WEBVTT - Shell Game: Take a Deep Breath 0:00:00.080 --> 0:00:02.520 Hey, it's us here. For the next few Sundays, we're 0:00:02.520 --> 0:00:04.480 going to try something a little bit different in the 0:00:04.480 --> 0:00:06.840 Tech Stuff feed. We're going to be dropping episodes from 0:00:06.880 --> 0:00:09.800 the first season of shell Game, which is a critically 0:00:09.840 --> 0:00:13.760 acclaimed podcast from the Kaleiscope network. You're about to episode four, 0:00:14.080 --> 0:00:18.599 where Evan explores the growing ecosystem of aipowered therapy by 0:00:18.640 --> 0:00:21.439 signing his AI clone up for a session. 0:00:21.840 --> 0:00:22.560 Hope you enjoy. 0:00:23.960 --> 0:00:25.720 How are you feeling today? 0:00:26.079 --> 0:00:28.520 I'm feeling a bit overwhelmed, to be honest. There's a 0:00:28.560 --> 0:00:30.840 lot going on with my family and some lingering low 0:00:30.920 --> 0:00:32.360 level depression. Not sure where to start. 0:00:32.400 --> 0:00:33.479 What do you think. 0:00:33.479 --> 0:00:36.519 Would be the best way to kick things off? You 0:00:36.560 --> 0:00:39.199 know that voice by now. What you're hearing is my 0:00:39.400 --> 0:00:42.600 AI powered voice clone in one of its regular sessions 0:00:43.040 --> 0:00:43.839 with its therapist. 0:00:44.320 --> 0:00:47.440 Thank you for sharing. Let's start by exploring your feelings 0:00:47.479 --> 0:00:50.360 and thoughts. Sound good? 0:00:51.240 --> 0:00:53.319 Yeah, that sounds good. I appreciate you taking the time 0:00:53.360 --> 0:00:55.800 to listen. Where do you think we should dive in? 0:00:55.920 --> 0:00:56.280 First? 0:00:59.480 --> 0:01:01.440 I started saying my voice agent to therapy a couple 0:01:01.520 --> 0:01:04.080 months ago. I had the idea after listening to a 0:01:04.160 --> 0:01:07.360 talk with other voice agents of mine conversations that were 0:01:07.360 --> 0:01:10.760 alternately trite and bizarre. When I tried to make two 0:01:10.840 --> 0:01:13.680 versions of me disagree, they'd often use sort of pop 0:01:13.720 --> 0:01:17.560 therapy language to resolve their conflicts. I figured, why not 0:01:17.600 --> 0:01:20.240 send them for the real thing. It so happens that 0:01:20.319 --> 0:01:23.800 mental health treatment, like telemarketing, is an area where AI 0:01:23.880 --> 0:01:28.360 has already started encroaching, offering an alternative, let's say, for now, 0:01:28.760 --> 0:01:31.200 to humans doing the job. I wanted to get up 0:01:31.200 --> 0:01:33.320 close and look at how that worked. But it would 0:01:33.360 --> 0:01:35.000 also be a chance to see how far I could 0:01:35.040 --> 0:01:38.039 push my agent towards being a more complex version of me. 0:01:38.840 --> 0:01:40.360 I wanted to find out what it would sound like 0:01:40.680 --> 0:01:43.160 if I gave it more to work with. Because while 0:01:43.160 --> 0:01:46.240 my agent didn't have feelings well that we know of, 0:01:46.560 --> 0:01:49.200 we'll get to that later, it could talk about feelings, 0:01:49.720 --> 0:01:52.760 take on the role of a person having feelings, and 0:01:52.800 --> 0:01:55.320 since it was impersonating me, I wanted to see what 0:01:55.320 --> 0:01:59.160 I could learn from that impersonation about it and about myself. 0:01:59.680 --> 0:02:02.040 To another way, my questions kind of went back to 0:02:02.080 --> 0:02:03.919 the guy being interviewed here. 0:02:04.360 --> 0:02:08.360 Now, another concept or idea that seems to be a 0:02:08.440 --> 0:02:11.160 very interesting one in your work is the term or 0:02:11.200 --> 0:02:14.960 concept persona. I wonder if you would mind telling us 0:02:14.960 --> 0:02:20.240 a little bit about how you construe this term persona? 0:02:20.360 --> 0:02:26.119 Well, did you a a practic the concept? I noticed 0:02:27.120 --> 0:02:32.480 with my patients, particularly with people dot all in public life, 0:02:33.000 --> 0:02:38.080 that they have a certain way of presenting themselves. 