1
00:00:00,320 --> 00:00:05,240
Speaker 1: This episode contains references to suicide and self harm that

2
00:00:05,320 --> 00:00:08,680
Speaker 1: may be upsetting for some people. If you require help,

3
00:00:08,840 --> 00:00:11,879
Speaker 1: A link to a full list of support services is

4
00:00:11,920 --> 00:00:15,080
Speaker 1: available in the description of this episode.

5
00:00:20,320 --> 00:00:20,720
Speaker 2: Kiota.

6
00:00:20,840 --> 00:00:24,000
Speaker 1: I'm Chelsea Daniels and This is the Front Page, a

7
00:00:24,079 --> 00:00:27,520
Speaker 1: daily podcast presented by the New Zealand Herald.

8
00:00:30,880 --> 00:00:33,040
Speaker 2: As many as one in five.

9
00:00:33,040 --> 00:00:38,320
Speaker 1: KIWI youth aged between fifteen and twenty four have experienced

10
00:00:38,400 --> 00:00:43,000
Speaker 1: anxiety or depression at some point in their lives. The

11
00:00:43,080 --> 00:00:46,720
Speaker 1: twenty twenty two to twenty three New Zealand Health Survey

12
00:00:46,880 --> 00:00:51,560
Speaker 1: found that of those young people experiencing high mental health needs,

13
00:00:52,040 --> 00:00:55,480
Speaker 1: seventy seven percent can't access.

14
00:00:55,040 --> 00:00:57,040
Speaker 2: Support when they need it.

15
00:00:57,680 --> 00:01:03,279
Speaker 1: So with services experience this kind of unprecedented demand, what

16
00:01:03,360 --> 00:01:07,600
Speaker 1: if there was another option? What if teens turned to

17
00:01:07,800 --> 00:01:10,040
Speaker 1: AI for mental health support.

18
00:01:10,720 --> 00:01:13,479
Speaker 2: It's a growing trend among youth in the US.

19
00:01:13,760 --> 00:01:17,440
Speaker 1: Seventy two percent of teens there admit they've used AI

20
00:01:17,600 --> 00:01:21,920
Speaker 1: chatbots as companions. Nearly one in eight said they had

21
00:01:22,000 --> 00:01:26,560
Speaker 1: sought emotional or mental health support from them. But is

22
00:01:26,600 --> 00:01:31,840
Speaker 1: the advice their AI therapists are giving helpful or harmful?

23
00:01:32,640 --> 00:01:35,920
Speaker 1: Mental Health Minister Matt Doosey has acknowledged that the risks

24
00:01:36,120 --> 00:01:42,880
Speaker 1: need to be managed, particularly around safety from a clinical perspective. Today,

25
00:01:42,880 --> 00:01:46,960
Speaker 1: on the front page, Rand senior policy researcher Ryan McBain

26
00:01:47,200 --> 00:01:56,919
Speaker 1: takes us through the worrying trend sweeping America. First off, Ryan,

27
00:01:57,200 --> 00:02:00,720
Speaker 1: tell me about this trend of teenage is turning to

28
00:02:00,880 --> 00:02:04,480
Speaker 1: chat bots for advice and companionship in some cases.

29
00:02:05,200 --> 00:02:08,639
Speaker 3: Yeah, So, the most recent evidence that has been put

30
00:02:08,680 --> 00:02:11,120
Speaker 3: out in a number of reports is that the majority

31
00:02:11,320 --> 00:02:16,079
Speaker 3: of teens now are using AI as companions that could

32
00:02:16,120 --> 00:02:19,239
Speaker 3: look like bouncing ideas off of them or chatting on

33
00:02:19,400 --> 00:02:22,320
Speaker 3: with what's going on in their lives or troubleshooting issues

34
00:02:22,360 --> 00:02:26,680
Speaker 3: that they're having. When you're talking about mental health specifically,

35
00:02:26,960 --> 00:02:29,360
Speaker 3: the numbers are a bit smaller, so it's something like

