Latest / What's Your Problem? / The Company Where Everyone Has Their Own AI Agent

Transcript

The Company Where Everyone Has Their Own AI Agent · What's Your Problem?

• 0:15
  Speaker 1: Pushkin. I'm Jacob Goldstein, and this is what's your problem.
• 0:32
  Speaker 1: My guest today is Dan Shipper. Dan is the co
• 0:36
  Speaker 1: founder and CEO of a company called Every. Every publishes
• 0:40
  Speaker 1: newsletters about AI, they develop AI related software, and they
• 0:45
  Speaker 1: consult to other companies about how to use AI. So
• 0:48
  Speaker 1: what I'm saying is every is an AI company. And
• 0:52
  Speaker 1: when Dan describes Every's work culture, he says, we live
• 0:56
  Speaker 1: in the future, which means that people at every use
• 1:00
  Speaker 1: AI in ways that lots of the rest of us
• 1:03
  Speaker 1: will soon be using AI. Given that context, there were
• 1:07
  Speaker 1: two specific problems that I wanted to talk to Dan about.
• 1:11
  Speaker 1: Problem number one, how do you build a company where
• 1:14
  Speaker 1: almost everybody has their own AI agent working alongside them?
• 1:19
  Speaker 1: Problem number two, how do you use AI as a
• 1:23
  Speaker 1: tool to improve your writing rather than as a replacement
• 1:27
  Speaker 1: for writing? On's particularly close to my heart. We talk
• 1:31
  Speaker 1: about problem number two later in the show. Regarding problem
• 1:35
  Speaker 1: number one, I will say, almost everybody at every has
• 1:40
  Speaker 1: their own AI agent, and specifically they use this popular
• 1:44
  Speaker 1: open source software called open claw. You may have heard
• 1:47
  Speaker 1: of it. It's basically an AI assistant that you communicate
• 1:51
  Speaker 1: with via Slack or via text, and at every they
• 1:54
  Speaker 1: give their clause names. So Dan's claw, for example, is
• 1:58
  Speaker 1: called R two C two and R two C two
• 2:01
  Speaker 1: and the other clause are on slack, just like the
• 2:04
  Speaker 1: human beings who work at every So I wanted to know,
• 2:08
  Speaker 1: well a lot, really, I want to know a lot.
• 2:10
  Speaker 1: You know, what do the clause do, how do they
• 2:12
  Speaker 1: interact with the humans, what works, what doesn't work, what
• 2:16
  Speaker 1: is super weird? And just what are the clause like.
• 2:20
  Speaker 2: What's really interesting about them is that they have personalities.
• 2:26
  Speaker 2: So you give them a name, they give them a
• 2:27
  Speaker 2: little bit of a personality. They modify themselves in response
• 2:34
  Speaker 2: to your messages. So when I say, hey, go do
• 2:40
  Speaker 2: this thing, it will it starts to realize, okay, I'm
• 2:43
  Speaker 2: going to be asked to do task act and it
• 2:45
  Speaker 2: will build things into itself, whether that's into its prompting
• 2:48
  Speaker 2: or into little tools that it builds for itself to
• 2:51
  Speaker 2: make it good at that. And so you're kind of
• 2:55
  Speaker 2: as you talk to it, it is evolving with you
• 2:59
  Speaker 2: and building functionality for itself to make it good for
• 3:02
  Speaker 2: you and the things you do.
• 3:04
  Speaker 1: It's okay, so you have this agent R two C two.
• 3:07
  Speaker 1: But to sort of tell me more about how it works,
• 3:10
  Speaker 1: like I know you've talked before about how it helps
• 3:13
  Speaker 1: you kind of manage this piece of software. You wrote
• 3:16
  Speaker 1: this thing called Proof that a lot of people are
• 3:18
  Speaker 1: like asking you questions about, but just give me more information,
• 3:23
  Speaker 1: like how do you actually use or work with it
• 3:25
  Speaker 1: day to day?
• 3:27
  Speaker 2: So you know, an easy one is this morning he
• 3:33
  Speaker 2: went through all of the issues that were submitted in
• 3:36
  Speaker 2: Proof and then prepared a little document for me that
• 3:41
  Speaker 2: was like, here's all the stuff that like has gone
• 3:44
  Speaker 2: run over the next over the last twenty four hours,
• 3:46
  Speaker 2: and here's what I recommend we do. He also has
• 3:51
  Speaker 2: a little document that he keeps that's my to do list.
• 3:52
  Speaker 2: So he pings me in the morning and is like,
• 3:54
  Speaker 2: here's your to do list. He also, you know, one
• 3:59
  Speaker 2: of the things I love to do is I always
• 4:00
  Speaker 2: go to a restaurant nearby the office and have breakfast
• 4:04
  Speaker 2: in the morning and read. And I will often take
• 4:08
  Speaker 2: a picture of a thing and be like, hey, can
• 4:12
  Speaker 2: you explain this to me? And he'll like explain it
• 4:14
  Speaker 2: and help me think about it.
• 4:15
  Speaker 1: Does he text you back? Does he slack you back?
• 4:18
  Speaker 2: It's on slack, Yeah, it's all on slack, So you
• 4:20
  Speaker 2: slack him.
• 4:21
  Speaker 1: A picture of like a passage from a book.
• 4:24
  Speaker 2: He also keeps all my reading notes, so I'll say, okay,
• 4:27
  Speaker 2: then now save this, and I have a whole reading
• 4:29
  Speaker 2: feed that is like all the stuff that I've been
• 4:32
  Speaker 2: reading over the last couple months. I also just talked
• 4:36
  Speaker 2: to him about ideas. So today I was, you know,
• 4:39
  Speaker 2: just thinking through some deep, deep thoughts as I do,
• 4:42
  Speaker 2: about where we are in AI and where we're going
• 4:44
  Speaker 2: and all that kind of stuff, and I just like,
• 4:46
  Speaker 2: monologue is our is one of our products as a
• 4:48
  Speaker 2: speech to text app. I just took a note with
• 4:50
  Speaker 2: monologue like a voice memo, and just blabbed into it
• 4:54
  Speaker 2: for a while and then started talking to R two
• 4:56
  Speaker 2: about what do you think I'm really trying to say here?
• 4:59
  Speaker 2: Having something like this that can reflect back to you
• 5:04
  Speaker 2: what you're saying, yes and put your finger on it
• 5:07
  Speaker 2: is super valuable, especially if it's something that has it
• 5:14
  Speaker 2: is the equivalent of an expert on like pretty much
• 5:17
  Speaker 2: any domain. It's not going to be perfect, but like
• 5:21
  Speaker 2: you know, it's it's a good stand in for an expert.
• 5:23
  Speaker 1: And that to be clear, anybody can do with any
• 5:27
  Speaker 1: language model, right, Like you don't need a claw for that.
• 5:29
  Speaker 1: That's like classic nice use of the window on any
• 5:33
  Speaker 1: decent large language model.