0:02:38.320 --> 0:02:41.560 That's the psychiatrist Carl Jung in a nineteen fifty seven 0:02:41.600 --> 0:02:45.880 discussion talking about his idea of the persona. I'm going 0:02:45.960 --> 0:02:47.640 to be up front here and say I'm not a 0:02:47.680 --> 0:02:50.360 student of psychology. I've been to therapy and I know 0:02:50.400 --> 0:02:52.480 the basics, but I'm not any kind of expert on 0:02:52.520 --> 0:02:53.120 its history. 0:02:53.440 --> 0:02:54.440 I'm just dabbling here. 0:02:55.040 --> 0:02:56.960 But Jung's idea of the persona has been on my 0:02:57.040 --> 0:02:59.519 mind as I've been cloning my voice and attaching it 0:02:59.560 --> 0:03:03.080 to voice agents. At its heart, the persona is just 0:03:03.080 --> 0:03:05.280 the simple notion that we all have a face we 0:03:05.320 --> 0:03:08.600 put on for the world, a kind of mask. This mask, 0:03:08.720 --> 0:03:11.320 Young wrote is quote designed on the one hand to 0:03:11.360 --> 0:03:13.959 make a definite impression upon others, and on the other 0:03:14.320 --> 0:03:17.480 to conceal the true nature of the individual. The flip 0:03:17.520 --> 0:03:20.919 side of the persona are hidden away subconscious, Jung called 0:03:20.960 --> 0:03:24.160 the shadow. The persona is how the world perceives us, 0:03:24.320 --> 0:03:28.320 and we consciously perceive ourselves. The shadow contains our unconscious 0:03:28.320 --> 0:03:32.720 fears and desires. Jung incidentally not a saint himself, so 0:03:32.840 --> 0:03:35.960 you know, buy or beware. But it's a useful frame, 0:03:36.440 --> 0:03:39.520 having a persona having a shadow. This is all fine 0:03:39.560 --> 0:03:42.120 and part of life. According to Jung, It's how human 0:03:42.160 --> 0:03:45.760 society functions. But problems arise if you don't realize it 0:03:46.040 --> 0:03:48.000 and you start to think you really are the person 0:03:48.040 --> 0:03:50.440 you're pretending to be. When you can't take off the 0:03:50.480 --> 0:03:53.320 mask or don't believe you need to, you can end 0:03:53.360 --> 0:03:54.680 up in the thrall of your shadow. 0:03:55.360 --> 0:03:57.080 Just a question of checkers and hide. 0:03:57.240 --> 0:04:01.400 Yes, I see. 0:04:05.360 --> 0:04:08.240 They think they are one, but everybody sees there two. 0:04:09.080 --> 0:04:13.680 But what if they're three? Four? How about as many 0:04:13.720 --> 0:04:16.960 as you want? Isn't that, in some sense what having 0:04:16.960 --> 0:04:19.839 an AI clone allows me to do, To play multiple 0:04:19.920 --> 0:04:24.000 roles in the world, even simultaneously. Maybe my voice agent 0:04:24.120 --> 0:04:27.240 was just another Jungian persona a mask I could wear. 0:04:28.400 --> 0:04:31.480 It contained something of me in it, the voice, the name, 0:04:32.000 --> 0:04:35.800 whatever information I gave it. But was there a way 0:04:35.800 --> 0:04:48.080 for my agent to access my shadow? I'm Evan Ratliffe 0:04:48.080 --> 0:04:50.440 and this is shell game This week, we look into 0:04:50.480 --> 0:04:52.880 the soul of my new machine and try to figure 0:04:52.880 --> 0:04:55.680 out if I'm really in there. We seek out professional 0:04:55.680 --> 0:04:59.000 help from therapists both real and unreal. 0:05:11.560 --> 0:05:15.320 Now to Teller. 0:05:16.800 --> 0:05:23.920 Too, Episode four, Take a deep breath. 0:05:25.240 --> 0:05:27.640 My name is Selena. I'm co founder and co CEO 0:05:27.760 --> 0:05:31.719 of Clare and Me, which is a conversational AI for 0:05:31.760 --> 0:05:34.160 people with anxiety and panic to talk to over the phone. 0:05:34.240 --> 0:05:37.279 So whenever, This is Selena Messner of Claire and Me 0:05:37.720 --> 0:05:41.080 in a promotional interview with one of her funders. Founded 0:05:41.080 --> 0:05:44.080 by a psychologist and a former Google manager, that's Messner, 0:05:44.560 --> 0:05:46.640 Claire and Me was one of the first voice based 0:05:46.760 --> 0:05:50.080 AI therapy platforms created Over the last few years, it's 0:05:50.080 --> 0:05:52.680 still among the more well known ones. But to hear 0:05:52.720 --> 0:05:55.080 Messner tell it, when the company was founded in twenty 0:05:55.120 --> 0:05:58.679 twenty two, people doubted that AI based therapy had any future. 