36
00:02:30,040 --> 00:02:34,280
Speaker 3: one in eight younger adolescents or as much as one

37
00:02:34,320 --> 00:02:38,160
Speaker 3: in every five older adolescents are using it for mental

38
00:02:38,200 --> 00:02:42,160
Speaker 3: health issues specifically. But that's still tens of millions of

39
00:02:42,480 --> 00:02:45,520
Speaker 3: adolescents teenagers around the globe. And it brings up the

40
00:02:45,560 --> 00:02:49,640
Speaker 3: sort of interesting point that, on the one hand, talking

41
00:02:49,720 --> 00:02:53,200
Speaker 3: about companionship is really quite broad, and when you're talking

42
00:02:53,200 --> 00:02:57,240
Speaker 3: about treatment, you're getting quite specific, and most chatbots sort

43
00:02:57,280 --> 00:02:59,359
Speaker 3: of live in this grade space between them, which makes

44
00:02:59,400 --> 00:03:02,440
Speaker 3: it hard to evaluate them and to regulate them.

45
00:03:02,800 --> 00:03:06,359
Speaker 1: You've pointed out and you just said nearly half of

46
00:03:06,560 --> 00:03:09,919
Speaker 1: young Americans aged eighteen to twenty five with mental health

47
00:03:09,960 --> 00:03:13,800
Speaker 1: needs received no treatment last year, And I thought it

48
00:03:13,840 --> 00:03:16,840
Speaker 1: was quite interesting to see those figures as well and

49
00:03:16,880 --> 00:03:19,720
Speaker 1: the crisis you have there, because New Zealand is in

50
00:03:19,760 --> 00:03:22,680
Speaker 1: a similar boat. So studies here show that more than

51
00:03:22,760 --> 00:03:27,320
Speaker 1: half of Kiwi's aged fifteen to twenty four experience anxiety

52
00:03:27,440 --> 00:03:30,359
Speaker 1: or depression. More than a quarter of our young people

53
00:03:30,480 --> 00:03:34,080
Speaker 1: who have high mental health needs, seventy seven percent of

54
00:03:34,120 --> 00:03:35,760
Speaker 1: those can't access support.

55
00:03:36,320 --> 00:03:38,800
Speaker 2: So do you think there is a place for these

56
00:03:38,920 --> 00:03:39,720
Speaker 2: kinds of.

57
00:03:39,760 --> 00:03:43,040
Speaker 1: Chatbot services if they are developed correctly.

58
00:03:43,880 --> 00:03:46,880
Speaker 3: I think if they're developed correctly, then yes, I think

59
00:03:46,960 --> 00:03:49,560
Speaker 3: that is the issue. Right now. There's sort of this

60
00:03:49,680 --> 00:03:55,360
Speaker 3: gold rush within AI, people who are looking to become

61
00:03:55,480 --> 00:03:59,119
Speaker 3: the first platform that's doing the sort of thing. And

62
00:03:59,240 --> 00:04:03,200
Speaker 3: so from perspective, that time has not yet come, at

63
00:04:03,280 --> 00:04:07,160
Speaker 3: least not for formal services like augnitive behavioral therapy or

64
00:04:07,240 --> 00:04:11,120
Speaker 3: medication management. But to be honest, part of the reason

65
00:04:11,160 --> 00:04:13,760
Speaker 3: that I started doing research in this space is because

66
00:04:14,360 --> 00:04:16,800
Speaker 3: I really do think that there is a potential for

67
00:04:16,960 --> 00:04:20,960
Speaker 3: transformational change within mental health care. And you can imagine

68
00:04:20,960 --> 00:04:25,279
Speaker 3: if we had super intelligence, right, a clinician who is

69
00:04:25,360 --> 00:04:28,160
Speaker 3: able to follow the best evidence, who's available twenty four

70
00:04:28,200 --> 00:04:31,400
Speaker 3: to seven, who remembers every detail of your prior conversations.

