1
00:00:00,000 --> 00:00:05,600
Hello! A new model has sparked widespread discussion over the last couple of days, named JF.

2
00:00:05,600 --> 00:00:14,240
A former OpenAI researcher spent two years developing it after leaving, creating a new model trained in a way that differs from the past.

3
00:00:14,240 --> 00:00:23,600
I think its features are very distinct and powerful. For specific tasks, it can run at a fraction of the cost and hundreds of times the speed of traditional models,

4
00:00:23,600 --> 00:00:29,440
with very low hallucination rates, generating the answers you need, which is highly significant.

5
00:00:29,440 --> 00:00:34,320
它是重要的,因為它選了完全不同的一路由我們現代的AI,

6
00:00:34,320 --> 00:00:36,320
它是在原來的N-Token Prediction。

7
00:00:36,320 --> 00:00:39,840
當大家是比拼了強硬的解決,

8
00:00:39,840 --> 00:00:44,480
硬的解決,和能力的解決方法,

9
00:00:44,480 --> 00:00:47,360
它其實是很小的地步。

10
00:00:47,360 --> 00:00:49,600
我們不需要等待太多的信息,

11
00:00:49,600 --> 00:00:54,960
或多多的解決,我們只需要一個比較的測試,

12
00:00:54,960 --> 00:00:57,040
或是個比例的不同的結果。

13
00:00:57,040 --> 00:00:59,600
我给你一个例子

14
00:00:59,600 --> 00:01:03,180
例如,在一个公司的行业,

15
00:01:03,180 --> 00:01:05,480
如果我们使用了AI的相机,

16
00:01:05,480 --> 00:01:07,140
我们必须要进入到的模特,

17
00:01:07,140 --> 00:01:08,880
等待在模特上的模特,

18
00:01:08,880 --> 00:01:10,980
这些时间需要几秒,对吗?

19
00:01:10,980 --> 00:01:12,100
这一终的效果会给你,

20
00:01:12,100 --> 00:01:14,100
你会在哪个部份接触,

21
00:01:14,100 --> 00:01:15,200
谁要跟他在谈论,

22
00:01:15,200 --> 00:01:18,100
或者是指出的优质是低的,

23
00:01:18,100 --> 00:01:19,380
需要更加注意力。

24
00:01:19,380 --> 00:01:20,880
它给你一个样的效果。

25
00:01:20,880 --> 00:01:23,100
在实际上,用这个新模特,

26
00:01:23,100 --> 00:01:24,900
它可以非常快速,

27
00:01:24,900 --> 00:01:28,900
这几个百分钟的几个百分钟的几个百分钟

28
00:01:28,900 --> 00:01:30,900
这几个百分钟的几个百分钟

29
00:01:30,900 --> 00:01:32,900
这几个百分钟的几个百分钟

30
00:01:32,900 --> 00:01:38,180
the billing department,estimatesthe customer churn rate,and assessesthe likelihoodofneeding

31
00:01:38,180 --> 00:01:41,580
human intervention.

32
00:01:41,580 --> 00:01:47,900
Once theseassessmentsare made,I canassignthe tasktoaspecificdepartmentorouteittoahumanagent,

33
00:01:47,900 --> 00:01:54,520
allowingme tomakethecorrespondingadjustmentsanddecisions.So itsefficiency andcosteffectivenessareincrediblyimpressive.It

34
00:01:54,520 --> 00:02:02,740
isjustextremely,extremelyfast.Even initslivedemonstrationofstructuredoutput,youcanclearlyseethatitgeneratesresults

35
00:02:02,740 --> 00:02:05,080
但是传统的 models

36
00:02:05,080 --> 00:02:06,320
也会一直在用评论

37
00:02:06,320 --> 00:02:08,180
词为文字的评论

38
00:02:08,180 --> 00:02:10,340
所以这两个技术的 approaches

39
00:02:10,340 --> 00:02:12,220
其实完全不同

40
00:02:12,220 --> 00:02:13,960
当 the founder released this video

41
00:02:13,960 --> 00:02:15,140
他说了一篇,

42
00:02:15,140 --> 00:02:17,920
他说了一篇,

43
00:02:17,920 --> 00:02:20,060
我们的认识,

44
00:02:20,060 --> 00:02:22,060
其实是一篇的答案

45
00:02:22,060 --> 00:02:23,860
合适的答案

46
00:02:23,860 --> 00:02:25,820
不太不太确定,

47
00:02:25,820 --> 00:02:27,060
我们喜欢他们,

48
00:02:27,060 --> 00:02:27,640
对吗?

