_logprob": -0.14696755466690983, "compression_ratio": 1.4623955431754874, "no_speech_prob": 2.2355646847604937e-11}, {"id": 80, "seek": 37216, "start": 381.16, "end": 385.16, "text": "那麼你就可以使用J5來去做這個郵船裡面的替換。", "tokens": [50815, 20504, 2166, 36781, 22982, 9254, 41, 20, 3763, 6734, 10907, 6287, 11178, 113, 10256, 117, 32399, 1546, 9531, 123, 36338, 1543, 51015], "temperature": 0, "avg_logprob": -0.14696755466690983, "compression_ratio": 1.4623955431754874, "no_speech_prob": 2.2355646847604937e-11}, {"id": 81, "seek": 37216, "start": 385.16, "end": 387.16, "text": "那麼你如果你想體驗這款模型的話,", "tokens": [51015, 20504, 2166, 13119, 38386, 23987, 24023, 245, 2664, 48798, 41908, 39823, 21358, 11, 51115], "temperature": 0, "avg_logprob": -0.14696755466690983, "compression_ratio": 1.4623955431754874, "no_speech_prob": 2.2355646847604937e-11}, {"id": 82, "seek": 37216, "start": 387.16, "end": 390.16, "text": "在OpenRot上還有這個官網上都可以去申請使用。", "tokens": [51115, 3581, 45569, 49, 310, 5708, 15569, 6287, 31929, 26352, 5708, 7182, 6723, 6734, 3526, 111, 16302, 22982, 9254, 1543, 51265], "temperature": 0, "avg_logprob": -0.14696755466690983, "compression_ratio": 1.4623955431754874, "no_speech_prob": 2.2355646847604937e-11}, {"id": 83, "seek": 37216, "start": 390.16, "end": 392.16, "text": "官網上這邊申請的話很快就通過了。", "tokens": [51265, 31929, 26352, 5708, 22821, 3526, 111, 16302, 21358, 4563, 10251, 3111, 19550, 8816, 2289, 1543, 51365], "temperature": 0, "avg_logprob": -0.14696755466690983, "compression_ratio": 1.4623955431754874, "no_speech_prob": 2.2355646847604937e-11}, {"id": 84, "seek": 37216, "start": 392.16, "end": 394.16, "text": "而且用起來速度也非常快。", "tokens": [51365, 22942, 9254, 21670, 31217, 13127, 6404, 14392, 10251, 1543, 51465], "temperature": 0, "avg_logprob": -0.14696755466690983, "compression_ratio": 1.4623955431754874, "no_speech_prob": 2.2355646847604937e-11}, {"id": 85, "seek": 37216, "start": 394.16, "end": 397.16, "text": "以上就是我對J5的一些初步看法了。", "tokens": [51465, 29497, 5620, 1654, 2855, 41, 20, 1546, 38515, 28727, 31429, 4200, 11148, 2289, 1543, 51615], "temperature": 0, "avg_logprob": -0.14696755466690983, "compression_ratio": 1.4623955431754874, "no_speech_prob": 2.2355646847604937e-11}, {"id": 86, "seek": 37216, "start": 397.16, "end": 399.16, "text": "那麼大家可以試一下自己的場景。", "tokens": [51615, 20504, 6868, 6723, 22099, 8861, 17645, 1546, 21446, 50218, 1543, 51715], "temperature": 0, "avg_logprob": -0.14696755466690983, "compression_ratio": 1.4623955431754874, "no_speech_prob": 2.2355646847604937e-11}, {"id": 87, "seek": 39916, "start": 399.16, "end": 402.16, "text": "真的有很多很多場景可以使用它來去體驗一下。", "tokens": [50365, 8034, 2412, 20778, 20778, 21446, 50218, 6723, 22982, 9254, 11284, 3763, 6734, 23987, 24023, 245, 8861, 1543, 50515], "temperature": 0, "avg_logprob": -0.2179184393449263, "compression_ratio": 0.9923076923076923, "no_speech_prob": 2.097524237631987e-11}, {"id": 88, "seek": 39916, "start": 402.16, "end": 403.16, "text": "OK,那本視頻就到這。", "tokens": [50515, 9443, 11, 