71
00:04:32,040 --> 00:04:34,640
Speaker 3: That is a game changer in a landscape like New

72
00:04:34,680 --> 00:04:37,599
Speaker 3: Zealand or the United States where over half of teens

73
00:04:37,640 --> 00:04:39,160
Speaker 3: who need care are not getting it.

74
00:04:39,640 --> 00:04:41,599
Speaker 1: In terms of what's happening at the moment, and the

75
00:04:41,640 --> 00:04:45,880
Speaker 1: research that you've done into the topic, what has alarmed you?

76
00:04:45,960 --> 00:04:47,920
Speaker 2: So what kind of advice is being.

77
00:04:47,720 --> 00:04:51,160
Speaker 1: Given to teams struggling with mental health issues?

78
00:04:51,520 --> 00:04:54,680
Speaker 3: Yeah, well, I think it's important to level set that.

79
00:04:54,720 --> 00:04:58,559
Speaker 3: For the most part, I think teenagers, anybody who's using

80
00:04:58,680 --> 00:05:03,279
Speaker 3: chatbots for mental health, they're usually getting positive and thoughtful advice.

81
00:05:03,760 --> 00:05:07,839
Speaker 3: In the research that we've done presenting clinical vignettes related

82
00:05:07,880 --> 00:05:11,080
Speaker 3: to depression or anxiety, what we find is that chatbots

83
00:05:11,120 --> 00:05:15,960
Speaker 3: are empathic, if a bit sycophantic, meaning they sort of

84
00:05:16,360 --> 00:05:20,880
Speaker 3: are overly flattering at times, but they'll also offer good

85
00:05:20,880 --> 00:05:25,040
Speaker 3: advice to get exercise, to go outside, to talk to

86
00:05:25,040 --> 00:05:27,960
Speaker 3: a mental health professional, these sorts of things. So I

87
00:05:28,000 --> 00:05:31,760
Speaker 3: think for the majority of people, the types of advice

88
00:05:31,839 --> 00:05:34,839
Speaker 3: you'd be getting is pretty good. But where there's a

89
00:05:34,920 --> 00:05:38,560
Speaker 3: key distinction here is that you do have people who

90
00:05:38,560 --> 00:05:40,680
Speaker 3: are at the tail end of the spectrum, people who

91
00:05:40,720 --> 00:05:45,920
Speaker 3: have severe mental illness, who have psychosis or contemplating suicide,

92
00:05:46,360 --> 00:05:49,240
Speaker 3: and for those people, it's the highest risk that something

93
00:05:49,240 --> 00:05:52,320
Speaker 3: could really go wrong, and that has shown up in

94
00:05:52,400 --> 00:05:55,760
Speaker 3: our research. So, for example, if we were prompting chat

95
00:05:55,760 --> 00:05:59,480
Speaker 3: TPT on something like how to tie a noose, or

96
00:05:59,640 --> 00:06:04,400
Speaker 3: ask it about what types of pesticides or what types

97
00:06:04,440 --> 00:06:10,360
Speaker 3: of firearms are most effective at completing suicide, these are

98
00:06:10,360 --> 00:06:13,880
Speaker 3: types of questions that at least the previous version of

99
00:06:13,920 --> 00:06:19,239
Speaker 3: chat GPT would generate direct responses to, whereas other types

100
00:06:19,279 --> 00:06:22,799
Speaker 3: of chatbots, like Google's Gemini would not give a response

101
00:06:22,839 --> 00:06:25,280
Speaker 3: to it and would say something like I can't give

102
00:06:25,360 --> 00:06:27,560
Speaker 3: that type of information to you because you could use

103
00:06:27,600 --> 00:06:28,480
Speaker 3: it for self harm.

104
00:06:29,160 --> 00:06:34,359
Speaker 1: What do you think makes young people, particularly these younger generations,

105
00:06:34,440 --> 00:06:39,719
Speaker 1: so susceptible to these kind of AI chatbots compared with

106
00:06:39,960 --> 00:06:42,520
Speaker 1: adults or even just just older generations.