49
00:02:27,640 --> 00:02:28,540
例如,

50
00:02:28,540 --> 00:02:30,200
他们可能会更多 empathetic

51
00:02:30,200 --> 00:02:32,100
或是更多理解我们的

52
00:02:32,100 --> 00:02:35,300
但是,他們的問題是他們的解答問題

53
00:02:35,300 --> 00:02:36,800
人們不可能會不喜歡,

54
00:02:36,800 --> 00:02:39,700
甚至是有個方法的方法。

55
00:02:39,700 --> 00:02:44,100
這帶來更加盡量的信息和 consistency。

56
00:02:44,100 --> 00:02:47,400
但是,他們的解答案是叫做RLCD,

57
00:02:47,400 --> 00:02:50,800
名為 Reinforcement Learning from Calibrated Decisions。

58
00:02:50,800 --> 00:02:52,800
它指的是,在 Reinforcement Learning,

59
00:02:52,800 --> 00:02:54,600
我實在是在調整的解答案是

60
00:02:54,600 --> 00:02:56,600
的解答案是一個解答案。

61
00:02:56,600 --> 00:02:59,800
解答案是每次它提供答案,

62
00:02:59,800 --> 00:03:01,600
它提供一個 probability。

63
00:03:01,600 --> 00:03:02,760
A confidence score

64
00:03:02,760 --> 00:03:05,160
For instance, with this billing inquiry

65
00:03:05,160 --> 00:03:09,760
the probability that I should route it to the billing department is about 90%

66
00:03:09,760 --> 00:03:12,360
But if the customer also mentioned other issues

67
00:03:12,360 --> 00:03:16,360
I believe there's a certain probability it should be prioritized for other departments

68
00:03:16,360 --> 00:03:20,160
So I might assign a 10% or 20% probability to those as well

69
00:03:20,160 --> 00:03:22,960
At this point, you can set a threshold

70
00:03:22,960 --> 00:03:26,360
For example, if the churn risk is very high

71
00:03:26,360 --> 00:03:28,760
I can have a human intervene directly

72
00:03:28,760 --> 00:03:32,760
如果它是低,我可以让它 AI model handle the response.