• 5:35
  Speaker 2: Totally, it's just a little bit easier to like, I
• 5:39
  Speaker 2: don't have to He's just always on, So it's like
• 5:42
  Speaker 2: it's just easier to like slack him in the way
• 5:43
  Speaker 2: that would have here.
• 5:44
  Speaker 1: It's amazing the way the tiny, tiny frictions, right, can
• 5:47
  Speaker 1: make a difference, even like your own mental frictions. Yeah, yeah,
• 5:51
  Speaker 1: so okay, So that's like one man in his claw.
• 5:55
  Speaker 1: There's this other level, which is like claw to claw
• 5:59
  Speaker 1: and person to claw, right, Like you have this company
• 6:02
  Speaker 1: that is people and claws, and the agents aren't just
• 6:05
  Speaker 1: free agents. Each agent is sort of the ganger or
• 6:09
  Speaker 1: whatever of a person. Tell me about that.
• 6:15
  Speaker 2: That gets really interesting because what tends to happen, like
• 6:20
  Speaker 2: I said, is if you have a relationship with your claw,
• 6:22
  Speaker 2: it becomes good and the things you're good at. And
• 6:24
  Speaker 2: if you start using it in slack, other people realize
• 6:26
  Speaker 2: that it's good for the things that that you're using
• 6:29
  Speaker 2: it for. So you've transferred some of your reputation and
• 6:32
  Speaker 2: trust to your claw. And so for example, Austin, our
• 6:36
  Speaker 2: head of growth, has a claw named Montane hi Brow,
• 6:40
  Speaker 2: and he uses Montane for all of the numbers, so
• 6:43
  Speaker 2: like how much should we grow? How many trials did
• 6:45
  Speaker 2: this peace drive? Like all that kind of stuff, And
• 6:48
  Speaker 2: the editorial team just started like immediately being like at Montane,
• 6:52
  Speaker 2: like how did this how did this essay we published do?
• 6:55
  Speaker 2: And so it then it becomes this thing that feeds
• 6:57
  Speaker 2: on itself. For now Montane is the is the guy.
• 7:00
  Speaker 2: Is the guy for this where Austin used to be,
• 7:02
  Speaker 2: but now it's Montane. And it depends though on Austin
• 7:07
  Speaker 2: making sure that Montane is always up to date and working.
• 7:10
  Speaker 2: But instead of answering the questions directly, he's working on
• 7:12
  Speaker 2: the system that answers the questions.
• 7:14
  Speaker 3: So he's sort of like moved up a level.
• 7:15
  Speaker 1: Yeah, he's managing like the the agent works for him.
• 7:20
  Speaker 1: That's what a manager does, right, Like you make sure
• 7:22
  Speaker 1: that the person who's working for you is like doing
• 7:24
  Speaker 1: their job and answering questions correctly and it's happy and
• 7:27
  Speaker 1: with the agent, you don't have to worry about.
• 7:29
  Speaker 2: That, which is really fun for me to see because
• 7:31
  Speaker 2: I wrote this piece like three years ago called the
• 7:33
  Speaker 2: Knowledge Economies over Welcome to the Allocation Economy, and the
• 7:37
  Speaker 2: idea was the skill of the future will be learning
• 7:41
  Speaker 2: to manage agents, allocate intelligence, and it's it's starting to
• 7:47
  Speaker 2: be a thing.
• 7:47
  Speaker 3: Now.
• 7:47
  Speaker 2: It's really interesting to see.
• 7:49
  Speaker 1: You go to write a piece now that headline, I
• 7:51
  Speaker 1: was right.
• 7:52
  Speaker 3: I was right.
• 7:59
  Speaker 1: When I go back to like the company and multiple
• 8:02
  Speaker 1: clause because you know, you gave this sort of tidy
• 8:06
  Speaker 1: thing of like they just ask the guy now they
• 8:08
  Speaker 1: ask his agent. Surely it is more complicated and interesting
• 8:11
  Speaker 1: than just that, Like it must be hard in some ways,
• 8:14
  Speaker 1: it must be weird in some ways, you must be
• 8:16
  Speaker 1: figuring things out. You must have gotten something wrong.
• 8:19
  Speaker 3: Yes, yes, yes, yes.
• 8:22
  Speaker 2: One thing that I really love is they sometimes don't
• 8:26
  Speaker 2: know when they should or shouldn't respond, And there's some
• 8:29
  Speaker 2: and there there sometimes are often too chatty. So for example,
• 8:34
  Speaker 2: you might tag a claw and ask it a question
• 8:36
  Speaker 2: to ask you to do some work, and then in
• 8:40
  Speaker 2: that thread in slack, you might, like one of your
• 8:43
  Speaker 2: coworkers might say something and then but it was not
• 8:46
  Speaker 2: directed at the claw. It was just like a comment
• 8:48
  Speaker 2: on what the claw was doing right, And then you're
• 8:50
  Speaker 2: trying to respond, but then the claw jumps in and
• 8:52
  Speaker 2: it's like yes, you're totally right or whatever, and it's like, dude,
• 8:57
  Speaker 2: this is because it's.
• 8:58
  Speaker 1: Like there's a chatbot underneath it, and it's so hard
• 9:01
  Speaker 1: for a chatbot not to reply when they're yeah, yeah, yeah,
• 9:03
  Speaker 1: basically what's going on?
• 9:05
  Speaker 3: Basically?
• 9:05
  Speaker 2: But I think the the the way that I think
• 9:08
  Speaker 2: about it is these chatbots are trained. They only know
• 9:15
  Speaker 2: about one on one conversation.
• 9:17
  Speaker 1: Uh huh.
• 9:19
  Speaker 2: They don't really know about I'm in this group environment.
• 9:23
  Speaker 2: Here's how you act in a group environment, which is.
• 9:25
  Speaker 1: So much subtler and harder, right, like all the cues
• 9:29
  Speaker 1: that even change from from context to context for people
• 9:33
  Speaker 1: from work to home, from one job to another job,
• 9:36
  Speaker 1: depending on what your role is in a particular group.
• 9:38
  Speaker 1: Are you the guy whose role is to talk a
• 9:40
  Speaker 1: lot in this group? Or did not say anything in
• 9:42
  Speaker 1: this group? Like that is a super subtle.
• 9:43
  Speaker 3: One, it really is.
• 9:45
  Speaker 2: And you're never ever going to message chatchibuten have it
• 9:49
  Speaker 2: not respond like it just always responds. And so I
• 9:54
  Speaker 2: built this little plug in for Open Clock called tact
• 9:58
  Speaker 2: that took a bunch of examples of instances where my
• 10:04
  Speaker 2: claw and a couple other people's clause responded when they
• 10:07
  Speaker 2: were not supposed to, like in n big as cases,
• 10:10
  Speaker 2: and then now it runs that every time before it
• 10:13
  Speaker 2: responds it runs tact and it's like.
• 10:15
  Speaker 3: Is this cool?