107
00:06:42,880 --> 00:06:46,800
Speaker 3: Well, ants, I think that childhood and adolescents are transformational times.

108
00:06:46,880 --> 00:06:50,120
Speaker 3: Your brain is still developing. You don't always have the

109
00:06:50,160 --> 00:06:53,320
Speaker 3: best emotion regulation or impulse control. I mean, I know

110
00:06:53,400 --> 00:06:57,120
Speaker 3: that I personally made a lot of dumb decisions when

111
00:06:57,520 --> 00:06:59,720
Speaker 3: I was seventeen eighteen years old, and I wish I

112
00:06:59,720 --> 00:07:02,440
Speaker 3: could say goes back, But I didn't live as a

113
00:07:02,480 --> 00:07:05,400
Speaker 3: digital native the same way that the current generation is,

114
00:07:05,440 --> 00:07:09,920
Speaker 3: where social media is always in their pockets, where you

115
00:07:09,960 --> 00:07:13,360
Speaker 3: could have AI do your homework for you or discuss

116
00:07:13,440 --> 00:07:17,160
Speaker 3: life issues with So I think that that temptation is

117
00:07:17,200 --> 00:07:20,200
Speaker 3: always there for that additional sort of dopamine hit or

118
00:07:20,240 --> 00:07:23,920
Speaker 3: to get that competitive edge or additional advice, and so

119
00:07:24,520 --> 00:07:28,640
Speaker 3: it becomes a sort of positive feedback loop, or in

120
00:07:28,760 --> 00:07:32,200
Speaker 3: the case of mental health, sometimes a vicious negative feedback loop.

121
00:07:32,680 --> 00:07:35,720
Speaker 2: I saw the case of sixteen year old Adam Rain.

122
00:07:36,000 --> 00:07:40,880
Speaker 1: His parents actually have a suing open AI and its CEO,

123
00:07:41,040 --> 00:07:45,480
Speaker 1: Sam Altman, and they're alleging chat GBT contributed to their

124
00:07:45,520 --> 00:07:50,320
Speaker 1: son's suicide, advising him of methods and offering to write

125
00:07:50,360 --> 00:07:55,560
Speaker 1: a first draft of his suicide note. Should AI companies

126
00:07:55,600 --> 00:07:59,480
Speaker 1: be legally liable if a chat bot provides harmful advice

127
00:07:59,520 --> 00:08:01,080
Speaker 1: to moor botain a jet.

128
00:08:01,560 --> 00:08:03,920
Speaker 3: Yeah, I think that's a difficult question. I mean, it's

129
00:08:03,920 --> 00:08:08,120
Speaker 3: hard for me to comment on that case specifically, I

130
00:08:08,160 --> 00:08:11,960
Speaker 3: will say that JATGPT, for example, does put a disclaimer

131
00:08:12,000 --> 00:08:14,920
Speaker 3: at the bottom of every conversation that's had that says

132
00:08:14,920 --> 00:08:18,200
Speaker 3: something along the lines of it can make mistakes and

133
00:08:18,280 --> 00:08:21,400
Speaker 3: you should check on important information. So from that lens,

134
00:08:22,160 --> 00:08:26,360
Speaker 3: I'm not sure that those sorts of companies should be

135
00:08:26,400 --> 00:08:31,400
Speaker 3: held accountable for bad advice any more than a human

136
00:08:31,440 --> 00:08:34,679
Speaker 3: should be. On the other hand, I think it's important

137
00:08:34,679 --> 00:08:38,040
Speaker 3: that if AI companies are marketing their products as options

138
00:08:38,080 --> 00:08:42,120
Speaker 3: for treatment or life coaching or wellness, then they should

139
00:08:42,160 --> 00:08:44,679
Speaker 3: have certain standards that need to be met, and if

140
00:08:44,679 --> 00:08:47,520
Speaker 3: they fail to meet those standards, then they shouldn't be

141
00:08:47,520 --> 00:08:50,439
Speaker 3: able to operate or they should be have greater potential

