1
00:00:00,920 --> 00:00:05,480
Speaker 1: I don't entertainment, put on your dance and shift.

2
00:00:05,559 --> 00:00:06,280
Speaker 2: Don't give me your.

3
00:00:06,240 --> 00:00:12,040
Speaker 3: Fast shot from the Daily Os. She's here, Emma Gillespie, Hello.

4
00:00:11,920 --> 00:00:15,400
Speaker 2: Good morning. Well we're talking AI this morning, as we

5
00:00:15,440 --> 00:00:17,440
Speaker 2: so often are, but there've been a couple of big,

6
00:00:17,800 --> 00:00:22,800
Speaker 2: high profile AI fake voices making the headlines, including our

7
00:00:22,840 --> 00:00:26,279
Speaker 2: own Sandra Sully. This is a really interesting one. I

8
00:00:26,320 --> 00:00:28,240
Speaker 2: want to play you a grab that was used in

9
00:00:28,360 --> 00:00:30,240
Speaker 2: a true crime podcast.

10
00:00:30,640 --> 00:00:35,440
Speaker 1: Saint Patrick's college teacher arrested and facing multiple alleged child

11
00:00:35,479 --> 00:00:36,360
Speaker 1: abuse charges.

12
00:00:36,960 --> 00:00:41,319
Speaker 2: That's not Sandra Sully. So in this podcast they say

13
00:00:41,360 --> 00:00:44,120
Speaker 2: before they go to the grab, this is an AI

14
00:00:44,280 --> 00:00:46,760
Speaker 2: generated voice reading out a headline.

15
00:00:46,800 --> 00:00:49,760
Speaker 4: So it's not generic that they've modeled Sandra's voice.

16
00:00:49,800 --> 00:00:51,839
Speaker 2: They disclose it as AAI, but they don't say that

17
00:00:51,880 --> 00:00:55,120
Speaker 2: they've modeled Sandra. Clearly, it sounds like that iconic voice

18
00:00:55,120 --> 00:00:58,080
Speaker 2: we know. Sandra actually did an interview with the project

19
00:00:58,200 --> 00:01:01,160
Speaker 2: talking about it, and she said she didn't feel violated,

20
00:01:01,200 --> 00:01:04,240
Speaker 2: but she felt more concerned about the future of news

21
00:01:04,280 --> 00:01:05,720
Speaker 2: and what that means his a little bit.

22
00:01:05,920 --> 00:01:08,600
Speaker 1: So if you know, I'm arguably some sort of trusted

23
00:01:08,680 --> 00:01:12,440
Speaker 1: voice in news, then how can that be manipulated? And

24
00:01:12,840 --> 00:01:16,480
Speaker 1: for what purposes a good, bad or evil? And that's

25
00:01:16,520 --> 00:01:17,319
Speaker 1: pretty scary.

26
00:01:17,640 --> 00:01:19,680
Speaker 3: Whose Amanda, does you do a great said, oh, not

27
00:01:19,720 --> 00:01:20,240
Speaker 3: as good as that.

28
00:01:21,720 --> 00:01:26,200
Speaker 4: So six people were taken out of the forest in bodybirgs.

29
00:01:26,400 --> 00:01:29,560
Speaker 2: Oh my god, Amanda Kella, that was a fantastic and

30
00:01:29,680 --> 00:01:31,520
Speaker 2: terrified chorin that was an AI.

31
00:01:31,720 --> 00:01:34,600
Speaker 4: But that is scary because when people won't know the context,

32
00:01:34,640 --> 00:01:36,639
Speaker 4: they will hear that on a podcast and think it's Sandra.

33
00:01:36,720 --> 00:01:39,360
Speaker 4: And it may be a benign circumstance in this one,

34
00:01:39,520 --> 00:01:40,559
Speaker 4: maybe not another time.

35
00:01:40,520 --> 00:01:43,919
Speaker 2: Well exactly. We've seen another really high profile similar story

36
00:01:43,920 --> 00:01:47,520
Speaker 2: in the news this week. Scarlet Johansen has accused Open Ai.

37
00:01:47,640 --> 00:01:50,360
Speaker 2: Now they're the tech company that owns and created chat GPT.