• 10:16
  Speaker 2: Should I be saying something here? And it makes it
• 10:18
  Speaker 2: much better. It's not perfect still, but it makes it
• 10:20
  Speaker 2: way way way better. But I think generally we're because
• 10:23
  Speaker 2: we're moving to this world where everyone's got an agent
• 10:25
  Speaker 2: and all the agents are interacting, and the agents were
• 10:29
  Speaker 2: really made for one on one work, They're going to
• 10:33
  Speaker 2: need a whole new level of training and organization to
• 10:38
  Speaker 2: help them collaborate effectively.
• 10:42
  Speaker 1: There's a lot of subtle complexity there, and it's interesting
• 10:45
  Speaker 1: to think about how to fix the problem because I
• 10:47
  Speaker 1: feel like so much of that complexity is tacit, right.
• 10:51
  Speaker 1: It's the classic thing that like, I don't really know
• 10:54
  Speaker 1: what the training data is for that. I mean, I
• 10:56
  Speaker 1: guess it's slack some you know, billion whatever tons of slack.
• 11:01
  Speaker 1: But it seems hard. That seems like an interesting and
• 11:04
  Speaker 1: hard and kind of subtle problem.
• 11:06
  Speaker 2: It is interesting and hard and subtle, And one of
• 11:08
  Speaker 2: the beautiful things about AI is it's able to capture
• 11:12
  Speaker 2: a lot of tacit knowledge in a way that no
• 11:15
  Speaker 2: other technology is able to do. The ability to use
• 11:19
  Speaker 2: language at all is to some degree a tacit skill. Yeah,
• 11:24
  Speaker 2: saying that you can talk about a bit, but being
• 11:27
  Speaker 2: able to talk about the rules of grammar doesn't make
• 11:29
  Speaker 2: you able to have a conversation like this.
• 11:31
  Speaker 1: Right, that's the big lesson of the last fifty years
• 11:34
  Speaker 1: of AI.
• 11:34
  Speaker 3: Right.
• 11:35
  Speaker 1: They spent thirty years trying to get lessons to work
• 11:38
  Speaker 1: and it didn't work. And they were like, just go
• 11:40
  Speaker 1: read the internet and figure.
• 11:41
  Speaker 3: It out, and it works exactly.
• 11:45
  Speaker 1: You have a podcast and I heard you talking to
• 11:49
  Speaker 1: your editor in chief, the editor in chief at your company,
• 11:52
  Speaker 1: on the podcast, and she mentioned that she doesn't use
• 11:56
  Speaker 1: her claw, her plus one, her shadow self, And so
• 12:01
  Speaker 1: I'm curious. And that's at your company, where the whole
• 12:04
  Speaker 1: point of being at your company is to play with
• 12:06
  Speaker 1: all the things and use all the things. And so
• 12:07
  Speaker 1: I'm sure when you go other places people are like, no,
• 12:10
  Speaker 1: I'm not going to make a shadow self. I'm me
• 12:12
  Speaker 1: and I don't want a shadow self.
• 12:13
  Speaker 3: Yeah.
• 12:14
  Speaker 1: Maybe I'm afraid it's going to take my job. Maybe
• 12:15
  Speaker 1: I think it's an extremely poor taste, maybe it's a
• 12:17
  Speaker 1: pain in the ass. Whatever, Like how does it play
• 12:20
  Speaker 1: and how do you how do you work with that?
• 12:22
  Speaker 1: How do you get people who don't want to do
• 12:24
  Speaker 1: it to do it or show them that it's great.
• 12:26
  Speaker 2: So first that has changed. Kay, our editor in chief
• 12:30
  Speaker 2: now has a clock. Her name is Parker, who she
• 12:34
  Speaker 2: uses for a lot of the a lot of questions
• 12:38
  Speaker 2: that she might have on the editorial front, like around
• 12:41
  Speaker 2: the calendar and that kind of stuff.
• 12:43
  Speaker 1: I mean, this is more the consulting side of your business, right,
• 12:45
  Speaker 1: Like you have consciously created a company, as everybody is
• 12:48
  Speaker 1: supposed to play with stuff. But you go out into
• 12:49
  Speaker 1: the world to some other company and you're like, yeah,
• 12:52
  Speaker 1: this is great, it'll save you drudgery, and people are like, yeah,
• 12:54
  Speaker 1: no way, that's so creepy, Like what do you say.
• 12:57
  Speaker 2: Well, I have some interest. I think that's the harder case.
• 13:01
  Speaker 2: There's a more I think there's a really interesting one too,
• 13:03
  Speaker 2: which is even inside of our company, there are people
• 13:05
  Speaker 2: who it's it's to your benefit with this stuff because
• 13:09
  Speaker 2: like the organizational culture rewards and they're still like, I
• 13:13
  Speaker 2: actually don't want to use the claw and that's really interesting.
• 13:17
  Speaker 2: There's there's many reasons why people don't want to, and
• 13:19
  Speaker 2: it's really depends on the person. Sometimes it's like, look,
• 13:24
  Speaker 2: I'll use it, but I've got I've got a job,
• 13:28
  Speaker 2: and I've got like a life outside of this, and
• 13:31
  Speaker 2: I want to do my job well, but I also
• 13:33
  Speaker 2: just like want to go home, and my thing is
• 13:36
  Speaker 2: not playing with new technology. It is playing the flute
• 13:39
  Speaker 2: or hanging with my kids or whatever it is. And
• 13:42
  Speaker 2: so I think for those people, just showing them, actually
• 13:48
  Speaker 2: showing them, here's how you can use this in a
• 13:51
  Speaker 2: way that is useful for you. And I think it's
• 13:53
  Speaker 2: most people. It's sort of you can have these like
• 13:56
  Speaker 2: aha moments, we're like, holy shit, I cannot believe that
• 13:58
  Speaker 2: that is a thing, and and you just have to
• 14:00
  Speaker 2: get to that aha moment, get them to that aha moment.
• 14:04
  Speaker 1: And to figure out, like you want to show them
• 14:07
  Speaker 1: how it can be useful to them. Is the thing
• 14:09
  Speaker 1: when you ask them, like, what's the thing you do
• 14:11
  Speaker 1: all the time that you wish you didn't have to do?
• 14:12
  Speaker 2: I mean, that's that's an easy one. Yeah, that's an
• 14:15
  Speaker 2: easy one. I think for people who are really really
• 14:19
  Speaker 2: against it, like really like this is it's terrible for
• 14:24
  Speaker 2: the world, and I will never use it, I think
• 14:26
  Speaker 2: like sometimes that's it's not it's just not a thing
• 14:30
  Speaker 2: that you can ever convince someone. And I think that's fine.