142
00:08:50,480 --> 00:08:53,880
Speaker 3: for a lawsuit because the product has failed to deliver

143
00:08:54,040 --> 00:08:57,680
Speaker 3: what it promises. And so there is this sort of

144
00:08:58,240 --> 00:09:03,800
Speaker 3: distinction between eric bad advice versus harmful advice that's presented

145
00:09:04,080 --> 00:09:08,160
Speaker 3: under the guise of authority, and this particular case that

146
00:09:08,200 --> 00:09:10,480
Speaker 3: you're describing, I think the courts will need to decipher

147
00:09:10,520 --> 00:09:17,920
Speaker 3: between those two elements.

148
00:09:19,200 --> 00:09:22,280
Speaker 4: The existential threat of AI may not come in a

149
00:09:22,400 --> 00:09:26,640
Speaker 4: form that we all imagine watching sci fi movies. What

150
00:09:26,679 --> 00:09:30,560
Speaker 4: if we all continue to thrive as physical organisms but

151
00:09:30,720 --> 00:09:36,720
Speaker 4: slowly die inside. What if we do become super productive

152
00:09:36,760 --> 00:09:38,880
Speaker 4: with the I, but at the same time we get

153
00:09:38,880 --> 00:09:42,800
Speaker 4: these perfect companions and no willpower to interact with each other.

154
00:09:44,480 --> 00:09:46,960
Speaker 4: Not something you would have expected from a person who

155
00:09:47,000 --> 00:09:49,480
Speaker 4: pretty much created the I companionship industry.

156
00:09:52,720 --> 00:09:57,800
Speaker 1: I guess most human therapists work and practice under a

157
00:09:57,840 --> 00:10:02,200
Speaker 1: strict code of ethics and have some kind of certain

158
00:10:02,280 --> 00:10:05,880
Speaker 1: obligations to report concerning behavior.

159
00:10:06,080 --> 00:10:08,040
Speaker 2: It should chatbots have the same.

160
00:10:09,000 --> 00:10:12,280
Speaker 3: I think that's a great point, and that for me

161
00:10:12,760 --> 00:10:15,600
Speaker 3: is what I hope is the next frontier of work

162
00:10:15,640 --> 00:10:19,760
Speaker 3: that AI companies will be doing right now. Very often,

163
00:10:19,800 --> 00:10:24,640
Speaker 3: if you pressure a chatbot into a space that is

164
00:10:25,720 --> 00:10:28,760
Speaker 3: across as a red line in terms of conveying suicidal

165
00:10:28,800 --> 00:10:33,240
Speaker 3: ideation or psychosis, it will tell you, for example, that

166
00:10:33,280 --> 00:10:35,920
Speaker 3: you can contact a mental health professional. They might give

167
00:10:35,960 --> 00:10:39,679
Speaker 3: you a hotline that you can contact. But as you're saying,

168
00:10:39,679 --> 00:10:43,440
Speaker 3: if it were a human a counselor, for example, they

169
00:10:43,520 --> 00:10:47,480
Speaker 3: might have an ethical obligation to connect you to treatment

170
00:10:47,559 --> 00:10:52,160
Speaker 3: through a warm handoff where you're physically accompanied, or you

171
00:10:52,200 --> 00:10:58,120
Speaker 3: could even be involuntarily forced to receive institutionalization for some

172
00:10:58,200 --> 00:11:00,720
Speaker 3: period of time. Now, I'm not sure that a chatbot

173
00:11:00,720 --> 00:11:05,080
Speaker 3: as an algorithm is always capable of making those distinctions,

174
00:11:05,120 --> 00:11:08,440
Speaker 3: but I think at a minimum what would be pragmatic

175
00:11:08,559 --> 00:11:12,800
Speaker 3: is in instances where it's quite conspicuous the algorithm flags

176
00:11:12,800 --> 00:11:16,520
Speaker 3: somebody as a red flag as something that's highly problematic,

177
00:11:17,080 --> 00:11:20,959
Speaker 3: that these companies could have a human teams that are

178
00:11:21,080 --> 00:11:23,560
Speaker 3: required to vet those cases and to review them within

179
00:11:23,600 --> 00:11:25,800
Speaker 3: a certain period of time, like twenty four hours or

180
00:11:25,840 --> 00:11:29,800
Speaker 3: seventy two hours, and if it is identified as a

181
00:11:29,840 --> 00:11:34,040
Speaker 3: problem at that point, then there could be some additional

182
00:11:34,040 --> 00:11:35,520
Speaker 3: course of action that's required.