4184, 8802, 27333, 29940, 45918, 2664, 1543, 50565], "temperature": 0, "avg_logprob": -0.2179184393449263, "compression_ratio": 0.9923076923076923, "no_speech_prob": 2.097524237631987e-11}, {"id": 89, "seek": 39916, "start": 403.16, "end": 404.16, "text": "希望這個視頻給你有所幫助。", "tokens": [50565, 29955, 6287, 27333, 29940, 17798, 43320, 5966, 32187, 37618, 1543, 50615], "temperature": 0, "avg_logprob": -0.2179184393449263, "compression_ratio": 0.9923076923076923, "no_speech_prob": 2.097524237631987e-11}, {"id": 90, "seek": 40416, "start": 404.16, "end": 405.16, "text": "希望這個視頻給你更多。", "tokens": [50365, 29955, 6287, 27333, 29940, 17798, 2166, 19002, 6392, 1543, 50415], "temperature": 0, "avg_logprob": -0.3243106842041016, "compression_ratio": 11.423076923076923, "no_speech_prob": 1.810114558242759e-11}, {"id": 91, "seek": 40416, "start": 405.16, "end": 406.16, "text": "希望這個視頻給你更多。", "tokens": [50415, 29955, 6287, 27333, 29940, 17798, 2166, 19002, 6392, 1543, 50465], "temperature": 0, "avg_logprob": -0.3243106842041016, "compression_ratio": 11.423076923076923, "no_speech_prob": 1.810114558242759e-11}, {"id": 92, "seek": 40416, "start": 406.16, "end": 407.16, "text": "希望這個視頻給你更多。", "tokens": [50465, 29955, 6287, 27333, 29940, 17798, 2166, 19002, 6392, 1543, 50515], "temperature": 0, "avg_logprob": -0.3243106842041016, "compression_ratio": 11.423076923076923, "no_speech_prob": 1.810114558242759e-11}, {"id": 93, "seek": 40416, "start": 407.16, "end": 408.16, "text": "希望這個視頻給你更多。", "tokens": [50515, 29955, 6287, 27333, 29940, 17798, 2166, 19002, 6392, 1543, 50565], "temperature": 0, "avg_logprob": -0.3243106842041016, "compression_ratio": 11.423076923076923, "no_speech_prob": 1.810114558242759e-11}, {"id": 94, "seek": 40416, "start": 408.16, "end": 409.16, "text": "希望這個視頻給你更多。", "tokens": [50565, 29955, 6287, 27333, 29940, 17798, 2166, 19002, 6392, 1543, 50615], "temperature": 0, "avg_logprob": -0.3243106842041016, "compression_ratio": 11.423076923076923, "no_speech_prob": 1.810114558242759e-11}, {"id": 95, "seek": 40416, "start": 409.16, "end": 410.16, "text": "希望這個視頻給你更多。", "tokens": [50615, 29955, 6287, 27333, 29940, 17798, 2166, 19002, 6392, 1543, 50665], "temperature": 0, "avg_logprob": -0.3243106842041016, "compression_ratio": 11.423076923076923, "no_speech_prob": 1.810114558242759e-11}, {"id": 96, "seek": 40416, "start": 410.16, "end": 411.16, "text": "希望這個視頻給你更多。", "tokens": [50665, 29955, 6287, 27333, 29940, 17798, 2166, 19002, 6392, 1543, 50715], "temperature": 0, "avg_logprob": -0.3243106842041016, "compression_ratio": 11.423076923076923, "no_speech_prob": 1.810114558242759e-11}, {"id": 97, "seek": 40416, "start": 411.16, "end": 412.16, "text": "希望這個視頻給你更多。", "tokens": [50715, 29955, 6287, 27333, 29940, 17798, 2166, 19002, 6392, 1543, 50765], "temperature": 0, "avg_logprob": -0.3243106842041016, "compression_ratio": 11.423076923076923, "no_speech_prob": 1.810114558242759e-11}, {"id": 98, "seek": 40416, "start": 412.16, "end": 