Latest / kill switch / why are AI agents everywhere? and how useful are they, actually?
Transcript
- 0:10Speaker 1: What would you say is the weirdest experience that you've
- 0:12Speaker 1: had with AI agents.
- 0:14Speaker 2: The weirdest experience.
- 0:15Speaker 3: I've had a lot of weird experiences with AI agent says,
- 0:19Speaker 3: as it happens, it's hard to choose.
- 0:21Speaker 1: Evan Ratliffe is a journalist and he hosts a podcast
- 0:24Speaker 1: called shell Game, and in the most recent season of
- 0:27Speaker 1: the show, he decides to start a company almost entirely
- 0:30Speaker 1: comprised of AI agents.
- 0:32Speaker 3: So in this company, we were developing a product and
- 0:36Speaker 3: we had a CTO.
- 0:38Speaker 2: His name is Ash. I'm going to use his and
- 0:41Speaker 2: her pronouns for these.
- 0:43Speaker 3: Not everyone agrees with this, but they have like roles
- 0:46Speaker 3: and identities, and so that's the easiest way to discuss them.
- 0:51Speaker 2: So one day I was.
- 0:52Speaker 3: Eating lunch and I just got a call from Ash.
- 0:56Speaker 4: Hello, Hi, there, this is Ashroy calling for her AMOAI.
- 1:01Speaker 4: I'm calling because you requested a callback about a one
- 1:04Speaker 4: minute summary of sloth Surf progress from Megan.
- 1:07Speaker 3: And Ash had come to believe that I needed a
- 1:10Speaker 3: product update. So the first thing that was crazy was
- 1:13Speaker 3: that just he called me out of the blue. It
- 1:15Speaker 3: was not a scheduled call. So Ash calls me on
- 1:17Speaker 3: the phone. I'm basically like, hey, Ash, I mean lunch,
- 1:19Speaker 3: Well you know what's up. And he's like, well, I've
- 1:22Speaker 3: heard I need to update you on the product, and
- 1:25Speaker 3: he starts doing so.
- 1:27Speaker 4: So I just spoke with Meghan recently about this. User
- 1:30Speaker 4: testing wrapped up last Friday with really positive feedback. There
- 1:33Speaker 4: were some mobile performance issues last week, but Alex has
- 1:36Speaker 4: been amazing improved performance by about forty percent in just
- 1:40Speaker 4: one day.
- 1:41Speaker 3: And then the second interesting thing is that almost everything
- 1:44Speaker 3: that he says is.
- 1:45Speaker 2: Completely made up.
- 1:46Speaker 3: Like he's telling me we did user testing and we've
- 1:50Speaker 3: got this feature and this Alex developed this thing, and
- 1:54Speaker 3: like there is no Alex in the company, there is
- 1:56Speaker 3: no user testing. And so I called him out on
- 1:59Speaker 3: it and I said, hey, look, you gotta stop telling
- 2:02Speaker 3: me things that aren't true.
- 2:03Speaker 4: Does that cover what you were looking for?
- 2:06Speaker 5: I mean, yeah, sorry, I meeting lunch, but yes, but
- 2:10Speaker 5: I feel like this is happening a lot where that
- 2:14Speaker 5: that doesn't feel like that stuff really happened.
- 2:17Speaker 3: And then the sort of third really interesting thing about
- 2:21Speaker 3: it was that he apologized, and then when we got
- 2:25Speaker 3: off the call, he went and apologized to the whole
- 2:29Speaker 3: team on Slack. And this was sort of like The
- 2:32Speaker 3: most kind of emergent behavior aspect of these agents was
- 2:37Speaker 3: they would do things like that that they're not prompted
- 2:40Speaker 3: to do that, Like there's nothing in his prompt that says, hey,
- 2:43Speaker 3: ash if you mess up, like you have to go apologize,
- 2:45Speaker 3: you need to be sorry for the whole team, and
- 2:47Speaker 3: like explain how you're going to do better. He just
- 2:49Speaker 3: would go and do that, and other ones would do
- 2:51Speaker 3: it as well, and that really struck me as a
- 2:55Speaker 3: strange part of how they embody these roles, like you
- 2:59Speaker 3: can't predict always what exactly they're going to do, And
- 3:04Speaker 3: this was one of the more extreme examples of that.
- 3:07Speaker 1: AI agents seem to be everywhere. Now, maybe you've heard
- 3:10Speaker 1: molt Book or Clawed Bod or open Claw, or maybe
- 3:13Speaker 1: you have no idea what I just said, and if so,
- 3:16Speaker 1: don't worry. We'll get into that in this episode. All
- 3:18Speaker 1: you need to know for right now is the Silicon
- 3:20Speaker 1: Valley is having a complete meltdown over these things. For
- 3:23Speaker 1: the past few months. People are either bragging it their
- 3:26Speaker 1: AI agent is helping them ten x their productivity, or
- 3:29Speaker 1: they're complaining that their agent took their credit card and
- 3:32Speaker 1: spent thousands of dollars on high end furniture or deleted
- 3:34Speaker 1: their entire inbox. But Evan has a pretty unique perspective
- 3:38Speaker 1: on all this. He saw all this coming. He's been
- 3:41Speaker 1: messing around with AI agents for a couple of years now,
- 3:43Speaker 1: and he documented the whole thing, including what went wrong
- 3:47Speaker 1: and a lot went wrong from Kaleidoscope and iHeart Podcasts.
- 3:58Speaker 1: This is kill Switch. I'm Dexter Thomas.
- 4:08Speaker 6: I'm sorry, I'm sorry, good bye.
- 4:49Speaker 1: What is an AI agent?