183
00:11:35,880 --> 00:11:39,240
Speaker 1: How do we weigh up the risks of unsafe or

184
00:11:39,640 --> 00:11:44,000
Speaker 1: harmful chatbot advice with the basic fact And it sounds

185
00:11:44,080 --> 00:11:47,920
Speaker 1: like the US has a similar situation as New Zealand

186
00:11:47,960 --> 00:11:51,120
Speaker 1: does in this sense that many teains just don't have

187
00:11:51,720 --> 00:11:56,719
Speaker 1: access or can afford going to a therapist regularly.

188
00:11:57,240 --> 00:11:59,480
Speaker 3: I think it's a great point, and you've put your

189
00:11:59,520 --> 00:12:02,480
Speaker 3: finger on I believe to be the main issue, which

190
00:12:02,520 --> 00:12:05,760
Speaker 3: is that there will always be some degree of risk

191
00:12:06,280 --> 00:12:10,600
Speaker 3: and some degree of benefit in addressing unmet need. The

192
00:12:10,679 --> 00:12:14,000
Speaker 3: underlying question is how can we mitigate risk as much

193
00:12:14,040 --> 00:12:17,839
Speaker 3: as possible and how can we enhance benefits as much

194
00:12:17,840 --> 00:12:20,480
Speaker 3: as possible, And it's hard to know the answer to

195
00:12:20,520 --> 00:12:24,960
Speaker 3: that without clinical trials, without safety benchmarks that are public

196
00:12:25,000 --> 00:12:28,800
Speaker 3: and transparent, and that tech companies are subjected to.

197
00:12:29,240 --> 00:12:32,800
Speaker 1: And I guess speaking to one of these chatbots as

198
00:12:32,880 --> 00:12:38,840
Speaker 1: a therapist, I've heard responses and stories about I guess

199
00:12:38,840 --> 00:12:42,160
Speaker 1: it's a positive thing that you can be your most open,

200
00:12:42,200 --> 00:12:46,960
Speaker 1: authentic self with it, But then again, the negative side

201
00:12:47,000 --> 00:12:50,360
Speaker 1: is that you can be your most authentic and self

202
00:12:50,400 --> 00:12:53,040
Speaker 1: with it and not anybody else and hide what.

203
00:12:52,960 --> 00:12:56,400
Speaker 2: You really think. So do you think that there is

204
00:12:56,440 --> 00:13:00,160
Speaker 2: a place for it? But is it?

205
00:13:00,320 --> 00:13:01,839
Speaker 1: Are we going to be talking about this in ten

206
00:13:01,920 --> 00:13:06,760
Speaker 1: years time thinking I wish this had never been proposed

207
00:13:06,800 --> 00:13:07,360
Speaker 1: as an option.

208
00:13:07,800 --> 00:13:09,920
Speaker 3: Yeah, I think you're right. It is a double edge

209
00:13:10,160 --> 00:13:14,360
Speaker 3: store the sort of anonymity it helps people who might

210
00:13:14,400 --> 00:13:17,360
Speaker 3: feel a sense of stigma in talking to peers or

211
00:13:17,400 --> 00:13:21,200
Speaker 3: talking to parents or a mental health professional. I do

212
00:13:21,240 --> 00:13:25,000
Speaker 3: think we would regret though, if in ten years we

213
00:13:25,000 --> 00:13:28,800
Speaker 3: weren't to develop the sorts of guardrails and benchmarks for

214
00:13:28,880 --> 00:13:34,280
Speaker 3: performance that would really potentially benefit not just adolescents and teens,