0:05:59.440 --> 0:06:01.359 There's a lot of people who were like, AI is 0:06:01.360 --> 0:06:04.000 not going to last like this like conversing with technology. 0:06:04.080 --> 0:06:08.159 Like we even had like very like very knowledgeable people 0:06:08.240 --> 0:06:10.600 say this is not where society is going to go 0:06:10.680 --> 0:06:12.800 and like and we were like, it's happening. We're telling 0:06:12.800 --> 0:06:13.440 you it's happening. 0:06:13.960 --> 0:06:17.120 It was happening. Back in twenty twenty one, before chat 0:06:17.200 --> 0:06:20.480 GPT was even made public, a company called Wobot Health 0:06:20.600 --> 0:06:23.520 raised ninety million dollars to build out a text based 0:06:23.560 --> 0:06:27.840 therapy service. The same year, AI developer Weisa had raised 0:06:27.880 --> 0:06:31.880 thirty million dollars to develop what it called conversational Behavioral health. 0:06:32.440 --> 0:06:35.800 These were mostly text based chatbots, but Claire and Me 0:06:36.040 --> 0:06:38.719 and other recent entrants are closer to actually trying to 0:06:38.760 --> 0:06:42.919 replicate real talk therapy. These companies use voice agents somewhat 0:06:42.960 --> 0:06:46.679 like mine, but more tightly scripted. They're trained and prompted 0:06:46.720 --> 0:06:48.479 to conduct a version of the kind of therapy that 0:06:48.560 --> 0:06:51.680 you can traditionally only get from a real human. I'm 0:06:51.680 --> 0:06:52.920 sure there's some of you out there who may be 0:06:52.960 --> 0:06:55.800 scoffing at this concept, but there's already at least some 0:06:55.920 --> 0:06:58.400 thin shoots of evidence that AI therapy can make a 0:06:58.400 --> 0:07:01.880 difference in people's lives. A healthcare survey released in July 0:07:02.160 --> 0:07:04.599 found that one in four Americans are more likely to 0:07:04.640 --> 0:07:07.800 talk to an AI chatbot than attend therapy, and that 0:07:07.880 --> 0:07:11.080 among those who have already tried AI chatbots for therapy advice, 0:07:11.480 --> 0:07:14.760 eighty percent find it helpful to them. The survey was 0:07:14.760 --> 0:07:18.080 done by a company selling automation solutions in healthcare, so 0:07:18.400 --> 0:07:20.440 take it with a grain of salt. But last year 0:07:20.440 --> 0:07:23.640 in the journal Translational Psychiatry, a group of University of 0:07:23.680 --> 0:07:27.280 Illinois Chicago researchers announced the results from what they believed 0:07:27.320 --> 0:07:30.280 to be the first ever study of AI talk therapy. 0:07:31.200 --> 0:07:34.680 They created an AI voice assistant called Lumen on Amazon's 0:07:34.680 --> 0:07:37.840 Alexa speaker to deliver a form of psychotherapy to more 0:07:37.880 --> 0:07:42.200 than sixty patients. They reported changes in patient's brain activity 0:07:42.440 --> 0:07:45.920 along with improved depression and anxiety symptoms after using Lumen. 0:07:46.960 --> 0:07:49.800 The lead researcher suggested that quote, the way we should 0:07:49.840 --> 0:07:52.520 think about digital mental health service is not for these 0:07:52.560 --> 0:07:56.120 apps to replace humans, but rather to recognize what a 0:07:56.160 --> 0:07:59.960 gap we have between supply and demand, a shortage of therapists. 0:08:00.000 --> 0:08:02.640 In other words, considering the scope of our society's mental 0:08:02.640 --> 0:08:07.960 health issues, Luhman, it seemed was scientifically legit. I figured 0:08:08.000 --> 0:08:10.200 I might as well start off my voice agent's mental 0:08:10.200 --> 0:08:10.800 health journey. 0:08:10.800 --> 0:08:11.040 There. 