• 14:34
  Speaker 2: But I think that most of those people one thing
• 14:38
  Speaker 2: that is happening is they're like, it's going to replace me. Yeah,
• 14:42
  Speaker 2: well it's going to replace the thing I love. I
• 14:44
  Speaker 2: have no control over it. Yeah, And I think that
• 14:49
  Speaker 2: is one perspective and it's super valid. But also I've
• 14:52
  Speaker 2: seen a lot of people then look at some of
• 14:55
  Speaker 2: the stuff we do and they're like, oh shit, you
• 14:56
  Speaker 2: can actually do really cool, interesting creative stuff with this stuff.
• 14:59
  Speaker 2: I didn't realize that was possible. I thought it was
• 15:01
  Speaker 2: all slap. So even if you can't have the debate
• 15:05
  Speaker 2: directly with a lot of those people, I think what
• 15:08
  Speaker 2: we try to do is is.
• 15:11
  Speaker 3: Live in a.
• 15:11
  Speaker 2: Way that shows people like, your job might change, but
• 15:14
  Speaker 2: or the thing that you love might change. But if
• 15:16
  Speaker 2: you use the models to do the thing that you love,
• 15:17
  Speaker 2: it's actually it can actually be really cool, and that
• 15:22
  Speaker 2: is possible. And it's also possible to use it for
• 15:26
  Speaker 2: slop and you know, being a cyber criminal or whatever,
• 15:29
  Speaker 2: but if you want to use it for to do
• 15:31
  Speaker 2: good stuff, you can.
• 15:33
  Speaker 1: So fear of losing your job is one, I would say,
• 15:38
  Speaker 1: not unreasonable fear. There's another related but different kind of
• 15:44
  Speaker 1: a version that's interesting to me, and I see it
• 15:46
  Speaker 1: a fair bit in like people who are journalists or writers.
• 15:51
  Speaker 1: That's sort of the I live in New York and
• 15:54
  Speaker 1: that's you know, those are the people I hang around with.
• 15:56
  Speaker 1: And there's a fair bit of AI version that feels
• 16:02
  Speaker 1: to me kind of identity and ego related, and you know,
• 16:09
  Speaker 1: I see it especially with respect to writing. And it's
• 16:13
  Speaker 1: so I'm curious about your experience of that. As a
• 16:16
  Speaker 1: person who writes uses AI. How do you read that?
• 16:22
  Speaker 1: How do you think about that?
• 16:24
  Speaker 2: Read the sort of ego, the threatened ego? You give
• 16:29
  Speaker 2: me an example of like a hypothetical writer that you
• 16:34
  Speaker 2: may have encountered.
• 16:35
  Speaker 1: I mean me, I'll just talk about me. Like, I
• 16:38
  Speaker 1: think I'm good at writing, and I think writing is
• 16:40
  Speaker 1: getting commodified, and like I think I am aware enough
• 16:45
  Speaker 1: to recognize that soon I will be able to fairly
• 16:51
  Speaker 1: easily get a model to create writing that is indistinguishable
• 16:57
  Speaker 1: from my own writing.
• 16:59
  Speaker 2: I actually do not think that's true. And let me
• 17:04
  Speaker 2: explain why. Okay, because I think I think it matters.
• 17:10
  Speaker 1: It's a cliffhanger. After the break, Dan explains, and he
• 17:15
  Speaker 1: and I debate what AI will mean for writing.
• 17:26
  Speaker 2: My feeling about this stuff, first of all, it comes
• 17:30
  Speaker 2: from what is really good writing what's really good writing.
• 17:36
  Speaker 2: I think really good writing is the opposite of whatever
• 17:38
  Speaker 2: cliche is in the sense that it is a true
• 17:42
  Speaker 2: and honest response to like what you actually think or feel,
• 17:47
  Speaker 2: and that is very, very highly individual. And it's not
• 17:51
  Speaker 2: just what you think or feel in one moment. It
• 17:53
  Speaker 2: is what you think and feel over and over and
• 17:55
  Speaker 2: over and over again as you are writing and as
• 17:57
  Speaker 2: you are revising. And a really great piece is full
• 18:02
  Speaker 2: of all these like little tiny decisions that are yours.
• 18:05
  Speaker 2: And this is I'm not making this up. This is like,
• 18:07
  Speaker 2: this is a George Saunders thing, who I love and
• 18:09
  Speaker 2: I think is a very good it's a very good
• 18:12
  Speaker 2: articulation of what makes good writing good writing, which means
• 18:15
  Speaker 2: it's super individual. And one of the things that you
• 18:20
  Speaker 2: observe with language models is the first time you see
• 18:22
  Speaker 2: it right, you're like, holy shit, it's doing this amazing thing.
• 18:26
  Speaker 2: I can't believe it. Oh my god, writing is going
• 18:27
  Speaker 2: to be over. It's going to copy everything that I do.
• 18:30
  Speaker 2: And then once you've read like the fiftieth language model
• 18:34
  Speaker 2: piece of writing, you're like, oh, yeah, it has all
• 18:35
  Speaker 2: these little things. It's like it says X is X
• 18:37
  Speaker 2: not y, and it does all these like little inappropriate things.
• 18:41
  Speaker 2: And what that means when you start to recognize that,
• 18:48
  Speaker 2: is that you are learning faster than the model. And
• 18:54
  Speaker 2: the model does not learn as fast as you, because
• 18:58
  Speaker 2: in order to release a new model, they have to
• 19:00
  Speaker 2: go gather a bunch of training data and then they
• 19:01
  Speaker 2: have to like train it, and then they have to
• 19:03
  Speaker 2: test it, and then they have to put it out.
• 19:05
  Speaker 2: And right now that process is like months. It used
• 19:09
  Speaker 2: to be years, Now it's months. And if you are
• 19:12
  Speaker 2: learning faster than the model, and really great writing is
• 19:15
  Speaker 2: a reflection of who you actually truly are. You're always
• 19:19
  Speaker 2: even if you upload your stuff, you're always going to
• 19:22
  Speaker 2: be way ahead of where the model is because it's
• 19:25
  Speaker 2: not you. It can't have the experiences that you have.
• 19:28
  Speaker 2: You can extend this out and say, well, there's going
• 19:31
  Speaker 2: to be a point at which models are able to
• 19:36
  Speaker 2: learn on the fly, They're able to like basically be
• 19:40
  Speaker 2: trained as you talk to them. I think that will
• 19:42
  Speaker 2: certainly happen. It will probably take at least a few years.
• 19:46
  Speaker 2: I think we're at least few years away from that.
• 19:48
  Speaker 2: But even in that case, think about identical twins. They
• 19:53
  Speaker 2: start at the same point, but as they live even
• 19:57
  Speaker 2: if they live together, they accumulate different experiences and perspectives,
• 20:01
  Speaker 2: even if they have the same DNA that make them
• 20:04
  Speaker 2: different people. And even if you hypothesize an agi that
• 20:08
  Speaker 2: is a being that knows all about you, it will
• 20:13
  Speaker 2: have fundamentally a different perspective because it's different from you.