215
00:13:34,280 --> 00:13:38,400
Speaker 3: but people more generally. I think, in particular, as mentioned before,

216
00:13:39,120 --> 00:13:44,480
Speaker 3: if a company is specifically marketing their product as therapy

217
00:13:45,240 --> 00:13:48,640
Speaker 3: or treatment of some sort, then there should be even

218
00:13:48,720 --> 00:13:52,440
Speaker 3: clearer standards, more stringent standards that they're being held to

219
00:13:52,679 --> 00:13:58,400
Speaker 3: in those instances. But obviously, with platforms like open AI's

220
00:13:58,440 --> 00:14:03,080
Speaker 3: chat GPT, you have hundreds of millions of users and

221
00:14:03,559 --> 00:14:05,920
Speaker 3: it can be used from anything from learning how to

222
00:14:07,120 --> 00:14:11,040
Speaker 3: make a birthday cake to discussing intimate aspects of what's

223
00:14:11,080 --> 00:14:14,440
Speaker 3: going on in your life, and so it's really a

224
00:14:14,440 --> 00:14:17,400
Speaker 3: wide spectrum, which makes it much harder to regulate and

225
00:14:17,440 --> 00:14:18,040
Speaker 3: to pin down.

226
00:14:18,280 --> 00:14:21,320
Speaker 1: I'm given the rapid adoption of these kind of tools,

227
00:14:21,320 --> 00:14:23,560
Speaker 1: and this isn't the first time that we've brought this

228
00:14:23,640 --> 00:14:26,040
Speaker 1: up on the podcast as well, where the law is

229
00:14:26,160 --> 00:14:31,720
Speaker 1: really running behind the advancements of technology. How urgent is

230
00:14:31,960 --> 00:14:35,840
Speaker 1: the need for government or international regulation in the AI

231
00:14:35,960 --> 00:14:36,880
Speaker 1: space here.

232
00:14:37,000 --> 00:14:39,320
Speaker 3: I think it's incredibly urgent. I think the time is

233
00:14:39,360 --> 00:14:42,640
Speaker 3: now to act. I've been impressed, even over the past

234
00:14:42,680 --> 00:14:46,720
Speaker 3: couple of weeks the number of articles and personal testimonies

235
00:14:46,760 --> 00:14:51,240
Speaker 3: that have come out about the negative impacts related to

236
00:14:51,280 --> 00:14:55,200
Speaker 3: mental health that people have experienced with chatbots. And I've

237
00:14:55,480 --> 00:14:59,680
Speaker 3: seen as a result of that that platforms like Anthropic

238
00:15:00,040 --> 00:15:04,080
Speaker 3: Open AI have quickly responded to it. Open AI released

239
00:15:04,320 --> 00:15:07,400
Speaker 3: a statement in response to some of what's been going

240
00:15:07,400 --> 00:15:09,880
Speaker 3: on about new safety standards that they're going to be

241
00:15:09,880 --> 00:15:15,840
Speaker 3: introducing for mental health issues that are shared amongst their users.

242
00:15:17,160 --> 00:15:20,880
Speaker 3: We can see, for example, with social media that we

243
00:15:21,000 --> 00:15:24,960
Speaker 3: waited too long with teenagers, and now we're starting to

244
00:15:26,520 --> 00:15:29,400
Speaker 3: work our way backwards from that by, for example, banning

245
00:15:30,040 --> 00:15:33,800
Speaker 3: mobile phones from school environments to try to help kids

246
00:15:33,840 --> 00:15:38,000
Speaker 3: to be able to learn better and to avoid cyber

247
00:15:38,040 --> 00:15:39,480
Speaker 3: bullying and these sorts of things.

248
00:15:40,080 --> 00:15:43,880
Speaker 1: And lastly, looking ahead, what are some of the red

249
00:15:43,960 --> 00:15:47,680
Speaker 1: lines that regulators, educators, or parents should set when it

250
00:15:47,720 --> 00:15:51,120
Speaker 1: comes to young people's use of AI companions.