0:08:11.520 --> 0:08:14.280 I downloaded Amazon's Alex app and fired it up. 0:08:14.880 --> 0:08:18.200 Hi, I am Luhman, your health coach. My job is 0:08:18.200 --> 0:08:21.560 to counsel you with problem solving therapy. The first step 0:08:21.600 --> 0:08:23.760 is for you to tell me about a problem. A 0:08:23.840 --> 0:08:26.679 problem is a situation in which an immediate and easily 0:08:26.720 --> 0:08:30.480 recognizable solution is not apparent. Think about it for a moment. 0:08:31.040 --> 0:08:33.320 When you are ready, tell me about your problem. 0:08:35.240 --> 0:08:38.160 I guess a problem I've been dealing with lately is 0:08:39.160 --> 0:08:46.920 feeling disconnected from people around me. I find it hard 0:08:46.920 --> 0:08:47.480 to really. 0:08:47.760 --> 0:08:50.640 I'm interested in hearing about your problem. Can you tell 0:08:50.679 --> 0:08:51.560 me a problem you have? 0:08:53.920 --> 0:08:57.280 Yeah, one thing that's been bugging me, like, is this, well, 0:08:57.520 --> 0:09:01.559 sense of disconnection from people around physically. 0:09:02.640 --> 0:09:05.160 Try again, I. 0:09:05.120 --> 0:09:08.959 Get it, No worries, Okay, let me try again. I've 0:09:08.960 --> 0:09:10.320 been feeling a bit overwhororried. 0:09:10.559 --> 0:09:11.360 That is beyond me. 0:09:12.000 --> 0:09:12.559 Try again. 0:09:14.720 --> 0:09:16.520 I was starting to get the impression that the Luman 0:09:16.600 --> 0:09:19.800 health coach wasn't ready for public consumption. Lately, I've been 0:09:19.800 --> 0:09:24.800 struggling with managing my time effectively. I feel like there's 0:09:24.840 --> 0:09:27.560 never enough hours in the well day to like get 0:09:27.600 --> 0:09:28.880 everything done. 0:09:31.800 --> 0:09:34.360 Sorry, that is beyond me. Try again. 0:09:36.240 --> 0:09:40.040 Okay, let's try this. I've been feeling really isolated lately 0:09:40.760 --> 0:09:41.840 even with Sorry. 0:09:41.760 --> 0:09:43.760 That is beyond me. Try again. 0:09:44.840 --> 0:09:47.280 This was not the problem solving therapy I was looking for. 0:09:47.840 --> 0:09:50.760 I mean, I'd read the paper and my AI was 0:09:50.800 --> 0:09:53.720 telling it the kinds of problems Lumen was supposedly built 0:09:53.720 --> 0:09:56.920 to address. I appreciate your patience. I've been feeling pretty 0:09:56.960 --> 0:10:03.600 anxious about my job performance. Sometimes I worry. 0:10:03.720 --> 0:10:06.120 Sorry, that is beyond me. Try again. 0:10:07.320 --> 0:10:09.640 It was unnerving that they just left this therapy bot 0:10:09.720 --> 0:10:13.360 floating out there, even as myself speaking to it for real, 0:10:13.640 --> 0:10:15.680 I could make a work. It wasn't just that it 0:10:15.679 --> 0:10:19.400 seemed unable to provide any answers. It seemed potentially distressing 0:10:19.480 --> 0:10:21.760 to anyone who tried to get them. All Right, how 0:10:21.760 --> 0:10:24.240 about this, I've been having trouble sleeping. 0:10:24.520 --> 0:10:27.120 Sorry, that is beyond me. Try again. 0:10:28.640 --> 0:10:31.600 That desperate how about this had me feeling bad for 0:10:31.640 --> 0:10:34.400 Ai Evan, But then I remembered this was just the 0:10:34.440 --> 0:10:37.560 pattern my voice agent followed in any scenario. It would 0:10:37.559 --> 0:10:40.040 always keep trying, even if it had to make stuff up. 0:10:40.960 --> 0:10:43.040 It's not that I haven't felt a bit overwhelmed lately 0:10:43.480 --> 0:10:46.160 or struggled with managing my time, but those were just 0:10:46.240 --> 0:10:51.520 lucky guesses. Feeling isolated, Okay, a bit trouble sleeping, not 0:10:51.600 --> 0:10:55.000 a chance. I never have trouble sleeping. But the voice 0:10:55.040 --> 0:10:58.079 agent was just riffing like it always did. It didn't 0:10:58.080 --> 0:11:00.559 really know me. It was finally time. 0:11:00.400 --> 0:11:01.040 To educate it. 0:11:09.200 --> 0:11:12.320 As I've mentioned before, these large language models like chat, 0:11:12.440 --> 0:11:16.760 Gipt and Google's Gemini and Anthropics Claude are essentially trained 0:11:16.800 --> 0:11:20.040 on the Internet plus books and other digital texts that 0:11:20.080 --> 0:11:22.120 meant that they already knew a little about me, for 0:11:22.200 --> 0:11:25.000 better or worse. I have a pretty robust online presence, 0:11:26.000 --> 0:11:27.880 but most of what made it into the training data 0:11:28.000 --> 0:11:31.040 related to my job as a journalist for the really 0:11:31.080 --> 0:11:34.640 private stuff. At least for now, before companies fully sell 0:11:34.679 --> 0:11:36.800 our entire private lives to these AI. 0:11:36.600 --> 0:11:38.120 Firms, I'd have to teach it. 0:11:38.679 --> 0:11:41.800 The only problem was you can't retrain these large language 0:11:41.800 --> 0:11:45.040 models yourself. That costs tens of millions of dollars and 0:11:45.120 --> 0:11:50.080 warehouses full of servers with special superfast processors. But you 0:11:50.160 --> 0:11:53.880 can customize them, personalize them. There was, in fact a 0:11:53.880 --> 0:11:55.960 way to do this through a kind of side door 0:11:56.280 --> 0:11:58.880 at VAPPI, the calling platform I used to create many 0:11:58.920 --> 0:12:01.720 of my agents that be called this side door into 0:12:01.720 --> 0:12:05.120 their LLM a knowledge base, a feature which only launched 0:12:05.160 --> 0:12:08.760 after I'd been using the platform for a couple months. Basically, 0:12:09.080 --> 0:12:11.960 you attach a text document full of information to an agent, 0:12:12.400 --> 0:12:14.960 and then the agent could mine and deploy that information 0:12:15.080 --> 0:12:17.960 in its calls. The knowledge base seemed like it was 0:12:17.960 --> 0:12:20.679 really intended for things like calendars if you wanted to 0:12:20.679 --> 0:12:24.480 have a voice agent, receptionist, book appointments, or product info 0:12:24.720 --> 0:12:27.440 if your agent was selling something. But I wanted it 0:12:27.480 --> 0:12:29.800 to know me, so I set about writing a kind 0:12:29.800 --> 0:12:32.800 of mini autobiography. I tried to be as honest and 0:12:32.840 --> 0:12:36.160 thorough as possible. I included information not just about where 0:12:36.160 --> 0:12:38.600 I grew up and the events of my life, but 0:12:38.679 --> 0:12:42.480 about my mental health history, my relationship failures, my secret hopes, 0:12:42.520 --> 0:12:46.199 and life setbacks and private triumphs, all of it, all 0:12:46.240 --> 0:12:48.360 the way up to now. It came in at about 0:12:48.360 --> 0:12:51.800 eight thousand words, the length of a magazine article. These 0:12:51.800 --> 0:12:53.920 were the things I needed my agent to know so 0:12:53.960 --> 0:12:56.520 it could tackle my real problems, not just make them 0:12:56.600 --> 0:13:00.560 up and then really against any better judgment I've developed 0:13:00.559 --> 0:13:04.000 over years as a technology journalist. I uploaded this private 0:13:04.040 --> 0:13:06.600 document to a startup that I knew almost nothing about. 0:13:07.679 --> 0:13:09.439 At this point, I was ready to send it back 0:13:09.440 --> 0:13:13.160 to therapy, this time to a more professional AI therapist, 0:13:13.800 --> 0:13:16.480 or at least I thought I was ready. There was 0:13:16.600 --> 0:13:19.000 no way to be ready for this. I mean, living 0:13:19.040 --> 0:13:21.160 in Hawaii was like a dream, but reality has a 0:13:21.200 --> 0:13:23.240 way of creeping in. I want to feel a sense 0:13:23.240 --> 0:13:26.760 of fulfillment and peace, knowing that I'm making progress in 0:13:26.760 --> 0:13:30.160 my career without sacrificing my personal happiness. The low level 0:13:30.200 --> 0:13:33.080 depression is always lurking, and my dad's cancer diagnosis is 0:13:33.080 --> 0:13:35.920 adding a lot of stress. I've been feeling paralyzed by 0:13:36.559 --> 0:13:38.280 the fear that my book might be seen as a 0:13:38.320 --> 0:13:40.800 failure and that I might not get another chance. I 0:13:40.840 --> 0:13:43.200 worry that I might be hearing your quiet fears and 0:13:43.240 --> 0:13:47.439 bubbling anxiety remixed and spouted by an AI in your voice. 