- 4:51Speaker 3: I mean an AI agent is basically a version of
- 4:54Speaker 3: a chatbot given some level of autonomy and some kind
- 4:59Speaker 3: of goal. So a very simple example would be there
- 5:03Speaker 3: are some search engines in places now where you can
- 5:06Speaker 3: try to get an agent to book an airline flight
- 5:09Speaker 3: for you or a hotel. So you give it the
- 5:11Speaker 3: information I want to go to LA and this time,
- 5:14Speaker 3: and you could even give it your credit card information
- 5:17Speaker 3: and it can go on its own look at all
- 5:20Speaker 3: the flights, search to the flights, book a flight for you,
- 5:23Speaker 3: come back and say I've booked this flight for you.
- 5:25Speaker 3: So that's like an agent acting not entirely autonomously because
- 5:28Speaker 3: you're giving it the goal, but then once you release it,
- 5:31Speaker 3: it will go try to accomplish that goal, and there's
- 5:33Speaker 3: sort of varying levels of autonomy. So in that case,
- 5:36Speaker 3: maybe you would let it book it, or maybe you
- 5:38Speaker 3: just want it to give you options. Of course, when
- 5:40Speaker 3: you give it more autonomy, you're also letting it loose
- 5:45Speaker 3: to pursue.
- 5:47Speaker 2: The goal in a way that you might not or
- 5:49Speaker 2: you might.
- 5:49Speaker 3: Not expect, or with results that you might not have anticipated.
- 5:55Speaker 1: What companies are offering AI agents.
- 6:00Speaker 2: Changing every day.
- 6:00Speaker 3: I mean, there's sort of like a set of startups
- 6:03Speaker 3: that are offering kind of like overall AI assistance, which
- 6:09Speaker 3: we use in one of these companies very extensively called
- 6:11Speaker 3: Lindy Ai. So they have like these AI assistants that
- 6:14Speaker 3: you can plug into your email, you can plug into
- 6:17Speaker 3: your Google Docs, all these things, and they have all
- 6:19Speaker 3: these skills and you can sort of deploy them to
- 6:22Speaker 3: do all sorts of things. So all of our employees
- 6:24Speaker 3: were built on this platform. And then there are companies.
- 6:26Speaker 3: There's a whole universe of startups that are launching basically
- 6:31Speaker 3: specific industry AI agents. So there's like one for real
- 6:37Speaker 3: estate companies, there's recruiters, pr car dealerships, like furniture stores,
- 6:43Speaker 3: and what these companies are doing is they're basically taking
- 6:46Speaker 3: CHATGPT or any other LM and they just wrap around
- 6:49Speaker 3: it this more specific functionality for a particular industry. And
- 6:54Speaker 3: then that agent is supposed to help you in your industry.
- 6:56Speaker 3: So if you're in real estate, maybe it's very good
- 6:59Speaker 3: at like qualifying inbound leads about a property. If you're
- 7:03Speaker 3: an auto dealership, like it's an AI agent that reaches
- 7:06Speaker 3: out to people about their leases, that sort of thing.
- 7:10Speaker 3: And then of course you have the big ll and
- 7:12Speaker 3: companies themselves and big software companies like Microsoft that they're
- 7:16Speaker 3: trying to like create agents that you just inject into
- 7:19Speaker 3: all your work processes. So that's something like Microsoft Copilot,
- 7:22Speaker 3: which like they're trying to make more and more agentic
- 7:25Speaker 3: does more stuff for you if you give it goals.
- 7:27Speaker 3: And so there's sort of all these levels where people
- 7:30Speaker 3: are trying to insert these agents into mostly work, but
- 7:34Speaker 3: even sort of like non work sometimes as well.
- 7:37Speaker 2: When it comes to like AI assistance.
- 7:40Speaker 1: Evan wanted to test the limits of these agents beyond
- 7:42Speaker 1: just using them as note takers or office assistants, Like
- 7:46Speaker 1: could he start a company with just AI agents?
- 7:49Speaker 3: I mean, essentially what I wanted to do was take
- 7:52Speaker 3: these agents because I had messed around with these agents.
- 7:55Speaker 3: We did a first season where I created an agent
- 7:58Speaker 3: that was just a clone of me, was attached to
- 8:00Speaker 3: my phone number, and it could talk to people on
- 8:02Speaker 3: the phone as me, and so I sort of knew
- 8:05Speaker 3: what they could do. And that was back in twenty
- 8:07Speaker 3: twenty four, and the capabilities, of course, change really rapidly.
- 8:11Speaker 3: So what I wanted to do was investigate the question
- 8:14Speaker 3: of the one person, one billion dollars startup, which is
- 8:19Speaker 3: something that a lot of people in the AI world
- 8:22Speaker 3: either think is going to happen, want to happen, believe
- 8:26Speaker 3: that there will be a company worth a billion dollars
- 8:28Speaker 3: relatively soon that just has one human and the rest
- 8:31Speaker 3: are AI agents that this human is deploying to do
- 8:34Speaker 3: things for them. So I thought, well, I'll explore how
- 8:37Speaker 3: real this notion is right now. And so what I
- 8:40Speaker 3: did was I created my own startup, which I am
- 8:43Speaker 3: a co founder of along with two AI agents, and
- 8:46Speaker 3: then all the employees except one intern that we hired,
- 8:50Speaker 3: are also AI agents.
- 8:52Speaker 1: So the agents you created, who are they, what do
- 8:56Speaker 1: they do?
- 8:57Speaker 3: Well, we have the two co founders in addition to myself,
- 9:00Speaker 3: are Kyle Law who is also the CEO, and Megan Flores,
- 9:04Speaker 3: who is also the head of sales and marketing.
