1
00:00:00,000 --> 00:00:15,840
[Music]

2
00:00:15,840 --> 00:00:18,320
Welcome back to Quietly Secure.

3
00:00:18,320 --> 00:00:22,400
In the last episode, we explored the psychology behind scams.

4
00:00:22,400 --> 00:00:26,560
We looked at why certain messages are so effective.

5
00:00:27,280 --> 00:00:31,760
Urgency, authority, familiarity, and emotion.

6
00:00:31,760 --> 00:00:34,400
And we discovered something important.

7
00:00:34,400 --> 00:00:38,880
Many successful attacks don't begin with technology.

8
00:00:38,880 --> 00:00:40,160
They begin with trust.

9
00:00:40,160 --> 00:00:46,320
Today we're going to explore what happens when technology itself starts challenging

10
00:00:46,320 --> 00:00:51,280
one of the oldest ways humans decide what is real, our senses.

11
00:00:51,280 --> 00:00:56,240
For most of human history, seeing something meant believing it.

12
00:00:56,960 --> 00:01:00,160
Hearing someone's voice meant they were probably there.

13
00:01:00,160 --> 00:01:03,120
A photograph was evidence that something happened.

14
00:01:03,120 --> 00:01:06,400
A video is proof that something was recorded.

15
00:01:06,400 --> 00:01:10,160
But technology has changed that relationship.

16
00:01:10,160 --> 00:01:12,960
Today images can be created.

17
00:01:12,960 --> 00:01:15,200
Voices can be copied.

18
00:01:15,200 --> 00:01:17,760
Faces can be generated.

19
00:01:17,760 --> 00:01:22,000
And videos can show people doing things they never actually did.

20
00:01:22,000 --> 00:01:26,240
This is the world of deep fakes and synthetic media.

21
00:01:26,800 --> 00:01:30,480
Unlike many technologies we've discussed throughout this podcast.

22
00:01:30,480 --> 00:01:37,200
The reality is both more interesting and more complicated than the headlines suggest.

23
00:01:37,200 --> 00:01:40,960
Let's start with the basics.

24
00:01:40,960 --> 00:01:47,520
Synthetic media is content created or modified using artificial intelligence.

25
00:01:47,520 --> 00:01:51,280
This can include images, videos, voices, music, text.

26
00:01:52,160 --> 00:01:58,480
Some examples are completely harmless, like the advertising graphics I do for the podcast.

27
00:01:58,480 --> 00:02:01,920
A filmmaker created a digital character.

28
00:02:01,920 --> 00:02:05,920
A game, studio, create in realistic environments.

29
00:02:05,920 --> 00:02:12,480
A company translating a video into another language while keeping the original person's voice.

30
00:02:12,480 --> 00:02:15,760
These are creative uses of the technology.

31
00:02:15,760 --> 00:02:20,800
The challenge begins when the same tools are used to mislead.

32
00:02:21,920 --> 00:02:24,960
This is where deep fakes enter the conversation.

33
00:02:24,960 --> 00:02:30,240
Throughout history people have manipulated information.

34
00:02:30,240 --> 00:02:35,520
Photographs have been edited. Audio recordings have been altered.

35
00:02:35,520 --> 00:02:37,920
Videos have been taken out of context.

36
00:02:37,920 --> 00:02:40,880
So what makes deep fakes different?

37
00:02:40,880 --> 00:02:44,320
The difference is scale and accessibility.

38
00:02:44,320 --> 00:02:50,400
Previously creating fake media required specialist equipment, expertise,

39
00:02:50,400 --> 00:02:51,920
and significant time.

40
00:02:51,920 --> 00:02:55,920
Today many of these tools are becoming easier to use.

41
00:02:55,920 --> 00:02:59,200
The barrier has well and truly lowered.

42
00:02:59,200 --> 00:03:02,320
This doesn't mean every video online is fake.

43
00:03:02,320 --> 00:03:05,200
It doesn't mean we can no longer trust anything we see.

44
00:03:05,200 --> 00:03:12,400
But it does mean we need to become slightly more thoughtful about how we verify information.

45
00:03:12,400 --> 00:03:15,840
Humans are highly visual.

46
00:03:15,840 --> 00:03:17,840
We naturally believe what we see.