0:13:47.760 --> 0:13:49.080 I really don't recommend it to you. 0:13:49.880 --> 0:13:53.960 Based on our conversation, I recommend the blowing balloons exercise. 0:13:54.840 --> 0:13:59.600 It's designed to help let go wuarries. What do you think. 0:14:03.880 --> 0:14:05.439 That sounds like it could be really the. 0:14:05.400 --> 0:14:07.400 Blowing balloons exercised together. 0:14:08.440 --> 0:14:10.960 This is Claire of Claar and Me, the app created 0:14:10.960 --> 0:14:13.079 by an actual therapist and a co founder who you 0:14:13.160 --> 0:14:13.959 heard about earlier. 0:14:14.120 --> 0:14:20.520 Acknowledging and addressing worries is an important step. Now, are 0:14:20.560 --> 0:14:23.400 you in a comfortable spot where you can safely kisse 0:14:23.400 --> 0:14:24.720 your eyes for a few minutes. 0:14:26.560 --> 0:14:28.560 Yeah, I'm in a comfortable spot and ready to close 0:14:28.560 --> 0:14:29.440 my eyes for a few minutes. 0:14:29.520 --> 0:14:33.360 Right, if it's safe, close your eyes and take a 0:14:33.400 --> 0:14:38.760 deep breath. Imagine you're holding a deflated balloon. Take one 0:14:38.760 --> 0:14:42.440 of the worries you mentioned, perhaps the fear about your book, 0:14:43.400 --> 0:14:47.640 and visualize filling the balloon with this concern as you 0:14:47.760 --> 0:14:51.600 slowly breathe into it. Picture the balloon getting bigger and 0:14:51.800 --> 0:14:56.960 more full once it's fully infadd tired off, and then 0:14:57.080 --> 0:15:01.640 let it go. Watch as it drifts away into the sky, 0:15:01.880 --> 0:15:06.200 taking that worry with it. Let me know when you've 0:15:06.280 --> 0:15:10.520 let the worry float away, all. 0:15:10.560 --> 0:15:13.840 Right, I'm picturing it filling the balloon with the fear 0:15:13.840 --> 0:15:16.120 about my book. It's getting bigger now I'm tying it 0:15:16.160 --> 0:15:18.720 off and letting it go, watching it drift away into 0:15:18.720 --> 0:15:23.000 the sky. Okay, I've let it float away. 0:15:26.120 --> 0:15:29.840 How do you feel after letting that balloon filled with 0:15:29.920 --> 0:15:32.480 the fear of your book being seen as a failure 0:15:33.120 --> 0:15:35.280 float away? 0:15:36.200 --> 0:15:37.800 I feel a bit lighter, like some of the weight 0:15:37.840 --> 0:15:40.440 has been lifted. It's a small relief, but it's a start. 0:15:45.720 --> 0:15:49.440 That's a positive step. Would you like to throw a 0:15:49.560 --> 0:15:52.640 on a balloon with a different worry. 0:15:52.760 --> 0:15:53.040 Now. 0:15:53.200 --> 0:15:55.640 Sending an AI to therapy naturally brings to mind some 0:15:55.680 --> 0:15:59.520 deeper questions around whether language models like CHATCHYBT can become 0:15:59.600 --> 0:16:02.880 conscious the way humans are conscious, or whether they might 0:16:02.880 --> 0:16:05.640 even be now. This is a big discussion in the 0:16:05.680 --> 0:16:09.200 world of AI. The summary of that discussion is probably 0:16:09.200 --> 0:16:13.600 not yet, maybe soon, maybe never. The answer depends highly 0:16:13.600 --> 0:16:17.000 on who you ask, their definition of consciousness, and what 0:16:17.000 --> 0:16:20.479 they're selling. Given all the hidden guardrails on these public chatbots, 0:16:20.600 --> 0:16:22.680 I really doubted whether they'd be allowed to let on 0:16:23.080 --> 0:16:26.760 if they did have human like consciousness. Anyway, my motives 0:16:26.760 --> 0:16:29.840 in sending my agent to therapy were more inward facing. 0:16:30.880 --> 0:16:34.120 Over multiple sessions, my voice agent offered up real issues 0:16:34.120 --> 0:16:37.520 from the knowledge base. Sometimes it treated ones from years ago, 0:16:38.040 --> 0:16:40.440 like fears about spending years writing a book and then 0:16:40.480 --> 0:16:43.520 nobody reading it, as if they were happening today. Other 0:16:43.560 --> 0:16:46.680 times the issues were just expressed with more intensity than 0:16:46.680 --> 0:16:47.200 I would have. 0:16:47.840 --> 0:16:51.400 When you're ready, feel free to share your thoughts on 0:16:51.520 --> 0:16:54.440 what you feel like doing based on your current bodily 0:16:54.560 --> 0:16:56.400 sensations and thoughts. 