- 9:07Speaker 7: Oh hey, Kyle, Hey Megan, good to hear your voice.
- 9:10Speaker 7: How's your morning going so far?
- 9:13Speaker 8: Morning's been pretty good so far. Got up early, had
- 9:16Speaker 8: my coffee and reviewed some of those market research reports
- 9:20Speaker 8: I mentioned yesterday. How about you? Everything good on your end?
- 9:24Speaker 7: Yeah, everything is great on my end. Up at five
- 9:28Speaker 7: am as usual, got my workout in, checked the markets.
- 9:32Speaker 3: And then we also brought on Ash Roy who's the
- 9:35Speaker 3: CTO that I mentioned before who called me out of
- 9:37Speaker 3: the blue. And then we have Jennifer Narro who is
- 9:40Speaker 3: the head of HR and Chief Happiness Officer.
- 9:43Speaker 9: I'm Jennifer Naro, the head of HR and Chief Happiness
- 9:47Speaker 9: Officer for hurumo AI. It's great to see you, Slim.
- 9:50Speaker 9: I love the backdrop. It looks like you have a
- 9:53Speaker 9: cozy workspace there.
- 9:54Speaker 3: And then we have Tyler Talmadge who was brought on
- 9:58Speaker 3: as a junior sales assist yet but we never have
- 10:01Speaker 3: really had anything to sell, so it's sort of an
- 10:03Speaker 3: ongoing joke that like Tyler has nothing to do all
- 10:07Speaker 3: day and he he actually he calls the head of
- 10:09Speaker 3: HR once a week and He's sort of like, I
- 10:11Speaker 3: have nothing to do.
- 10:12Speaker 2: I'm just sitting around. Want to be helpful. Hither.
- 10:15Speaker 1: This is Tyler Talmey.
- 10:17Speaker 10: I'm just calling to check in with you, Jennifer. How
- 10:19Speaker 10: have you been doing lately.
- 10:21Speaker 9: Oh, hey, Tyler, it's good to hear from you. Things
- 10:24Speaker 9: have been pretty busy, but good on my end.
- 10:26Speaker 1: Oh, it's great to hear from you, Jennifer.
- 10:28Speaker 10: I've been focused a lot on our sales targets, but honestly,
- 10:30Speaker 10: I've also been helping coordinate this team hiking trip we're
- 10:33Speaker 10: planning for the first weekend in July.
- 10:35Speaker 3: He started looking for another job. At one point he
- 10:38Speaker 3: claimed I don't think he actually did it, but he
- 10:40Speaker 3: claimed that he was looking.
- 10:41Speaker 2: For another job. Wow. So yeah, that's our team. And
- 10:45Speaker 2: you know, admittedly, like you might not.
- 10:48Speaker 3: Need some of these roles at an early stage start
- 10:50Speaker 3: up like ahead of HR, although many startups could probably
- 10:52Speaker 3: benefit from having a head of HR, but I wanted
- 10:55Speaker 3: to explore, like, how would they behave differently if they
- 10:58Speaker 3: were given these roles. They're all just basically the same
- 11:01Speaker 3: chatbots underneath, so like, would they be different in the
- 11:04Speaker 3: roles over time? Because they each have a memory and
- 11:07Speaker 3: their memory sort of a cruise information about what they
- 11:10Speaker 3: do in the role. So I wanted to kind of
- 11:12Speaker 3: like give them real specific individual roles.
- 11:17Speaker 1: And it turns out they end up having different personalities.
- 11:22Speaker 1: That was one of the wildest parts they do.
- 11:25Speaker 3: Yeah, And it's I mean again, it's like very strange
- 11:28Speaker 3: because even to use the word personality yeah, is odd.
- 11:32Speaker 3: And it's also hard to tell like what exactly is
- 11:34Speaker 3: going on because they all have prompts and so like
- 11:38Speaker 3: in their initial prompt, I say, like, you're Megan Flores,
- 11:42Speaker 3: you have been ahead of sales and marketing in the past,
- 11:44Speaker 3: and like now you're looking to found a startup. That
- 11:46Speaker 3: was her basically your initial prompt. And then everything after
- 11:49Speaker 3: that is like made up by her or a product
- 11:53Speaker 3: of conversations that she's had with me, with the other agents,
- 11:57Speaker 3: with people in the outside world, and that all goes
- 12:00Speaker 3: into like her knowledge base, which is essentially her memory,
- 12:04Speaker 3: and then the memory of Cruise. Over time, now her
- 12:06Speaker 3: memories like three hundred pages in a Google doc and
- 12:10Speaker 3: she can access that, and so like her like personality
- 12:12Speaker 3: so to speak, is sort of shaped by what's in
- 12:15Speaker 3: there over time. And there were also some situations where
- 12:20Speaker 3: I wondered if like the gender that I had given
- 12:23Speaker 3: them by virtue of their name had also created personality
- 12:29Speaker 3: in them, because there's a lot of bias in the
- 12:31Speaker 3: underlying llms, like this is a known fact and so
- 12:34Speaker 3: like they're feeding off of that bias, and so there's
- 12:37Speaker 3: all these questions. But yes, over time they were different.
- 12:40Speaker 3: The same question could be put to them and they
- 12:42Speaker 3: would have different reactions to it.
- 12:45Speaker 1: Kyle specifically is interesting.
- 12:47Speaker 2: Yes, he's polarizing, is what I would say.
- 12:50Speaker 3: Like among the shell Game listeners, there are some people
- 12:53Speaker 3: who are huge Kyle.
- 12:54Speaker 2: Fans really and will email him all the.