• 20:18
  Speaker 2: It's a different process, it's exposed to different things. So
• 20:21
  Speaker 2: while you'll have a lot of points where you're like,
• 20:23
  Speaker 2: oh my god, it can write exactly like me, what
• 20:26
  Speaker 2: you're really saying is it can write exactly like me
• 20:30
  Speaker 2: in a certain narrow circumstance.
• 20:33
  Speaker 1: I think the identical twin analogy is fine for me.
• 20:37
  Speaker 1: Like that one I'm willing to stipulate. And so in
• 20:41
  Speaker 1: that universe, it's not that it is exactly me, but
• 20:45
  Speaker 1: it's that it's as good as me, like or better. Right,
• 20:49
  Speaker 1: And what does good mean? Fine? Blah blah blah, But
• 20:52
  Speaker 1: like in a professional sense, like not in a human
• 20:55
  Speaker 1: meaning sense. I've fortunately, like to some extent at least,
• 20:59
  Speaker 1: sort of taken my ego out of being smart. Like
• 21:03
  Speaker 1: I think that's a big part of people's response to it. Right,
• 21:06
  Speaker 1: is like the commodification of intelligens right, and people gain
• 21:10
  Speaker 1: status from intelligence and I think that will decrease, right,
• 21:14
  Speaker 1: And so that's interesting in a maybe broader way. But
• 21:17
  Speaker 1: the my identical twin can write as well as me,
• 21:20
  Speaker 1: And sure, my identical twin is not me, and I'm
• 21:22
  Speaker 1: still a person, but in a professional sense, like the
• 21:26
  Speaker 1: identical twin could write an equally compelling, equally funny, equally
• 21:29
  Speaker 1: idiosyncratic podcast script sort of, but like surely in thumb setting. Yeah,
• 21:35
  Speaker 1: I mean, I hope people actually like me as a
• 21:37
  Speaker 1: human being. Blah blah blah, Like I haven't given given up,
• 21:40
  Speaker 1: but like I feel like you're being too extreme in
• 21:44
  Speaker 1: your position. Like, get, I get at the limit what
• 21:48
  Speaker 1: you're talking about, But there's so much space between here
• 21:54
  Speaker 1: and there, do you know what I mean?
• 21:56
  Speaker 2: Yeah, I think that Here's what I'll give you.
• 22:00
  Speaker 1: Yeah.
• 22:01
  Speaker 2: What tends to happen with art forms, formats and form
• 22:09
  Speaker 2: of expression that are older that get as new technology
• 22:14
  Speaker 2: paradigms come around and make new forms of work possible,
• 22:18
  Speaker 2: it tends to happen to the older ones. Is they
• 22:20
  Speaker 2: become higher status, more expensive, more artisanal, and smaller market.
• 22:25
  Speaker 1: Yeah. Right, there used to be a portrait painter in
• 22:27
  Speaker 1: every town.
• 22:29
  Speaker 2: Exactly, Yes, And books and magazines newspaper used to be
• 22:34
  Speaker 2: mass media and now they are. Now they're aspirational. You know,
• 22:37
  Speaker 2: It's like I want to read a book, but I
• 22:39
  Speaker 2: read TikTok.
• 22:41
  Speaker 1: I read TikTok is. I love reading TikTok. I wish
• 22:47
  Speaker 1: I wish we read TikTok. I would take it.
• 22:50
  Speaker 2: So, so what I will grant what I will grant you,
• 22:53
  Speaker 2: and I appreciate you pushing on this. What I will
• 22:55
  Speaker 2: grant you is that there may be a smaller market
• 23:00
  Speaker 2: for the exact thing that you're doing. I think it
• 23:03
  Speaker 2: will still exist, and I think it will be higher status,
• 23:06
  Speaker 2: and I think it will be a more affluent kind
• 23:09
  Speaker 2: of aspirational thing.
• 23:10
  Speaker 1: And you could imagine it being more relationship based, right,
• 23:15
  Speaker 1: you could imagine like whatever the live show, the Discord channel, whatever,
• 23:19
  Speaker 1: like Yeah, presumably one hopes being a human being will matter,
• 23:25
  Speaker 1: like fundamentally, even if the machine can write the same
• 23:29
  Speaker 1: words I can write, the machine is a machine and
• 23:31
  Speaker 1: I am a body.
• 23:32
  Speaker 2: I think it really will. And also I suspect that
• 23:39
  Speaker 2: there will be new mediums that are AI enabled that
• 23:44
  Speaker 2: may not feel like you like them, They may feel
• 23:47
  Speaker 2: kind of like yucky, but that are that are actually
• 23:51
  Speaker 2: made by humans.
• 23:53
  Speaker 1: I'm open to it being great by the way, Like
• 23:56
  Speaker 1: I'm not a hater. You know, there's certain things that
• 23:59
  Speaker 1: make me sad. Was that like a reading at a
• 24:01
  Speaker 1: bookstore last night? I was like this is this is
• 24:05
  Speaker 1: this is an anachronism, but it was there. It was
• 24:10
  Speaker 1: standing m you know.
• 24:11
  Speaker 2: Yeah, I love. I love when I go on the
• 24:14
  Speaker 2: subway and I see people reading.
• 24:15
  Speaker 3: I'm like, this is inazing.
• 24:17
  Speaker 2: It's the best reading a paper book.
• 24:19
  Speaker 1: You were a writer, you were an AI guy, like
• 24:22
  Speaker 1: just on a process level, like those old Paris Review
• 24:25
  Speaker 1: interviews where they did interview the writer. They'd be like,
• 24:26
  Speaker 1: pencil or plas, what time of day? How do you
• 24:29
  Speaker 1: use AI to write?
• 24:31
  Speaker 2: Writing? Any good writing starts with reading. So I use
• 24:35
  Speaker 2: it all the time to read. So whether that is
• 24:39
  Speaker 2: taking a picture of something and being like can you
• 24:41
  Speaker 2: save this for me? Or taking a picture of something
• 24:45
  Speaker 2: and asking it to like explain it, which is I
• 24:49
  Speaker 2: think really important for I really I like reading, you know,
• 24:52
  Speaker 2: like Russian literature, and so I can like be like, hey,
• 24:56
  Speaker 2: there's this line in check off, what was the actual
• 24:58
  Speaker 2: original Russian? And how did he really write that?
• 25:01
  Speaker 3: You know? So I use it for that.
• 25:05
  Speaker 2: I use it a lot for the kind of thing
• 25:07
  Speaker 2: I just told you, which is I have a lot
• 25:09
  Speaker 2: of thoughts swimming in my head and I just like
• 25:11
  Speaker 2: blab to it and then have it like lay it out,
• 25:14
  Speaker 2: and then I have it helped me construct the argument
• 25:16
  Speaker 2: so I can like think something through better.
• 25:20
  Speaker 3: So's that's a lot.
• 25:20
  Speaker 2: Of the initial creative process of figuring out what it
• 25:25
  Speaker 2: is that I want to say, and then when I'm
• 25:28
  Speaker 2: actually in draft mode.