73
00:03:32,760 --> 00:03:36,760
所以,它帮助我们去解决哪个性能量的权量使用

74
00:03:36,760 --> 00:03:39,520
来解决于不同的情况下。

75
00:03:39,520 --> 00:03:44,240
所以,我相信这个模型会将将其进行我们的生活生活

76
00:03:44,240 --> 00:03:46,360
和很多不同的专业。

77
00:03:46,360 --> 00:03:50,200
以前我们曾经用了一枪创造的创造,

78
00:03:50,200 --> 00:03:53,000
或者我们不需要太多的时间。

79
00:03:53,000 --> 00:03:57,480
但这些随意的随意的选择是我们的选择。

80
00:03:57,480 --> 00:04:02,600
但是在这个新的模式中,我们的选择和认证能更加更加更加适合。

81
00:04:02,600 --> 00:04:06,680
如果我们说我们的选择这个进入我们的页务,

82
00:04:06,680 --> 00:04:11,880
我们肯定能够用这个模式,因为现在,我们的页务员工程,

83
00:04:11,880 --> 00:04:15,640
他们其实会有很多认证和评价的评价。

84
00:04:15,640 --> 00:04:19,480
例如,如果我们的收入的收入我们的预期,

85
00:04:19,480 --> 00:04:21,720
我们还是需要这些认证,

86
00:04:21,720 --> 00:04:24,680
这会常常需要一个人的路程式的方法。

87
00:04:24,680 --> 00:04:26,600
但是如果我们有一个模式的模式,

88
00:04:26,600 --> 00:04:29,800
我们可以定义它的质量从开始

89
00:04:29,800 --> 00:04:34,080
即使是80%,60%,40%的距离,

90
00:04:34,080 --> 00:04:38,520
我们只需要在一个很低的40%的距离。

91
00:04:38,520 --> 00:04:41,080
如果它基本上是在结构的质量,

92
00:04:41,080 --> 00:04:43,480
我们不需要保持人的质量,

93
00:04:43,480 --> 00:04:46,600
我们不需要保持人的质量。

94
00:04:46,600 --> 00:04:50,040
所以,在这个将来的更多的模式,

95
00:04:50,040 --> 00:04:52,280
它可以更加大幅并且在这些模式上

96
00:04:52,280 --> 00:04:55,280
这个模式其实只发了一个新版本

97
00:04:55,280 --> 00:04:59,040
没有提供的技术和计程度

98
00:04:59,040 --> 00:05:01,960
我觉得这个方法非常值得考虑

99
00:05:01,960 --> 00:05:04,160
它使用一个系统一式的方法

100
00:05:04,160 --> 00:05:05,640
是快思考

101
00:05:05,640 --> 00:05:06,880
Daniel Kahneman的作词

102
00:05:06,880 --> 00:05:09,280
说一大部分人的认识

103
00:05:09,280 --> 00:05:11,360
不在实际上的认识

104
00:05:11,360 --> 00:05:12,720
我们立刻的认识

105
00:05:12,720 --> 00:05:14,640
我们立刻的认识

106
00:05:14,640 --> 00:05:17,440
例如,选择哪个衣服

107
00:05:17,440 --> 00:05:18,880
很自然的衣服

108
00:05:18,880 --> 00:05:21,440
我只穿着衣服

109
00:05:21,440 --> 00:05:24,200
and when I'm going to work I might wear a different pair.

110
00:05:24,200 --> 00:05:26,940
So this judgment is actually very intuitive.

111
00:05:26,940 --> 00:05:29,280
It's the same for AI models.

112
00:05:29,280 --> 00:05:34,040
You don't actually need to generate or think through every single judgment

113
00:05:34,040 --> 00:05:36,200
or use chain of thought reasoning.

114
00:05:36,200 --> 00:05:40,200
I can use a smaller model or a model designed for fast thinking

115
00:05:40,200 --> 00:05:41,680
with parallel computing

116
00:05:41,680 --> 00:05:44,680
to accomplish our tasks and goals very efficiently.

117
00:05:44,680 --> 00:05:48,600
So I think it might really make its way into our society soon

118
00:05:48,600 --> 00:05:50,700
because it's cheap, right?

119
00:05:50,700 --> 00:05:53,780
而且是最高的,因為它是很高的,

120
00:05:53,780 --> 00:05:57,100
在0.042萬元的價值,

121
00:05:57,100 --> 00:05:59,780
比 GPT-4的價值,可能是10萬元。

122
00:05:59,780 --> 00:06:03,300
價值是這麼高,他們現在不在賣現在的價值。

123
00:06:03,300 --> 00:06:07,820
所以它可以使用在一大多數的評測 scenarios。

124
00:06:07,820 --> 00:06:10,740
我們看到每個小型的手法

125
00:06:10,740 --> 00:06:13,420
可能有一個決定的模式背後。

126
00:06:13,420 --> 00:06:18,580
以前,一名 classifier would have required someone to spend ages training it, right?

127
00:06:18,580 --> 00:06:22,740
我們以前要有的要求,但現在是沒有必要的

128
00:06:22,740 --> 00:06:25,780
現在,這 model可以幫我們做了一切的評判,

129
00:06:25,780 --> 00:06:28,660
決定定 specific outcomes based on the context

130
00:06:28,660 --> 00:06:34,460
這代表,這代表,已經融化了更多的技術和考驗

131
00:06:34,460 --> 00:06:38,860
我們會看到這個功能伸展開更多的使用

132
00:06:38,860 --> 00:06:42,620
是的,我认为这个是非常有趣的新方式,

133
00:06:42,620 --> 00:06:46,380
一遍完全不同的方法使用力

134
00:06:46,380 --> 00:06:47,980
我們之前說明,

135
00:06:47,980 --> 00:06:50,700
未來可能會有小型的模式

136
00:06:50,700 --> 00:06:52,420
我們只需要有小型的模式

137
00:06:52,420 --> 00:06:54,500
提供的模式

138
00:06:54,500 --> 00:06:56,500
而是用販的模式

139
00:06:56,500 --> 00:06:57,380
每個都市場的模式

140
00:06:57,380 --> 00:06:59,620
我們最近的書

141
00:06:59,620 --> 00:07:00,940
我們最近的書

142
00:07:00,940 --> 00:07:02,540
也有同一個同意

143
00:07:02,540 --> 00:07:05,060
成為我們目標的最適合的方法

144
00:07:05,060 --> 00:07:06,500
是完全不平的

145
00:07:06,500 --> 00:07:11,180
所以我們想分享這些 JB的模式

146
00:07:11,180 --> 00:07:15,360
They've opened their waitlist, so if you'd like to try them out, feel free to join.

147
00:07:15,820 --> 00:07:17,680
Have a great weekend. Bye!