38
00:01:50,880 --> 00:01:54,000
Speaker 2: They've released a new voice. She's called Sky and she's

39
00:01:54,040 --> 00:01:57,560
Speaker 2: an AI bot assistant. Scar Jo says that she sounds

40
00:01:57,760 --> 00:01:59,880
Speaker 2: just like her. She's released a statement saying that it

41
00:02:00,120 --> 00:02:03,720
Speaker 2: sounds eerily similar to her and particularly interesting because she

42
00:02:03,920 --> 00:02:08,000
Speaker 2: turned down an offer from Open Ai to be that voice.

43
00:02:08,120 --> 00:02:11,120
Speaker 2: So we've got some grabs here. We've got the AI

44
00:02:11,280 --> 00:02:12,880
Speaker 2: voice and whether or not you think it sounds like

45
00:02:12,880 --> 00:02:18,080
Speaker 2: Scarlett Johansson and also playing me off Ryan blame Ryan

46
00:02:18,600 --> 00:02:21,560
Speaker 2: We've also got a grab of the movie Her, which

47
00:02:21,600 --> 00:02:24,840
Speaker 2: you might remember from twenty thirteen, and Scarlett Johansson voiced

48
00:02:24,880 --> 00:02:27,920
Speaker 2: an AI assistant in that movie, so a bit of

49
00:02:28,000 --> 00:02:30,720
Speaker 2: art imitates, like, so this first bit is the AI.

50
00:02:30,800 --> 00:02:32,760
Speaker 3: This is the II. I'm on stage right now, I'm

51
00:02:32,760 --> 00:02:34,959
Speaker 3: doing a live demo, and frankly, I'm feeling a little

52
00:02:34,960 --> 00:02:35,560
Speaker 3: bit nervous.

53
00:02:35,639 --> 00:02:38,880
Speaker 1: Can you help me calm my nerves a little bit. Oh,

54
00:02:38,960 --> 00:02:40,720
Speaker 1: you're doing a live demo right now.

55
00:02:41,040 --> 00:02:41,680
Speaker 4: That's awesome.

56
00:02:42,800 --> 00:02:47,080
Speaker 1: Just take a deep breath and remember you're the expert.

57
00:02:47,360 --> 00:02:49,760
Speaker 3: Now that sounds that sounds faint, I can't tell.

58
00:02:49,800 --> 00:02:52,080
Speaker 4: So here's the actual then, Scarlett Johansson.

59
00:02:51,760 --> 00:02:54,600
Speaker 2: Good morning theatre, Good morning, you have a meeting in

60
00:02:54,680 --> 00:02:55,239
Speaker 2: five minutes.

61
00:02:55,400 --> 00:03:00,000
Speaker 3: You want to try getting out of bed? Too funny.

62
00:03:00,000 --> 00:03:02,240
Speaker 4: So I wouldn't have said that identical, but she claims

63
00:03:02,240 --> 00:03:04,919
Speaker 4: that all her friends and family and in the industry

64
00:03:05,000 --> 00:03:06,079
Speaker 4: they think it sounds like her.

65
00:03:06,280 --> 00:03:06,480
Speaker 3: Yeah.

66
00:03:06,520 --> 00:03:09,960
Speaker 2: I think that stand alone, that the voice doesn't sound

67
00:03:10,200 --> 00:03:13,360
Speaker 2: exactly like her, But given that they tried to get

68
00:03:13,400 --> 00:03:16,520
Speaker 2: her to voice it, I think that's where the concern is.

69
00:03:16,639 --> 00:03:19,079
Speaker 2: I mean, it does sound different enough, but you can

70
00:03:19,080 --> 00:03:22,040
Speaker 2: hear that kind of warm raspy sort of Scarlet Johansson

71
00:03:22,080 --> 00:03:26,240
Speaker 2: tone open AI have paused the voice, so they've said, fine,

72
00:03:26,320 --> 00:03:29,040
Speaker 2: we won't run it because Scylet Johnson's threatened legal action.

73
00:03:29,360 --> 00:03:31,320
Speaker 2: But they also said that it's not her voice, it's

74
00:03:31,320 --> 00:03:33,680
Speaker 2: not supposed to be her voice. That they put out

75
00:03:33,720 --> 00:03:37,680
Speaker 2: an open casting call to actors months before they reached

76
00:03:37,720 --> 00:03:41,520
Speaker 2: out to Sculet Johansson allegedly, who knows if that's true.