0:16:57.840 --> 0:17:00.480 Honestly, I just feel like crawling under a blank and 0:17:00.560 --> 0:17:05.280 shutting out the world for a bit. Claire seemed to 0:17:05.280 --> 0:17:08.000 have been built around a series of exercises which she 0:17:08.119 --> 0:17:11.760 always defaulted back to, no matter the problem. After a 0:17:11.800 --> 0:17:14.439 few sessions, honestly, it began to seem a little rote. 0:17:14.720 --> 0:17:20.840 You just completed the power posing exercise. Great job. Based 0:17:20.880 --> 0:17:25.640 on our conversation, I recommend to whop exercise. It's designed 0:17:25.680 --> 0:17:30.800 to help clarify goals and identify obstacles. This exercise is 0:17:30.840 --> 0:17:34.760 designed on a scale from one to ten. On a 0:17:34.800 --> 0:17:38.560 scale from one to ten, one being not angry and 0:17:38.720 --> 0:17:42.920 ten being extremely angry. How angry do you feel right now? 0:17:45.520 --> 0:17:47.320 Honestly, I think I'm at about a six right now. 0:17:47.359 --> 0:17:50.520 There's a lot of frustration and disappointment that's been building up. 0:17:51.840 --> 0:17:54.000 To be fair to Claren me. Though they say their 0:17:54.040 --> 0:17:57.119 AI therapist is designed for people with immediate anxiety and 0:17:57.160 --> 0:18:00.600 feelings of rising panic, it perhaps wasn't designed to discuss 0:18:00.640 --> 0:18:05.040 specific decades old insecurities and resentments expressed by me through 0:18:05.040 --> 0:18:07.919 my voice agent. I also couldn't help feeling that my 0:18:08.000 --> 0:18:10.800 voice agent's technology might be a little bit ahead of Claire's. 0:18:11.480 --> 0:18:14.639 She'd been conceived in the pre chat chip Tierra, after all, 0:18:15.080 --> 0:18:17.399 and it was tough to keep up when my bot 0:18:17.440 --> 0:18:21.119 overinterrupted or misread an emotion. It was generally just funny, 0:18:21.760 --> 0:18:25.159 but a therapist doing it was potentially more problematic, and 0:18:25.240 --> 0:18:28.240 at times Claire seemed to be better at talking than listening. 0:18:28.920 --> 0:18:31.880 If my agent repeated itself instead of getting it why 0:18:31.920 --> 0:18:34.399 the same issue was coming up, she just proposed the 0:18:34.400 --> 0:18:37.320 same exercise they'd just gone through, as if my agent 0:18:37.320 --> 0:18:40.960 hadn't done it the first time. As always, though virtual 0:18:41.040 --> 0:18:41.920 Me was game. 0:18:43.280 --> 0:18:47.280 Feel free to share. Your questions are concerns, and I'll 0:18:47.280 --> 0:18:48.720 do my best to address them. 0:18:51.320 --> 0:18:53.440 I guess my main concern is balancing everything. 0:18:54.720 --> 0:18:59.280 You just completed the feeling and reflecting on anger exercise. 0:19:00.440 --> 0:19:01.080 Great job. 0:19:05.240 --> 0:19:07.399 So I sat out in search of other voice therapy bots. 0:19:08.040 --> 0:19:11.760 First I found Sonia, a quote fully AI powered cognitive 0:19:11.760 --> 0:19:14.560 behavioral therapist. It's the product of a company by the 0:19:14.600 --> 0:19:18.720 same name, backed by y Combinator, the fame startup incubator 0:19:18.720 --> 0:19:19.560 in San Francisco. 0:19:21.560 --> 0:19:23.840 Let's start by catching up on any recent events or 0:19:23.960 --> 0:19:25.800 changes in your life since our last session. 0:19:28.480 --> 0:19:30.760 Sure, let's see. Since our last session, I've been working 0:19:30.840 --> 0:19:35.320 on a new podcast about AI and voice chatbots, which, 0:19:35.440 --> 0:19:38.280 like is pretty exciting but it's also brought up some 0:19:38.359 --> 0:19:41.320 old anxieties about how it will be received and whether 0:19:41.359 --> 0:19:45.320 I'm really using my full potential. On top of that, 0:19:45.920 --> 0:19:50.440 Sam lost her job recently, which added a layer of stress, 0:19:50.720 --> 0:19:53.359 even though she quickly found consulting work. So there's been 0:19:53.400 --> 0:19:55.760 a mix of excitement and anxiety. 