47
00:03:18,720 --> 00:03:23,680
A picture creates an emotional reaction faster than a written explanation.

48
00:03:23,680 --> 00:03:26,800
A video feels more convincing.

49
00:03:26,800 --> 00:03:30,800
Seeing a person speak creates a sense of connection.

50
00:03:30,800 --> 00:03:35,600
We see facial expression, body language, tone.

51
00:03:35,600 --> 00:03:40,080
All of these signals help our brain decide is this person real.

52
00:03:40,080 --> 00:03:45,280
Deep fakes work because they exploit some of their signals.

53
00:03:45,920 --> 00:03:51,840
They don't just create information, they create familiarity, they create confidence,

54
00:03:51,840 --> 00:03:55,440
they create the feeling that something is genuine.

55
00:03:55,440 --> 00:04:00,560
Many early deep fake examples focused on celebrities.

56
00:04:00,560 --> 00:04:03,440
Actors appearing to say things they never said.

57
00:04:03,440 --> 00:04:06,800
Public figures appearing in fake videos.

58
00:04:06,800 --> 00:04:11,200
This attracted attention because the subjects were recognisable.

59
00:04:11,200 --> 00:04:15,520
But the bigger concern isn't necessarily famous people.

60
00:04:15,520 --> 00:04:16,800
It's ordinary people.

61
00:04:16,800 --> 00:04:21,200
Imagine receiving a video message from someone you know,

62
00:04:21,200 --> 00:04:26,960
a family member, a colleague, a manager, the voice sounds right, the first looks right,

63
00:04:26,960 --> 00:04:28,800
the request seems believable.

64
00:04:28,800 --> 00:04:32,480
This is where technology becomes more concerning.

65
00:04:32,480 --> 00:04:39,200
Not because the technology is magical, but because humans naturally trust familiar people.

66
00:04:39,200 --> 00:04:43,920
Which brings us back to something we discussed last episode.

67
00:04:44,560 --> 00:04:46,880
Familiarity is powerful.

68
00:04:46,880 --> 00:04:52,000
One of the areas developing quickly is voice cloning.

69
00:04:52,000 --> 00:04:56,000
A person's voice contains many unique characteristics.

70
00:04:56,000 --> 00:05:00,400
Their accent, their rhythm, their expression, their tone.

71
00:05:00,400 --> 00:05:08,720
AI systems can now analyse recordings of voices and generate new speech that sounds remarkably similar.

72
00:05:08,720 --> 00:05:12,400
This creates possibilities for useful applications.

73
00:05:12,960 --> 00:05:17,760
Accessibility, translation, helping people who've lost their voice.

74
00:05:17,760 --> 00:05:20,880
But it can also create risk.

75
00:05:20,880 --> 00:05:25,040
A scammer doesn't need to perfectly recreate someone's voice forever.

76
00:05:25,040 --> 00:05:31,280
They only need to create enough uncertainty, enough realism, enough urgency.

77
00:05:31,280 --> 00:05:36,720
The same psychological principles we discussed before still apply.

78
00:05:36,720 --> 00:05:41,440
The technology changes, the manipulation does not.

79
00:05:42,640 --> 00:05:49,360
The obvious question is, if AI can create realistic content, how do we know what is real?

80
00:05:49,360 --> 00:05:52,560
The answer is not one single trick.

81
00:05:52,560 --> 00:05:54,720
There is no magic rule.

82
00:05:54,720 --> 00:05:59,200
Instead, we go back to good digital habits.

83
00:05:59,200 --> 00:06:01,120
Consider the source.

84
00:06:01,120 --> 00:06:03,280
Where did this come from?

85
00:06:03,280 --> 00:06:04,560
Who published it?

86
00:06:04,560 --> 00:06:07,760
Is there another independent source confirming it?

87
00:06:07,760 --> 00:06:10,640
Does this situation make sense?

88
00:06:11,360 --> 00:06:16,480
Ask whether the content is trying to create an immediate emotional response

89
00:06:16,480 --> 00:06:19,360
like fear and shock excitement?

90
00:06:19,360 --> 00:06:24,240
Because emotional reactions are exactly what misinformation relies on.

91
00:06:24,240 --> 00:06:30,560
One of the biggest lessons of the internet age is that information rarely exists alone.