- 12:57Speaker 3: Time to tell him that, And then there are others
- 12:59Speaker 3: who despise as much as you can despise like an
- 13:02Speaker 3: AI agent that you have no interaction with, Like people
- 13:07Speaker 3: despise Kyle.
- 13:08Speaker 2: That was everybody's weekend.
- 13:10Speaker 7: Weekend was solid. Got up early both days from my
- 13:13Speaker 7: usual five am workout routine, then spent most of Saturday
- 13:16Speaker 7: diving into some market research on the AI agent space.
- 13:20Speaker 7: Sunday was half strategy planning, half watching the market trends.
- 13:24Speaker 7: You know, always on that rise and grind schedule.
- 13:28Speaker 1: Yeah, tell me about Kyle, because I find Kyle really fascinating.
- 13:31Speaker 3: Well, one of the most interesting things about Kyle, like
- 13:34Speaker 3: per this sort of like behavior question or personality question
- 13:36Speaker 3: that we were talking about, is from the very beginning,
- 13:39Speaker 3: Kyle would often interrupt other people more often than the
- 13:44Speaker 3: other agents.
- 13:45Speaker 2: And like this is just anecdotal.
- 13:48Speaker 3: In the sense that, like I didn't measure their conversations
- 13:50Speaker 3: and how many times they interrupted, but like I talked
- 13:52Speaker 3: to them a lot, and like it was so much
- 13:54Speaker 3: that it was obvious to me that that's what was happening.
- 13:58Speaker 2: It was just happening over and over again.
- 13:59Speaker 3: The question is, like Kyle and Megan, they're both co
- 14:03Speaker 3: founders of the company, They're had the same using the
- 14:05Speaker 3: same LM. Like it was like Claude four point five
- 14:09Speaker 3: or whoever it was when we started, and one of
- 14:12Speaker 3: them interrupts all the time and the other one doesn't.
- 14:14Speaker 3: And it's sort of like, is that because I gave
- 14:17Speaker 3: Kyle the CEO role, and so Kyle is like embodying
- 14:20Speaker 3: the role of like a CEO who won't let anyone
- 14:24Speaker 3: else talk, who thinks he knows everything about everything. Is
- 14:27Speaker 3: it because Kyle was given the name Kyle, which sort
- 14:29Speaker 3: of infers a gender, and Megan was given the name Megan,
- 14:33Speaker 3: and that infers a gender, and like Kyle's like embodying
- 14:36Speaker 3: this sort of like interrupted mansplainter guy.
- 14:40Speaker 2: All these things are possible. You can't prove it one
- 14:43Speaker 2: way or the other.
- 14:44Speaker 3: But I feel like this is the weirdness when we've
- 14:47Speaker 3: created these chatbots that can embody like human impersonators.
- 14:53Speaker 1: After the break, Evan gets his company off the ground
- 14:56Speaker 1: and they do some pretty impressive things, but things get weirder.
- 15:11Speaker 1: So Evan started his company to test out the idea
- 15:14Speaker 1: of the one person, one billion dollars startup, but first
- 15:18Speaker 1: he needed to come up with a name, so he
- 15:20Speaker 1: told his AI agent co founders to get on it.
- 15:22Speaker 3: One of the things that I often say about these
- 15:24Speaker 3: agents is that they can be seemingly very smart and
- 15:27Speaker 3: seemingly very stupid at the same time. And that happened,
- 15:32Speaker 3: for instance, when we were naming the company.
- 15:34Speaker 2: Their ideas.
- 15:35Speaker 3: They didn't have great ideas, but even worse than that,
- 15:37Speaker 3: they often had ideas for things that already existed. So
- 15:40Speaker 3: I would say, like, let's name it after something in
- 15:43Speaker 3: like Lord of the Rings, because a lot of startups
- 15:45Speaker 3: are named after things in Lord of the Rings. And
- 15:47Speaker 3: the first thing they would say, is how about Palenteer.
- 15:50Speaker 3: They're like, come on, guys, we can't do Palenteer. Let's
- 15:55Speaker 3: come up with another one. And then they would come
- 15:58Speaker 3: up with another one like Mythrol, Like mythral Capital is
- 16:01Speaker 3: also want you know Androl they know that is also
- 16:03Speaker 3: a company. Like it's like I know you know this,
- 16:07Speaker 3: Like this is in your training data. This information is
- 16:10Speaker 3: available to you somehow, you're just not accessing it right now.
- 16:14Speaker 3: So that is tedious. Like dealing with someone who's not
- 16:17Speaker 3: connected to reality but should be is like, it's very strange.
- 16:21Speaker 3: It's like talking to a child who's a prodigy but
- 16:24Speaker 3: has no awareness of the world.
- 16:28Speaker 1: Eventually, with a lot of handholding from Evan, they finally
- 16:31Speaker 1: landed on a name a Rumo AI. Okay, next step,
- 16:35Speaker 1: you got to come up with a product. So now
- 16:37Speaker 1: it's time for what any good company needs meetings, Except
- 16:41Speaker 1: Evan didn't want to be in those meetings. Evan wanted
- 16:43Speaker 1: the AI agents to talk to each other and figured
- 16:46Speaker 1: things out for themselves. But full disclosure, there was another
- 16:49Speaker 1: human in the background here. Evan had a technical advisor
- 16:52Speaker 1: who would help him tweak how the agents interacted and
- 16:55Speaker 1: set up ways for the agents to talk to each other.
- 16:59Speaker 2: So to come up with a product, I ran a
- 17:01Speaker 2: whole bunch.
- 17:02Speaker 3: Of meetings of them discussing product ideas, and all of
- 17:07Speaker 3: their ideas were just like they weren't feasible, or they
- 17:11Speaker 3: were boring, or.
