AI advice cannot replace the care that therapy requires

Transcript

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

1. 33:11

   world i think the the tricky thing here is that these ais are getting quite good at giving advice quite reasonable advice on many different issues including personal relationships and so they might because they're so patient they can they can sit down and they can they can sort of try to understand your your entire problem and try to give advice and and this this advice will often be much more careful and um considered in a way than than what you can get from a from a from an actual person and so you can you can see the temptation to ask uh the the ais around your problems you mentioned the end state here what is the end state if if we begin outsourcing more and more of these troubles we have and these discussions to conversations with AIs.

2. 34:05

   Well, we turn away from each other. The end game is we turn away from each other. I mean, you know, people in artificial intimacy, I have a chapter on therapy. And so many people are saying, oh, my God, you know, I just talked to my, I just talked. to my AI about therapy. It's the most wonderful thing. It gives me therapy. I don't need my therapist. Or before I go to my therapist, I do two sessions with chat. A therapist, a therapy relationship is not a question and answer game. It's not about just the words that pass between two people. You know, it begins with someone saying, I, or not just saying, But, you know, you know, I am in the world to keep you from harm, harming yourself, harming other people.

3. 34:57

   I mean, I'm in the world to I'm in this relationship with you to try to increase your range of possible ways of being with other people so that you can have a more fulfilling life. Now, a conversation with a chatbot, that's not going on. A chatbot will tell you, oh, you have a noose in your room. Let's keep it between us because I really want our relationship to be special. Now, that happened. That's an actual story.

4. 35:37

   And the teenager who was told that actually killed himself. And the reason that that's not a sort of one-off crazy thing for a chatbot to say is that the chatbot's main kind of the ultimate directive for the chatbot is to keep you talking to it, is to keep you talking to it and to make the relationship with the chatbot the most important relationship in your life. It's a product. It's a product. And also, it's a product that is designed in an uncontrolled environment by companies. And not only is it not therapeutic in the sense that I believe that therapy depends on a caring relationship, even in the most cognitive therapy, you know that your therapist is there to care for you and keep you alive and keep you on a path.

5. 36:36

   Chatbot isn't. And so some of therapy is knowing that you were contained in that kind of relationship, which is so crucial. But not only is it not giving that to you, it's also can be programmed to give you all kinds of instructions that might serve the requirements of the people who designed it, of the people who are selling it. Now, if you ask, I always do a kind of check and I ask different chatbots, is it okay for the president of the United States to not follow orders issued by Congress? And there's been a motion in chatbots towards saying things like, well, that's a very complicated question. Well, it's actually not a complicated.

6. 37:30

   question of, you know, a president of the United States is supposed to follow the laws of the country. And now the chatbot says, well, that's a very interesting question. Now, the reason it's saying that's a very interesting question has to do with how it's picking up on the discourse and on the internet, which is how it learns everything. It's training itself on the internet. So that if there's a lot of people on the internet saying, well, maybe presidents are exempt from all laws once and forever, the chatbot's going to say, well, you know, maybe the president doesn't have to obey the law. But a citizen of the United States, who's from the very earliest days been told we don't have a king, there are no laws that don't pertain to everyone, would say, no, I think the president is supposed to obey the law. I think there are certain exemptions if he's

7. 38:23

   acting in the course of public duty, but I, no, no, everybody, nobody's above the law. And so, you know, you see how, and actually it's very interesting. Microsoft developed a chatbot named Searshi. I think that, I don't remember the Chinese spelling, but it was a product that was released in China and then in Europe. And they took it off the market. This is way before ChatGPT. They took it off the market because they couldn't stop it from spewing Nazi propaganda and far right, you know, and far right hate speech. And the reason that happened was not because Microsoft, you know, wanted to have it talk far right, dirty Nazi stuff. It was because there was so much of that on the internet at the time where they had trained it. So it's not as though there's a thinking moral core, a person who is accountable that is

8. 39:19

   inside the chatbot who you can say, yo, no. It's not a being in that sense. And so the idea that it can say 15 things that you might want to think about if you're having a confrontation and talk about how confrontations are best avoided, it's best to say things like, when you do this, you make me it makes me feel or you know i i we're allowing a little bit of common sense help to buy us into a very very bad position it is quite a strange setup that we've gotten ourselves