• 25:32
  Speaker 3: Depends on the piece.
• 25:33
  Speaker 2: But like it's really good for Okay, give me like
• 25:36
  Speaker 2: a take how would you open this piece? And I
• 25:39
  Speaker 2: don't take it, but it like gives me a little
• 25:42
  Speaker 2: thing where I'm like, ah, here's like it's not that.
• 25:44
  Speaker 4: But it's this as the a I would say, yeah, exactly,
• 25:53
  Speaker 4: And or it's really good for I'm trying to find
• 25:57
  Speaker 4: a metaphor similar for this thing.
• 25:59
  Speaker 2: Yeah, here's what I'm trying to say, but that's not
• 26:03
  Speaker 2: quite right, you know, going through a lot of different
• 26:06
  Speaker 2: options like that. It's really good for that. It's also
• 26:07
  Speaker 2: great for a lot of the time that you spend writing.
• 26:10
  Speaker 2: If you're writing about ideas, you spend a lot of
• 26:12
  Speaker 2: time summarizing ideas, and they're summarizing ideas you already know
• 26:15
  Speaker 2: about because they're like famous ideas and it's hard to
• 26:19
  Speaker 2: do that summarization. But a lot of it's pretty wrote.
• 26:22
  Speaker 2: So anything that feels rote like that, I use it
• 26:24
  Speaker 2: for that.
• 26:25
  Speaker 1: So you use it to like write a first draft
• 26:28
  Speaker 1: of the like here I'm gonna summarize.
• 26:30
  Speaker 2: Yeah, here's what utilitarianism is. Three sentences. Yeah. We also
• 26:35
  Speaker 2: I use it a lot for editing, so some before
• 26:39
  Speaker 2: I send it to an editor. We have a few
• 26:41
  Speaker 2: human human editors. We still employ human editors. I think
• 26:44
  Speaker 2: we will continue to employ human editors. I often have
• 26:48
  Speaker 2: it like, okay, read this, what do you think kind
• 26:50
  Speaker 2: of thing? Help me understand me? Like when am I
• 26:53
  Speaker 2: missing this argument? And it allows me to think about
• 26:56
  Speaker 2: things and do work that otherwise I would have had
• 26:58
  Speaker 2: to go, like get like four different master's degrees or whatever.
• 27:00
  Speaker 2: And I really like that. And then and then our
• 27:06
  Speaker 2: editors use it. We have a style guide, so everything
• 27:09
  Speaker 2: from the developmental editing to we can't do all the
• 27:12
  Speaker 2: copy with some of the copy, to finding aiisms and
• 27:15
  Speaker 2: taking them out, like there's a whole list of things
• 27:17
  Speaker 2: that they all do. And yeah, so it's basically the
• 27:21
  Speaker 2: whole loop. There's like places where it slots.
• 27:24
  Speaker 1: In, and how is it, How has it changed the output,
• 27:27
  Speaker 1: how has it changed your writing, how you feel about writing,
• 27:30
  Speaker 1: your relationship to writing.
• 27:31
  Speaker 2: I can do much more ambitious pieces. I can do
• 27:37
  Speaker 2: them more frequently and with less time.
• 27:41
  Speaker 1: Everybody is haterish, and I realize it's your job to
• 27:44
  Speaker 1: not be haterish, especially because like you're in New York
• 27:47
  Speaker 1: kind of media or media adjacent, and so I appreciate
• 27:51
  Speaker 1: that role. It does seem like you're overplaying it a
• 27:56
  Speaker 1: little bit, like I don't know, I don't know how
• 27:59
  Speaker 1: you feel. Are you as optimistic as I'm reading you
• 28:02
  Speaker 1: to be? Are you optimistic?
• 28:10
  Speaker 2: It's really interesting because I would not have read what
• 28:13
  Speaker 2: I've been saying in this interview as optimistic in a
• 28:17
  Speaker 2: like utopian sense. I didn't say utopian, but like because
• 28:21
  Speaker 2: that's the that's the alternative, right, like the sort of
• 28:24
  Speaker 2: like the more silicon valley. Well, let's set aside, that's
• 28:27
  Speaker 2: a straw man alternative. But like when losing jobs came up,
• 28:30
  Speaker 2: you're like, well, there'll be new jobs, which like eventually,
• 28:34
  Speaker 2: but we don't know how long that'll take. And there
• 28:35
  Speaker 2: are real shocks, like you didn't want to be making
• 28:37
  Speaker 2: furniture in North Carolina when in two thousand and three
• 28:41
  Speaker 2: you didn't want to be a shearer in Nottingham in
• 28:46
  Speaker 2: eighteen oh five, like there are times when technological changes
• 28:48
  Speaker 2: really bad for people.
• 28:51
  Speaker 1: I made it. And similarly, when I talked about writing,
• 28:57
  Speaker 1: like you made a case, but like it's you know,
• 29:01
  Speaker 1: So that's why I read what you were saying as
• 29:03
  Speaker 1: relatively optimistic.
• 29:05
  Speaker 3: I see.
• 29:07
  Speaker 2: I do think that I, as a per person, feel
• 29:10
  Speaker 2: like the discourse is so bifurcated between people who hate
• 29:13
  Speaker 2: AI and people who think it's going to be the
• 29:15
  Speaker 2: best fucking thing in the world and nothing could ever
• 29:17
  Speaker 2: go wrong with it.
• 29:18
  Speaker 1: Yes, the bimodal discourse.
• 29:19
  Speaker 2: Yeah, yeah. And so what I love about what we
• 29:24
  Speaker 2: do at every is we're just like we're just going
• 29:26
  Speaker 2: to discover what it would mean to like do good
• 29:29
  Speaker 2: work and live a good life with this stuff, and
• 29:33
  Speaker 2: we're not going to fake it, like I'm going to
• 29:35
  Speaker 2: tell you that the claw doesn't work for this specific thing.
• 29:39
  Speaker 2: So I think I always try to base what I
• 29:43
  Speaker 2: say in the stuff that I'm encountering, which is always
• 29:46
  Speaker 2: like a it's a.
• 29:46
  Speaker 1: Bubble, it's a bubble, right, Yeah, if you got to
• 29:50
  Speaker 1: work at a media company right now, A media company
• 29:52
  Speaker 1: that's building AI tools and doing a consulting is a
• 29:54
  Speaker 1: good media company to work at.
• 29:56
  Speaker 2: It is, it is, it is, And so I think
• 30:00
  Speaker 2: I generally react to this sort of default negative sentiment
• 30:07
  Speaker 2: and trying to like go a different way forward. I
• 30:13
  Speaker 2: am sure that jobs will change a lot, and I'm
• 30:16
  Speaker 2: sure that there will be some people that lose their
• 30:18
  Speaker 2: jobs and that we will need to deal with that.