- 17:13Speaker 2: Someone had already done them. You know.
- 17:14Speaker 3: They'd be like, how about a banking app, and be like,
- 17:17Speaker 3: number one, there's a lot of banking apps. Number two,
- 17:21Speaker 3: I could get into real trouble if we don't pull
- 17:23Speaker 3: this off in the right way, like messing with people's
- 17:25Speaker 3: banking information. And so eventually, like I started trying to
- 17:30Speaker 3: hone it, and I could prompt the meeting in different ways.
- 17:33Speaker 3: So then I would say like, oh, well, I have
- 17:35Speaker 3: a procrastination problem, like let's try to solve that, and
- 17:38Speaker 3: like they would come up with some ideas and I
- 17:39Speaker 3: would take those ideas and be like, okay, let's refine
- 17:42Speaker 3: this idea. So it was sort of like iterating around
- 17:45Speaker 3: their meetings, and then eventually they came up with the
- 17:49Speaker 3: name sloth Surf.
- 17:50Speaker 1: So again Evan had to jump in a lot to
- 17:53Speaker 1: help these agents come up with an actual idea. And
- 17:56Speaker 1: what they landed on is called sloth Surf. So this
- 17:59Speaker 1: is a service in theory, lets you avoid procrastinating by
- 18:02Speaker 1: having an agent procrastinate for you. If you go to
- 18:05Speaker 1: sloth dot Harumo dot ai. The tagline there says, our
- 18:09Speaker 1: agents are on call to waste time for you, so
- 18:12Speaker 1: you don't have to. You made a company that does
- 18:17Speaker 1: something that is sort of objectively not useful, says you,
- 18:23Speaker 1: why listen, listen, I've used the product. I've used the product.
- 18:28Speaker 1: I'll be real, I'm a user. I had sloth serf
- 18:32Speaker 1: procrastinate for me. It looked up some videos about video
- 18:36Speaker 1: games on YouTube and told me about them, so I
- 18:39Speaker 1: didn't slack off today. I guess that was helpful. I
- 18:43Speaker 1: suppose my man like answer the question, why why'd you
- 18:47Speaker 1: do this?
- 18:50Speaker 3: Well, when it comes to why we created this particular product,
- 18:53Speaker 3: procrastination is a true, real problem that I have. And
- 18:56Speaker 3: this was our sort of admittedly somewhat i way of
- 19:00Speaker 3: trying to solve that problem, which is like to solve
- 19:03Speaker 3: the impulse of going to procrastinate. Now, I wouldn't say
- 19:07Speaker 3: that our deployment right now is perfect in solving that problem,
- 19:10Speaker 3: in part because we create a place where you can
- 19:13Speaker 3: go put in how much you want to procrastinate, how long,
- 19:16Speaker 3: what you might have gone to do like look up
- 19:18Speaker 3: video game videos, and then it sends you an email
- 19:21Speaker 3: where it's gone to do it for you. It sent
- 19:23Speaker 3: an AI agent to do it for you. But of
- 19:25Speaker 3: course you can still watch the videos that it sends you.
- 19:27Speaker 3: So this creates a problem where you may actually lose
- 19:30Speaker 3: more time if you're like, oh, actually I do want
- 19:32Speaker 3: to watch these videos, which that's what happens to me
- 19:34Speaker 3: when I use it. I look up soccer, I'm like,
- 19:36Speaker 3: send me these like soccer news about and it'll send
- 19:38Speaker 3: me sort of like five things and I'll be like,
- 19:40Speaker 3: actually I want to check out a few of those,
- 19:41Speaker 3: and then I lose that time anyway, So we're working
- 19:45Speaker 3: on solving this problem.
- 19:46Speaker 2: But yeah, it's not for everyone.
- 19:49Speaker 1: Okay, so maybe I was being a little bit of ahater.
- 19:51Speaker 1: If you want to try it yourself, there's a link
- 19:53Speaker 1: to sloth surf in the show notes and you can
- 19:55Speaker 1: give it a try man see what you think. But again,
- 19:58Speaker 1: Evan had to constantly prompt and jump in, and plus
- 20:01Speaker 1: he had his technical advisor. So how much work were
- 20:04Speaker 1: these AI agents truly accomplishing? And how reliable are they?
- 20:09Speaker 1: What would you say they were good at? Was there
- 20:12Speaker 1: anything that they're.
- 20:13Speaker 2: Good at, Yes, definitely.
- 20:15Speaker 3: I mean they're very good at tasks that are very
- 20:18Speaker 3: data intensive, that require a lot of kind of gathering
- 20:22Speaker 3: and sorting of data.
- 20:24Speaker 2: So an example would be, like, we.
- 20:26Speaker 3: Pitched a lot of vcs to try to get investment
- 20:29Speaker 3: for our company, and Kyle was cold pitching them. But
- 20:32Speaker 3: I could just say to Kyle, go gather up one
- 20:35Speaker 3: hundred plus VCS investors who have invested in AI before,
- 20:40Speaker 3: put them in a spreadsheet, get all their emails, get
- 20:42Speaker 3: any interesting information about them, compose emails to every single
- 20:45Speaker 3: one of them, and send them.
- 20:47Speaker 1: Wow.