92
00:06:30,560 --> 00:06:34,080
A screenshot without context can mislead.

93
00:06:34,080 --> 00:06:37,440
A quote without context can mislead.

94
00:06:37,440 --> 00:06:40,720
A video without context can mislead.

95
00:06:41,680 --> 00:06:47,760
Even real content can create false impressions when removed from its original situation.

96
00:06:47,760 --> 00:06:51,600
Deep facts are a new version of an old problem.

97
00:06:51,600 --> 00:06:56,000
The challenge isn't simply identifying fake things.

98
00:06:56,000 --> 00:06:59,200
It's understanding the bigger picture.

99
00:06:59,200 --> 00:07:05,040
This is where the conversation about AI often becomes pessimistic.

100
00:07:05,040 --> 00:07:10,000
People ask if everything can be faked, how can we believe anything?

101
00:07:10,720 --> 00:07:13,040
But history suggests something different.

102
00:07:13,040 --> 00:07:18,320
Every major communication technology has created similar concerns.

103
00:07:18,320 --> 00:07:22,480
Photography, radio, television, the internet.

104
00:07:22,480 --> 00:07:25,520
Each one changed how information spread.

105
00:07:25,520 --> 00:07:28,320
Each one created new challenges.

106
00:07:28,320 --> 00:07:30,720
But society adapted.

107
00:07:30,720 --> 00:07:34,320
The answer was never to stop using technology.

108
00:07:35,280 --> 00:07:40,240
It was to develop better ways of understanding and verifying information.

109
00:07:40,240 --> 00:07:44,960
The girl isn't to become suspicious of every video.

110
00:07:44,960 --> 00:07:46,800
That will be exhausting.

111
00:07:46,800 --> 00:07:50,960
The girl is to become comfortable asking better questions.

112
00:07:50,960 --> 00:07:54,320
Not is this definitely fake.

113
00:07:54,320 --> 00:07:56,560
But what evidence supports this?

114
00:07:56,560 --> 00:07:58,800
Where did it come from?

115
00:07:58,800 --> 00:08:00,880
Who benefits if I believe in it?

116
00:08:00,880 --> 00:08:04,800
Those questions are useful, whether AI exists.

117
00:08:05,200 --> 00:08:05,760
or not.

118
00:08:05,760 --> 00:08:10,400
Deep facts are often described as a technology problem.

119
00:08:10,400 --> 00:08:13,120
But they are really a trust problem.

120
00:08:13,120 --> 00:08:16,480
The technology makes creation easier.

121
00:08:16,480 --> 00:08:20,480
But human behaviour determines whether it succeeds.

122
00:08:20,480 --> 00:08:25,760
We trust familiar voices, we trust confident appearances,

123
00:08:25,760 --> 00:08:27,760
and we trust things that fail real.

124
00:08:27,760 --> 00:08:31,840
Understanding that doesn't mean becoming cynical.

125
00:08:31,840 --> 00:08:33,680
It means becoming more aware.

126
00:08:34,240 --> 00:08:37,120
The future will contain more synthetic content.

127
00:08:37,120 --> 00:08:39,360
That's an almost certainty.

128
00:08:39,360 --> 00:08:45,120
But a world with more artificial content does not have to become a world without trust.

129
00:08:45,120 --> 00:08:51,760
It simply means trust will need to be supported by something stronger than the appearance

130
00:08:51,760 --> 00:08:52,240
alone.

131
00:08:52,240 --> 00:08:56,800
Like evidence, context, and thoughtful judgment.

132
00:08:56,800 --> 00:09:00,880
Thank you for listening to Quietly Secure.

133
00:09:00,880 --> 00:09:05,360
Next time we'll explore one of the biggest questions of the modern technology age.

134
00:09:05,360 --> 00:09:07,840
Can you trust AI answers?

135
00:09:07,840 --> 00:09:13,120
Because if artificial intelligence can create images, voices, and videos,

136
00:09:13,120 --> 00:09:16,560
what happens when it starts answering our questions?

137
00:09:16,560 --> 00:09:24,240
Until then, stay curious, stay calm, and as always, stay Quietly Secure.

138
00:09:24,400 --> 00:09:34,640
[Music]

139
00:09:34,640 --> 00:09:36,640
[music]