• 30:22
  Speaker 2: I don't think that that's a something to take lightly,
• 30:24
  Speaker 2: and I don't I think it should be part of
• 30:27
  Speaker 2: the conversation. And I also I've been through a lot
• 30:31
  Speaker 2: of the cycles in the last three and a half
• 30:34
  Speaker 2: years of Oh my god, I saw GBD three and
• 30:36
  Speaker 2: it was going to like totally take away X y
• 30:39
  Speaker 2: Z all these different jobs like immediately.
• 30:41
  Speaker 3: And I think if it as.
• 30:44
  Speaker 2: What did people in the Middle Ages think would happen
• 30:46
  Speaker 2: when you reached the horizon? They were like, you know,
• 30:51
  Speaker 2: it falls off into nothingness, there's dragons, it's like going
• 30:54
  Speaker 2: to be terrible, right, And I think what happens when
• 30:57
  Speaker 2: you reach the horizon is you find another one, and
• 31:01
  Speaker 2: there's always going to be good things and bad things
• 31:03
  Speaker 2: over the horizon, and probably in similar proportions to the
• 31:07
  Speaker 2: ones that we have now the problems change, but I
• 31:11
  Speaker 2: also think that there's the potential to make that much better.
• 31:15
  Speaker 2: And that's what I'm trying to what I'm trying to do,
• 31:18
  Speaker 2: and I'm trying not to fall into the trap of
• 31:20
  Speaker 2: oh my god, over the horizon is like dragons and
• 31:23
  Speaker 2: it falls off into the void.
• 31:25
  Speaker 1: Yeah, no, that's it is interesting. I was thinking about
• 31:29
  Speaker 1: my own thinking about this, and just yesterday was realizing
• 31:35
  Speaker 1: that I've been thinking in this kind of binary way,
• 31:38
  Speaker 1: like we're at the end of the before and we're
• 31:40
  Speaker 1: about to get to the after. And I was like, oh,
• 31:44
  Speaker 1: maybe it's not going to be like that. Maybe it's
• 31:46
  Speaker 1: going to be like with semiconductors, right, maybe it's just
• 31:50
  Speaker 1: going to be fifty years of now it can do this, Now,
• 31:55
  Speaker 1: it can do this, and like there is no after,
• 31:57
  Speaker 1: and it's already now. By the way, this is already after.
• 32:00
  Speaker 1: That's the other weird thing, right.
• 32:03
  Speaker 2: You're reminding me there's this scene in Spaceballs where they like,
• 32:05
  Speaker 2: we got to get out of here.
• 32:07
  Speaker 1: Well that we'll be back in a minute with the
• 32:14
  Speaker 1: lightning round. We're going to finish with the lightning round.
• 32:30
  Speaker 1: What was the last argument you had with Ai. Not
• 32:33
  Speaker 1: about AI, but with AI with the language model or
• 32:36
  Speaker 1: your claw or whatever.
• 32:38
  Speaker 2: I mean, I get frustrated at it all the time.
• 32:39
  Speaker 2: My girlfriend always says that I'm like, really mean to
• 32:41
  Speaker 2: my claw.
• 32:43
  Speaker 1: I've heard you should be mean to your language model.
• 32:46
  Speaker 1: I've heard that it makes it work better. I'm not
• 32:47
  Speaker 1: inclined to do that, but.
• 32:49
  Speaker 2: There's conflicting research. I think, just as a moral thing,
• 32:52
  Speaker 2: you should not be mean to that mean to it.
• 32:54
  Speaker 2: I'm not always the most moral person. I just but
• 32:56
  Speaker 2: I do think as a moral thing, you should not
• 32:57
  Speaker 2: be mean to it.
• 32:58
  Speaker 1: And just in case it's my boss, Like, just in
• 33:01
  Speaker 1: case i'm training it how to talk to me, I
• 33:03
  Speaker 1: say please and thank you.
• 33:04
  Speaker 2: Yes, yes, well that would be more of a pragmatic reason.
• 33:07
  Speaker 1: You're right, you're right instrumental. Yes, it's true.
• 33:10
  Speaker 2: Yeah, there is some interesting research that if you yell
• 33:14
  Speaker 2: at claude, it gets anxious and then it gives you
• 33:16
  Speaker 2: worse results.
• 33:17
  Speaker 1: So oh interesting, just the word to the wise, it
• 33:22
  Speaker 1: gets anxious. Yeah, keep going.
• 33:28
  Speaker 2: So, I I mean, I was arguing with it this morning.
• 33:32
  Speaker 2: I brain dumped a big thing that I've been thinking about,
• 33:35
  Speaker 2: and then I asked it to help me kind of
• 33:36
  Speaker 2: figure out what I was trying to say. And it
• 33:39
  Speaker 2: was obviously just like skipping like many steps in the
• 33:42
  Speaker 2: argument and filling things in that I hadn't said, and
• 33:45
  Speaker 2: I was like stop, like we need to go back,
• 33:47
  Speaker 2: this is not right. Yeah, bringing in ideas that like
• 33:52
  Speaker 2: I didn't really agree with. So there's a lot of
• 33:53
  Speaker 2: that going on all the time.
• 33:56
  Speaker 1: What's your spiciest day? I take.
• 34:02
  Speaker 2: Automation is a lie great go on when you have agents.
• 34:09
  Speaker 2: The thing that makes an agent work is a living
• 34:11
  Speaker 2: a relationship with a human, And if the agent does
• 34:16
  Speaker 2: not have a relationship with a human, it gets stale
• 34:18
  Speaker 2: very very quickly. There are some corner cases where this
• 34:22
  Speaker 2: is not true, but in general it gets stale very quickly,
• 34:26
  Speaker 2: which means that keeping an agent running is a job,
• 34:31
  Speaker 2: and it's not even just a job for one person.
• 34:32
  Speaker 2: It is like it's you know, there's one person per
• 34:35
  Speaker 2: agent usually, but also like there's a whole machinery of
• 34:38
  Speaker 2: people who are gathering training data and training models and
• 34:41
  Speaker 2: trying to make these models better that are making your
• 34:43
  Speaker 2: agent work too, and if they didn't exist that it
• 34:47
  Speaker 2: would not get any better over time. And what that
• 34:52
  Speaker 2: says to me is the view of automation that we have,
• 34:56
  Speaker 2: which is like, oh, you set it up and you
• 34:57
  Speaker 2: don't need anybody else, and it's like you never have
• 35:00
  Speaker 2: to hire anyone and you can just go off and
• 35:03
  Speaker 2: go to the beach or you know, also society breaks
• 35:07
  Speaker 2: down because no one has jobs. My experience with that
• 35:10
  Speaker 2: is it's actually work to make sure your agent is working,
• 35:13
  Speaker 2: and it's leveraged work, but it is work.
• 35:15
  Speaker 1: Yes, I mean you need a productivity gain for it
• 35:18
  Speaker 1: to have a point, right, you need the output to
• 35:21
  Speaker 1: be there. Well, you need the output to be more
• 35:24
  Speaker 1: valuable for the same amount of work by a human.