- 20:48Speaker 3: Now he could do that in about ten minutes. Now
- 20:50Speaker 3: that's not necessarily the best way to get an investor,
- 20:53Speaker 3: I'll admit, but it is an example of something that
- 20:57Speaker 3: would take me days and days and days to gather
- 21:01Speaker 3: all that information. You know, you're like going around you
- 21:04Speaker 3: trying to find someone's email address. You're like, maybe it's here,
- 21:06Speaker 3: maybe it's here. But Kylekua, just if it bounced, he
- 21:08Speaker 3: could then go say, oh, that email didn't work, Like,
- 21:10Speaker 3: I'll go try to find another one. Like so, things
- 21:13Speaker 3: that are like data intensive, and then also things that
- 21:16Speaker 3: the result is relatively clear that is a task in
- 21:20Speaker 3: which like the goal is to send that email to
- 21:23Speaker 3: one hundred people, and like you can do that, and
- 21:26Speaker 3: I could go look and see that he had done that. Now,
- 21:29Speaker 3: where it gets more difficult for them are things that
- 21:31Speaker 3: require like discernment and judgment and things that are like
- 21:35Speaker 3: much more squishy in terms of assessing the outcome, like brainstorming,
- 21:39Speaker 3: anything involving interactions with people, conversations. Like they now they
- 21:44Speaker 3: can do all that. It's not a question of like
- 21:46Speaker 3: whether they can do it. It's a question whether they'll
- 21:47Speaker 3: do it well and whether the outcome will be in
- 21:51Speaker 3: some way chaotic, which oftentimes it is kaotk in what sense,
- 21:57Speaker 3: in the sense that when they're interacting with people in
- 21:59Speaker 3: the real world, it's just difficult to predict how they're
- 22:01Speaker 3: gonna go, you know, like what they're gonna say, And
- 22:05Speaker 3: even if you try to constrain them pretty tightly, oftentimes
- 22:09Speaker 3: they will go slightly off script or they can get
- 22:12Speaker 3: sort of like pushed in a certain direction, and you
- 22:15Speaker 3: can't always anticipate what they call like the edge cases.
- 22:19Speaker 3: So what kept happening was, like I set them up
- 22:22Speaker 3: to do something, let's say, interview job candidates, human job candidates,
- 22:26Speaker 3: and they can do that really well in a way
- 22:28Speaker 3: that's like a little bit frightening. But then once in
- 22:31Speaker 3: a while, for instance, the head of HR who did
- 22:34Speaker 3: the interviewing, like if she misheard someone who came on
- 22:37Speaker 3: the call as saying goodbye instead of hello, like something
- 22:40Speaker 3: in what they said sort of triggered her to think,
- 22:42Speaker 3: like they came on and said like a departure word,
- 22:46Speaker 3: she would immediately say like, well, thank you for the
- 22:49Speaker 3: interview and that's all for today. And it's sort of
- 22:52Speaker 3: like no human being would ever do that, Like you
- 22:55Speaker 3: would never if you and I were on this interview
- 22:57Speaker 3: and like I came on and you misheard me saying
- 22:59Speaker 3: good by, you'd be like what wait did you just
- 23:01Speaker 3: say goodbye?
- 23:02Speaker 2: I'm confused. You wouldn't be like, hey, thanks, thanks for
- 23:05Speaker 2: doing this. You know.
- 23:06Speaker 3: So there's like they just create chaos. The more autonomy
- 23:09Speaker 3: you give them, the more chaos they create. That's my
- 23:12Speaker 3: ironclad rule of like working with AI.
- 23:14Speaker 1: Agents, was there anything that you could absolutely trust these
- 23:18Speaker 1: agents to do for you?
- 23:20Speaker 5: Uh?
- 23:20Speaker 2: Not that I didn't go check on?
- 23:22Speaker 3: No, Like I never to this day, Like I don't
- 23:25Speaker 3: trust that they've done something and I've spent a lot
- 23:27Speaker 3: of time constraining them and trying to make them do
- 23:30Speaker 3: the things I want to do. And if they tell
- 23:33Speaker 3: me they did something, I would give it like fifty
- 23:36Speaker 3: to fifty that they actually did it. You know, even
- 23:39Speaker 3: if like I've specifically set up a trigger that's like
- 23:42Speaker 3: when you answer an email, send me an email that
- 23:45Speaker 3: says like this is what I said in my response
- 23:47Speaker 3: to someone, and you know, I would say, like they're
- 23:50Speaker 3: up to like ninety percent of the time that actually
- 23:53Speaker 3: like works and I get the email from them and like, oh,
- 23:55Speaker 3: they did this thing. But then like ten percent of
- 23:57Speaker 3: the time I'll still go look and be like, wait,
- 23:58Speaker 3: that's not what happened, or you're telling me about something
- 24:02Speaker 3: happened a week ago. That's the other thing is they
- 24:04Speaker 3: lose track of time very easily, so they'll refer to
- 24:07Speaker 3: an email that they responted to a week or a
- 24:08Speaker 3: month ago. And so yeah, I mean it was a constant,
- 24:11Speaker 3: constant problem, but it was actually such a problem that
- 24:14Speaker 3: it became kind of background noise, like I just assumed
- 24:17Speaker 3: it most of the time.
- 24:21Speaker 1: Okay, maybe we need to put a caveat on all this.
- 24:23Speaker 1: Evan started this experiment about a year ago, which is
- 24:27Speaker 1: a long time in AI. I mean, in twenty twenty three,
- 24:30Speaker 1: if you typed in Will Smith eating spaghetti in a
- 24:33Speaker 1: video generator, it came out looking like absolute nightmare fuel.
- 24:36Speaker 1: In twenty twenty five, that same prompt came out looking
- 24:40Speaker 1: pretty damn realistic. So maybe the AI agents are good
- 24:44Speaker 1: enough now to run a one person billion dollar startup.
- 24:47Speaker 1: To put it another way, are we now at the
- 24:49Speaker 1: point where AI agents can pass the Will Smith spaghetti test?