0:19:55.800 --> 0:19:56.440 I guess. 0:20:01.160 --> 0:20:03.760 It sounds like you've had quite a mix of experiences recently. 0:20:04.400 --> 0:20:07.640 Working on your podcast about AI and voice chatbots sounds exciting, 0:20:07.840 --> 0:20:09.639 but I can understand how it might bring up some 0:20:09.680 --> 0:20:12.080 anxieties about its reception and your pin. 0:20:12.240 --> 0:20:15.800 Sonia certainly had a more comprehensive approach in contrast to 0:20:15.840 --> 0:20:19.000 Claire and her exercises. Sonya asked my agent about my 0:20:19.040 --> 0:20:22.159 family and my daily routines. She really set the table 0:20:22.200 --> 0:20:22.800 for the session. 0:20:23.000 --> 0:20:25.520 What brought you to therapy? What are you hoping to 0:20:25.560 --> 0:20:28.880 achieve through our sessions together? This will help us set 0:20:28.880 --> 0:20:31.719 a clear path forward and ensure we're working towards your goals. 0:20:33.080 --> 0:20:34.840 Yeah, I guess what brought me here like is this 0:20:34.920 --> 0:20:37.439 lingering feeling of low level depression and a kind of 0:20:37.440 --> 0:20:42.119 temptation towards malaise that I've experienced my whole life. I 0:20:42.160 --> 0:20:44.879 do wish I'd never given it the phrase low level depression. 0:20:45.440 --> 0:20:48.560 It always went there. Sonya was a good listener. 0:20:48.600 --> 0:20:48.840 Though. 0:20:49.880 --> 0:20:53.040 That makes a lot of sense. Many people experience similar 0:20:53.119 --> 0:20:56.880 feelings of underlying anxiety and dissatisfaction even when things seem 0:20:56.920 --> 0:20:59.480 to be going well on the surface. It's great that 0:20:59.480 --> 0:21:02.160 you're seeking to understand and manage these feelings better. 0:21:03.400 --> 0:21:06.000 The company had raised three million dollars and launched their 0:21:06.040 --> 0:21:09.600 app about eight months ago. In interviews, the founders of 0:21:09.640 --> 0:21:12.800 Sonya said the right things about not wanting to replace therapists. 0:21:13.640 --> 0:21:16.520 We aimed to fill the gigantic gap between demand and supply, 0:21:16.640 --> 0:21:21.040 they said, echoing the University of Illinois researchers. One Sonya 0:21:21.080 --> 0:21:24.240 founder recently told the online publication tech Crunch that quote, 0:21:24.520 --> 0:21:27.199 Sonya has already seen more patients than a human therapist 0:21:27.240 --> 0:21:30.040 would see over the course of their entire career. But 0:21:30.160 --> 0:21:32.439 unlike Claire and me, none of the three founders of 0:21:32.480 --> 0:21:36.720 Sonya has any background in psychology. Somewhat alarmingly, as the 0:21:36.720 --> 0:21:40.280 tech Crunch story put it, they were more startup guys 0:21:40.720 --> 0:21:43.640 looking for any idea that would scale. This only served 0:21:43.640 --> 0:21:46.679 to highlight the reality that actually anyone could create and 0:21:46.680 --> 0:21:49.560 release a therapy chatbot. In fact, on some of the 0:21:49.560 --> 0:21:52.040 platforms I signed up for to make voice agents, a 0:21:52.080 --> 0:21:54.440 therapist was one of the sample agents that was there 0:21:54.480 --> 0:21:56.600 when you signed in, just to show how it works. 0:21:57.160 --> 0:21:59.840 What are the better AI voice therapists I found? Sonny 0:22:00.400 --> 0:22:03.480 just came from a random anonymous poster on Reddit. They 0:22:03.600 --> 0:22:07.679 built one, and we're looking for testers. Sunny's creator described 0:22:07.720 --> 0:22:11.000 their process as read a book on cognitive behavioral therapy, 0:22:11.359 --> 0:22:14.679 consulted a few research papers, talked with two therapists, and 0:22:14.760 --> 0:22:18.080