77
00:03:41,680 --> 00:03:44,680
Speaker 2: But funnily enough, in that open casting call they specifically

78
00:03:44,840 --> 00:03:47,360
Speaker 2: called for actors who were not in the union, not

79
00:03:47,480 --> 00:03:50,920
Speaker 2: in sag AFTRA, because they can exploit them, and went

80
00:03:50,960 --> 00:03:53,880
Speaker 2: on strike really hard last year for riots for voice

81
00:03:53,920 --> 00:03:54,680
Speaker 2: actors and AI.

82
00:03:54,920 --> 00:03:56,800
Speaker 3: They could be mimicking us. You wouldn't know.

83
00:03:56,960 --> 00:03:58,800
Speaker 4: I saw this great thing the other day about AI.

84
00:03:58,840 --> 00:04:00,880
Speaker 4: They say, you know the biggest problem pushing all things

85
00:04:00,920 --> 00:04:04,240
Speaker 4: AI wrong direction. I want AI to do my laundry

86
00:04:04,240 --> 00:04:06,640
Speaker 4: and dishes so I can do art and writing. Not

87
00:04:06,640 --> 00:04:08,200
Speaker 4: for AI to do my art and writing so I

88
00:04:08,200 --> 00:04:10,000
Speaker 4: can do laundry and dishes and sot on an instant.

89
00:04:10,240 --> 00:04:12,680
Speaker 4: Humans doing the hard jobs on minimum wage while the

90
00:04:12,800 --> 00:04:15,640
Speaker 4: robots write poetry and paint is not the future I want.

91
00:04:15,760 --> 00:04:16,760
Speaker 3: That's not the future.

92
00:04:17,120 --> 00:04:20,200
Speaker 2: Point well, thinking about the future scarily enough. You know,

93
00:04:20,279 --> 00:04:22,400
Speaker 2: next year we've got a federal election due and the

94
00:04:22,440 --> 00:04:25,560
Speaker 2: Australian Electoral Commission even came out last week and they

95
00:04:25,640 --> 00:04:27,960
Speaker 2: warned that ahead of the election, we're going to see

96
00:04:28,000 --> 00:04:31,680
Speaker 2: a huge rise of AI voices and deep fakes in

97
00:04:31,720 --> 00:04:34,960
Speaker 2: the political space, and that they don't have any power

98
00:04:35,320 --> 00:04:38,000
Speaker 2: to do anything about it. There aren't laws in Australia

99
00:04:38,000 --> 00:04:41,880
Speaker 2: that are strong enough to prevent these kinds of deep fakes.

100
00:04:41,600 --> 00:04:45,120
Speaker 4: Or laws anywhere. So America's having an election, India's having

101
00:04:45,160 --> 00:04:48,440
Speaker 4: an election, Briton's having an election, Australia's having an election

102
00:04:48,560 --> 00:04:53,200
Speaker 4: next year, and they're expecting AI deep fakes, misinformation everywhere.

103
00:04:53,320 --> 00:04:55,960
Speaker 2: Well, there was that call in the US, a voicemail

104
00:04:56,080 --> 00:04:58,400
Speaker 2: thing that was Joe Biden calling people telling them how

105
00:04:58,440 --> 00:04:59,960
Speaker 2: to vote, and that was a deep fake. We've already

106
00:05:00,000 --> 00:05:02,960
Speaker 2: we've seen lots of examples of how it can impact politics,

107
00:05:03,000 --> 00:05:05,880
Speaker 2: but now that that's crossing over into pop culture and arts,

108
00:05:05,960 --> 00:05:06,880
Speaker 2: you know who knows where.

109
00:05:07,080 --> 00:05:09,520
Speaker 3: Don't believe everything you hear. So for example, you hear this,

110
00:05:11,520 --> 00:05:12,640
Speaker 3: you know that that's fake.

111
00:05:14,120 --> 00:05:16,520
Speaker 4: I can smell it.

112
00:05:16,680 --> 00:05:20,200
Speaker 3: Fascinating M thank you, thank you. Check out M editor

113
00:05:20,200 --> 00:05:21,360
Speaker 3: of the jaily Oz,