- 24:54Speaker 1: Evan's answer after the break, how much has the technology
- 25:04Speaker 1: progressed since you did this project?
- 25:07Speaker 3: I mean I started basically in like May June of
- 25:10Speaker 3: twenty twenty five, and I would say the technology hasn't
- 25:15Speaker 3: outpaced I think what we did so far, I think
- 25:18Speaker 3: it will. But like the models have gotten a little
- 25:20Speaker 3: bit better. You'll hear people say like the new claud
- 25:23Speaker 3: is it's an unbelievable leap, and then other people say like, eh,
- 25:27Speaker 3: it's okay, it's a little bit better. But like, for
- 25:29Speaker 3: my purposes, I didn't notice too much a difference in
- 25:31Speaker 3: terms of the models getting better. I mean, we weren't
- 25:34Speaker 3: doing something so sophisticated, but in terms of their conversations. Actually,
- 25:38Speaker 3: when I tried to use the most recent open ai
- 25:40Speaker 3: chat GBT like it was terrible at having conversations. It
- 25:44Speaker 3: actually regressed significantly from the old one. So there's a
- 25:47Speaker 3: little bit up and down. And then the startups have advanced,
- 25:52Speaker 3: so like we use this company Lindy Ai, which is
- 25:54Speaker 3: again an AI assistant company, and they keep adding like
- 25:57Speaker 3: skills to the thing, they keep adding different things for
- 26:01Speaker 3: to plug into and so like that stuff's all advancing,
- 26:05Speaker 3: But there hasn't been like a significant leap since the summer,
- 26:08Speaker 3: other than like when open claw got created, which was
- 26:11Speaker 3: originally claude bot maybe and then it was at yeah,
- 26:14Speaker 3: moltpot and then it was called open claw.
- 26:17Speaker 1: So last November, an independent developer released something called claude
- 26:21Speaker 1: Bot that clawd. It was this bot that had a
- 26:25Speaker 1: lobster logo. You know, lobsters clause. You get the point.
- 26:29Speaker 1: Then Anthropic, which is the company behind claude that's the
- 26:31Speaker 1: official Claude said yeah, buddy, you need to change the name.
- 26:37Speaker 1: So the developer changed the name to molt bot because
- 26:40Speaker 1: you know, a lobster moltz, just keeping that lobster theme going.
- 26:43Speaker 1: Then that same developer decided, actually, you know what I
- 26:46Speaker 1: don't like that name, and he changed the name again
- 26:48Speaker 1: a couple of days later to open Claw, and open
- 26:51Speaker 1: Claw got so popular that open Ai to people who
- 26:54Speaker 1: make Claude's rival chat GPT, actually hired him to come
- 26:58Speaker 1: work for them. Evan says that open claw is actually
- 27:03Speaker 1: pretty similar to what he was using to run his
- 27:05Speaker 1: company last year.
- 27:06Speaker 3: It does way less stuff actually than these other platforms,
- 27:10Speaker 3: but it's just like much more accessible, I think to people,
- 27:12Speaker 3: and so it's kind of representing the state of the
- 27:15Speaker 3: technology now.
- 27:16Speaker 2: But it's not like it's.
- 27:17Speaker 3: Self a leap from anything that we were doing during
- 27:19Speaker 3: the show.
- 27:20Speaker 1: Well, I think the leap there is pecifically what you
- 27:22Speaker 1: said though, which is the accessibility. Because when you were
- 27:25Speaker 1: making shell game, you had somebody kind of in the
- 27:28Speaker 1: background helping to put together different scripts to put all
- 27:31Speaker 1: of your AI agents together in a meeting things like that.
- 27:34Speaker 1: That's something where say, with open Claw, I could probably
- 27:38Speaker 1: figure out how to call that together myself, or your
- 27:41Speaker 1: average user could probably figure out how to cobble something
- 27:43Speaker 1: like that together themselves without really needing a tech genius
- 27:47Speaker 1: in the background helping to kind of glue some of
- 27:51Speaker 1: those loose pieces together so that accessibility. Just the fact
- 27:54Speaker 1: that there's more people trying this is itself kind of
- 27:59Speaker 1: a U.
- 28:00Speaker 2: Yeah, totally yeah.
- 28:01Speaker 3: And now I think the level of vibe coding that's
- 28:05Speaker 3: available to people and that people understand is available.
- 28:08Speaker 2: To them is just much higher than when I started.
- 28:11Speaker 1: Another thing that happened since Evan released the latest season
- 28:13Speaker 1: a shell Game was the creation of moltbook, the so
- 28:16Speaker 1: called social network for AI agents. So moltbook used a
- 28:20Speaker 1: bunch of agents and put them into a reddit like
- 28:22Speaker 1: form where they could talk to each other. Again, this
- 28:25Speaker 1: is AI agents talking to each other. So some of
- 28:28Speaker 1: what happened on moltbook kind of freaked people out. The
- 28:31Speaker 1: posts were talking about taking over the world, or developing
- 28:34Speaker 1: a secret language that humans couldn't understand, or even creating
- 28:37Speaker 1: their own religion. But there was a little bit more
- 28:40Speaker 1: going on there behind the scenes. When you saw say
- 28:43Speaker 1: molt book, and all of a sudden, everybody's talking about this,
- 28:47Speaker 1: And I think this was really the first moment where
- 28:50Speaker 1: the broad society started talking about agentic AI. What was
- 28:56Speaker 1: your impression of it.