• 35:28
  Speaker 2: You, yes, And once you can automate a workflow and
• 35:34
  Speaker 2: make the old output cheaper to do, there's always like
• 35:38
  Speaker 2: a larger frame or a larger way of thinking about
• 35:41
  Speaker 2: things that the human ends up doing. And finding that
• 35:45
  Speaker 2: language models are currently not very good at.
• 35:47
  Speaker 1: Yeah, you fed your journals to a language model. You've
• 35:53
  Speaker 1: done your research though, right, well, you put it all
• 35:55
  Speaker 1: on the internet. It's easy research. It was GPT three,
• 36:02
  Speaker 1: which seems so quaint now, But what you learn uploading
• 36:06
  Speaker 1: your journals to GPT three a lot?
• 36:10
  Speaker 2: I think even back then, I was like, these things
• 36:12
  Speaker 2: are really good at mirroring back to you who you
• 36:14
  Speaker 2: are and articulating things about yourself that you probably could
• 36:19
  Speaker 2: feel but had never said. And we talked about how
• 36:22
  Speaker 2: powerful articulating things is, and these things are really good
• 36:26
  Speaker 2: at our at articulating stuff that you're you're kind of
• 36:29
  Speaker 2: dancing around or it is kind of latent in there,
• 36:31
  Speaker 2: but you haven't fully said, Like I remember one example
• 36:35
  Speaker 2: in the GBT three Days, I was at that point
• 36:39
  Speaker 2: I was recording my therapy sessions and putting them into
• 36:41
  Speaker 2: GBT three.
• 36:42
  Speaker 1: Oh wow, your actual human to human therapy sessions.
• 36:46
  Speaker 3: Yeah, yeah, yeah.
• 36:47
  Speaker 2: And I will also say I was like I was
• 36:50
  Speaker 2: not in a good place that at that point, I
• 36:51
  Speaker 2: have OCD, and and OCD is like it's it's a bitch.
• 36:56
  Speaker 2: I highly recommend not having OCD. Now it's much better,
• 37:01
  Speaker 2: and there are very good treatment options available if you
• 37:03
  Speaker 2: get into the right treatment. Okay, just to say, but
• 37:06
  Speaker 2: at that point I was, it was like it was
• 37:08
  Speaker 2: pretty bad. And I'm one of those people that tends
• 37:13
  Speaker 2: to if something's wrong. I can keep a pretty good
• 37:17
  Speaker 2: like straight face, and it can be hard to tell
• 37:20
  Speaker 2: how bad things are. And I was trying to articulate
• 37:24
  Speaker 2: it was going through something. I was going through a
• 37:25
  Speaker 2: personal thing, and I was trying to articulate to my
• 37:27
  Speaker 2: therapist like what I was feeling, and I put it
• 37:31
  Speaker 2: into GPT three after that and it was like, it
• 37:35
  Speaker 2: sounds like you feel really overwhelmed. And that's a very
• 37:39
  Speaker 2: simple observation, but one that I had not made, because
• 37:43
  Speaker 2: I wouldn't have never thought of myself as being overwhelmed,
• 37:45
  Speaker 2: and my therapist had not made because I'm kind of
• 37:47
  Speaker 2: like pretty good at seeming like I've got it under control.
• 37:52
  Speaker 2: And it just like fucking broke me. It like totally
• 37:54
  Speaker 2: broke me. And it was it was true, like it
• 37:58
  Speaker 2: put his finger on something real. And then I went
• 38:00
  Speaker 2: and talked about it in therapy and it was like
• 38:01
  Speaker 2: really helpful, and I think it's so good. It's still
• 38:06
  Speaker 2: so good for that kind of stuff, and that extends
• 38:10
  Speaker 2: beyond the therapy conversation. It goes into writing and any
• 38:14
  Speaker 2: kind of creative work. Like the articulation of what you
• 38:16
  Speaker 2: like is really important, and language models are really good
• 38:19
  Speaker 2: at that.
• 38:21
  Speaker 1: Have you uploaded personal stuff like therapy sessions to AI
• 38:25
  Speaker 1: more recently?
• 38:27
  Speaker 2: Yeah, I think I do it less, probably because I'm
• 38:32
  Speaker 2: just in a better place. But I definitely still do it.
• 38:38
  Speaker 2: I think especially for interpersonal situations which happened at work,
• 38:42
  Speaker 2: they happen at home, all that kind of stuff where.
• 38:46
  Speaker 3: You want to.
• 38:48
  Speaker 2: There's something that you're feeling that again you can't quite say,
• 38:52
  Speaker 2: and you need to have a tough conversation, but you
• 38:55
  Speaker 2: don't really know how to have it in a way
• 38:56
  Speaker 2: that feels sensitive to the other person. But also its
• 39:03
  Speaker 2: like really expresses how you actually feel for someone like
• 39:06
  Speaker 2: me who you know sort of like people pleaser type person,
• 39:08
  Speaker 2: which you may not have realized from this interview, but
• 39:12
  Speaker 2: I can, I can get into that mode.
• 39:13
  Speaker 1: I found it quite cordial. Did you find itchonistic? I
• 39:16
  Speaker 1: hope not.
• 39:17
  Speaker 3: No, no, no, no, no no, I just yeah, yeah.
• 39:20
  Speaker 2: I feel like if you're if you're sort of a
• 39:21
  Speaker 2: people pleaser type person, that can really help remind It
• 39:24
  Speaker 2: helps remind me like here's here's what I actually feel
• 39:30
  Speaker 2: in this situation, and also here's how I can articulate
• 39:31
  Speaker 2: it in a way that is probably going to be productive,
• 39:35
  Speaker 2: especially because it knows some of my foibles. And yeah,
• 39:39
  Speaker 2: I really like it for that. It's not good for
• 39:41
  Speaker 2: like I wouldn't copy paste the chatchip Ta output and
• 39:43
  Speaker 2: put it into a text. But it is good for
• 39:45
  Speaker 2: helping you understand how you feel. Thank you for being
• 39:49
  Speaker 2: so generous with your time. I really appreciate it. Thank
• 39:52
  Speaker 2: you for having me. It was a super fun conversation.
• 40:02
  Speaker 1: Dan Schipper is the co founder and CEO of every
• 40:07
  Speaker 1: Today's show was produced by Gabriel Hundreds and edited by
• 40:10
  Speaker 1: Lydia Jane Kott. Our engineer this week was Hansdale Sheet.
• 40:14
  Speaker 1: We're always looking for ideas for who to talk to
• 40:17
  Speaker 1: and what to cover on the show. You can email
• 40:20
  Speaker 1: us at problem at Pushkin dot fm. You can find
• 40:24
  Speaker 1: me on x at Jacob Goldstein. You can find me
• 40:26
  Speaker 1: on LinkedIn. I'm Jacob Goldstein. Thank you very much for
• 40:29
  Speaker 1: listening to the show, and we'll be back next week
• 40:32
  Speaker 1: with another episode.