413.16, "text": "希望這個視頻給你更多。", "tokens": [50765, 29955, 6287, 27333, 29940, 17798, 2166, 19002, 6392, 1543, 50815], "temperature": 0, "avg_logprob": -0.3243106842041016, "compression_ratio": 11.423076923076923, "no_speech_prob": 1.810114558242759e-11}, {"id": 99, "seek": 40416, "start": 413.16, "end": 414.16, "text": "希望這個視頻給你更多。", "tokens": [50815, 29955, 6287, 27333, 29940, 17798, 2166, 19002, 6392, 1543, 50865], "temperature": 0, "avg_logprob": -0.3243106842041016, "compression_ratio": 11.423076923076923, "no_speech_prob": 1.810114558242759e-11}, {"id": 100, "seek": 40416, "start": 414.16, "end": 415.16, "text": "希望這個視頻給你更多。", "tokens": [50865, 29955, 6287, 27333, 29940, 17798, 2166, 19002, 6392, 1543, 50915], "temperature": 0, "avg_logprob": -0.3243106842041016, "compression_ratio": 11.423076923076923, "no_speech_prob": 1.810114558242759e-11}, {"id": 101, "seek": 40416, "start": 415.16, "end": 416.16, "text": "希望這個視頻給你更多。", "tokens": [50915, 29955, 6287, 27333, 29940, 17798, 2166, 19002, 6392, 1543, 50965], "temperature": 0, "avg_logprob": -0.3243106842041016, "compression_ratio": 11.423076923076923, "no_speech_prob": 1.810114558242759e-11}, {"id": 102, "seek": 40416, "start": 416.16, "end": 417.16, "text": "希望這個視頻給你更多。", "tokens": [50965, 29955, 6287, 27333, 29940, 17798, 2166, 19002, 6392, 1543, 51015], "temperature": 0, "avg_logprob": -0.3243106842041016, "compression_ratio": 11.423076923076923, "no_speech_prob": 1.810114558242759e-11}, {"id": 103, "seek": 40416, "start": 417.16, "end": 418.16, "text": "希望這個視頻給你更多。", "tokens": [51015, 29955, 6287, 27333, 29940, 17798, 2166, 19002, 6392, 1543, 51065], "temperature": 0, "avg_logprob": -0.3243106842041016, "compression_ratio": 11.423076923076923, "no_speech_prob": 1.810114558242759e-11}, {"id": 104, "seek": 40416, "start": 418.16, "end": 419.16, "text": "希望這個視頻給你更多。", "tokens": [51065, 29955, 6287, 27333, 29940, 17798, 2166, 19002, 6392, 1543, 51115], "temperature": 0, "avg_logprob": -0.3243106842041016, "compression_ratio": 11.423076923076923, "no_speech_prob": 1.810114558242759e-11}, {"id": 105, "seek": 40416, "start": 419.16, "end": 420.16, "text": "希望這個視頻給你更多。", "tokens": [51115, 29955, 6287, 27333, 29940, 17798, 2166, 19002, 6392, 1543, 51165], "temperature": 0, "avg_logprob": -0.3243106842041016, "compression_ratio": 11.423076923076923, "no_speech_prob": 1.810114558242759e-11}, {"id": 106, "seek": 40416, "start": 420.16, "end": 421.16, "text": "希望這個視頻給你更多。", "tokens": [51165, 29955, 6287, 27333, 29940, 17798, 2166, 19002, 6392, 1543, 51215], "temperature": 0, "avg_logprob": -0.3243106842041016, "compression_ratio": 11.423076923076923, "no_speech_prob": 1.810114558242759e-11}, {"id": 107, "seek": 40416, "start": 421.16, "end": 422.16, "text": "希望這個視頻給你更多。", "tokens": [51215, 29955, 6287, 27333, 29940, 17798, 2166, 19002, 6392, 1543, 51265], "temperature": 0, "avg_logprob": -0.3243106842041016, "compression_ratio": 11.423076923076923, "no_speech_prob": 1.810114558242759e-11}], "language": "zh"}