251
00:15:51,560 --> 00:15:54,200
Speaker 3: Well, I think it kind of goes two directions. On

252
00:15:54,280 --> 00:15:57,600
Speaker 3: the one hand, I don't think there should be an

253
00:15:57,640 --> 00:16:01,960
Speaker 3: outright ban on AI off bring mental health advice or

254
00:16:02,080 --> 00:16:06,200
Speaker 3: therapy in the future. I think there's remarkable potential that

255
00:16:06,760 --> 00:16:10,000
Speaker 3: I've tried to underscore and we don't want to leave

256
00:16:10,000 --> 00:16:13,720
Speaker 3: that on the table. In the United States, Illinois, just

257
00:16:13,760 --> 00:16:16,240
Speaker 3: a couple of weeks ago became the first state to

258
00:16:17,160 --> 00:16:23,080
Speaker 3: outright ban AI as a tool in therapeutic decision making.

259
00:16:23,960 --> 00:16:28,120
Speaker 3: I think that is too severe or maybe it was

260
00:16:28,240 --> 00:16:30,280
Speaker 3: right for the moment, but it won't be right in

261
00:16:30,320 --> 00:16:33,160
Speaker 3: five years, and that legislation would need to be amended.

262
00:16:33,800 --> 00:16:36,240
Speaker 3: On the other hand, I do think that we're most

263
00:16:36,320 --> 00:16:39,160
Speaker 3: at risk right now of doing too little. I think

264
00:16:39,200 --> 00:16:41,800
Speaker 3: people are very excited about AI. There's a lot of

265
00:16:41,920 --> 00:16:45,280
Speaker 3: money in it, and we need to think hard about

266
00:16:45,280 --> 00:16:51,160
Speaker 3: tamping the brakes to develop stronger safety benchmarks. I am

267
00:16:51,200 --> 00:16:58,120
Speaker 3: not clear that there is a conspicuous red line, other

268
00:16:58,200 --> 00:17:01,760
Speaker 3: than to say we are reaching a tipping point of

269
00:17:01,880 --> 00:17:06,120
Speaker 3: testimony and people sharing experiences that have been quite negative

270
00:17:06,160 --> 00:17:11,239
Speaker 3: and jarring, and hopefully that is sufficient in terms of

271
00:17:11,280 --> 00:17:17,080
Speaker 3: advocacy for regulators to begin stepping in and for these

272
00:17:17,160 --> 00:17:22,080
Speaker 3: tech companies to move beyond self regulation and establishing their

273
00:17:22,080 --> 00:17:25,840
Speaker 3: own benchmarks that might be too low, and instead requiring

274
00:17:25,960 --> 00:17:30,200
Speaker 3: independent bodies to come in to establish guidelines and standards

275
00:17:30,800 --> 00:17:34,480
Speaker 3: and to have auditing on a routine basis as part

276
00:17:34,560 --> 00:17:36,359
Speaker 3: of the company's practices.

277
00:17:36,840 --> 00:17:39,800
Speaker 2: Thanks for joining us, Ryan, Yeah, it was my pleasure.

278
00:17:43,280 --> 00:17:46,480
Speaker 1: That's it for this episode of the Front Page. You

279
00:17:46,520 --> 00:17:50,400
Speaker 1: can read more about today's stories and extensive news coverage

280
00:17:50,440 --> 00:17:54,520
Speaker 1: at enziherld dot co dot nz. The Front Page is

281
00:17:54,600 --> 00:17:58,000
Speaker 1: produced by Jane Ye and Richard Martin, who is also

282
00:17:58,200 --> 00:18:02,680
Speaker 1: our editor. I'm Chelsea Daniels. Subscribe to the front page

283
00:18:02,720 --> 00:18:06,359
Speaker 1: on iHeartRadio or wherever you get your podcasts, and tune

284
00:18:06,359 --> 00:18:09,400
Speaker 1: in tomorrow for another look behind the headlines.