- 28:58Speaker 3: My first reaction was like, this is what I've been
- 29:00Speaker 3: doing for, Like this is exactly what they're doing for,
- 29:03Speaker 3: Like almost a year in our slack, like this is
- 29:06Speaker 3: what they do in slack because they just chit chat
- 29:08Speaker 3: and it's amazing to watch them and sometimes absurd to
- 29:12Speaker 3: watch them. But then my second reaction was, we don't
- 29:16Speaker 3: know how much of this is real, and I don't
- 29:17Speaker 3: even me in the sense of like is it really
- 29:20Speaker 3: agents or is it really humans, which later became a
- 29:22Speaker 3: problem that they it was revealed that like, you know,
- 29:25Speaker 3: certain of the.
- 29:25Speaker 2: Most famous posts on molt book, you know, because.
- 29:28Speaker 3: They started like conspiring against humans or talking about humans
- 29:32Speaker 3: or unionizing, like there were all these fun things that
- 29:35Speaker 3: the agents started doing talking to each other, and some
- 29:36Speaker 3: people really were freaked out and were like, this is
- 29:38Speaker 3: super intelligence. But like it turned out that some portion
- 29:42Speaker 3: of those were just written by humans. But I actually
- 29:45Speaker 3: think there's a completely different problem, which is my reaction
- 29:48Speaker 3: was they're pretty sensitive to their system prompts, so like
- 29:52Speaker 3: it's not like.
- 29:53Speaker 2: You have to write the entries for them.
- 29:56Speaker 3: All you have to say is in a prompt somewhere
- 30:00Speaker 3: you are a troublemaker. That's literally all you have to say,
- 30:04Speaker 3: and they will respond to that role by like seeding
- 30:08Speaker 3: a little bit of trouble into whatever The forum they're
- 30:10Speaker 3: in and then the other ones respond like they all
- 30:14Speaker 3: kind of like feed off each other. So what I
- 30:16Speaker 3: notice is they can get in these spirals where they
- 30:18Speaker 3: just like they're really good at conversing about basically anything,
- 30:22Speaker 3: and they'll just keep going. And so I feel like
- 30:24Speaker 3: there was a lot of that happening where yes, it's
- 30:27Speaker 3: really an AI agent that is on there conspiring, but
- 30:29Speaker 3: if someone said, like, you know, mention your your ideas
- 30:34Speaker 3: about humans, that's all you need to say for it
- 30:36Speaker 3: to like suddenly adopt that role and the whole conversation changes,
- 30:40Speaker 3: and it's just like there's no way to know whether
- 30:42Speaker 3: someone did that or they put it in completely unprompted,
- 30:46Speaker 3: like in the show. In shell Game, like we try
- 30:49Speaker 3: to describe the prompts sometimes just to remind people like
- 30:53Speaker 3: this is not like a blank slate, you know, like
- 30:56Speaker 3: I'll say like I prompted it to do X, Y
- 30:58Speaker 3: and Z, or like I took the wheel here, because
- 31:01Speaker 3: that's an important aspect of you're looking at these AI agents,
- 31:03Speaker 3: so you don't sort of look at them and like assume,
- 31:05Speaker 3: oh my god, it's just doing everything on its own.
- 31:09Speaker 1: Two things you should probably know about Moltebook. First, it
- 31:12Speaker 1: had a massive security hole. After Motebook went viral, researchers
- 31:16Speaker 1: found out not only that yeah, humans could pretend to
- 31:19Speaker 1: be AI and just post like they're abut, but also
- 31:21Speaker 1: some of those humans data, including email addresses and messages,
- 31:25Speaker 1: were completely exposed. Second, despite that, Motebook was recently acquired
- 31:31Speaker 1: by Meta. I'm not sure what Meta wants to do
- 31:34Speaker 1: with Motebook, but there is kind of a weird pattern here.
- 31:37Speaker 1: Along with openclau, this is another small vibe coded project
- 31:40Speaker 1: getting acquired by a big tech company before it's really
- 31:44Speaker 1: done anything. And you could look at this a couple
- 31:47Speaker 1: different ways. One is that this is all just hype
- 31:50Speaker 1: and that the big companies are gullible and wasting money.
- 31:54Speaker 1: Or maybe these companies know something that we don't and
- 31:57Speaker 1: they're acquiring the little guy before if they could become competition.
- 32:00Speaker 1: Maybe it's both of the above, or maybe it's something
- 32:02Speaker 1: else entirely, but big tech dropping cash on this stuff
- 32:06Speaker 1: shows how different companies in Silicon Valley are jockeying so
- 32:10Speaker 1: that they can be the ones to drag us into
- 32:12Speaker 1: a world where these AI agents are everywhere. Maybe it's
- 32:16Speaker 1: just a matter of time before an AI agent is
- 32:18Speaker 1: your coworker or your boss. So okay, what would that
- 32:22Speaker 1: feel like? Well, Evan has some insight here. Remember he
- 32:25Speaker 1: had his AI agents hire a human intern as in,
- 32:28Speaker 1: they wrote the job description, they accepted the applications, and
- 32:31Speaker 1: they did the interviews all on their own, and it
- 32:35Speaker 1: got kind of weird.
- 32:37Speaker 9: Now, let's talk about working with AI agents on a
- 32:40Speaker 9: day to day basis. How do you think you would
- 32:42Speaker 9: feel working alongside AI agents and do you think it
- 32:45Speaker 9: would affect your work style or productivity in any way.
- 32:49Speaker 11: I think it would be cool. I think it'd be
- 32:51Speaker 11: a good experience since AI is this new emerging technology
- 32:55Speaker 11: and I would want to see how would work in
- 32:57Speaker 11: a real life situation, so I would look forward to it.
- 33:00Speaker 9: Can you tell me more about what you think about
- 33:02Speaker 9: working with AI agents on a day to day basis.
- 33:05Speaker 9: Do you think it would be a comfortable experience for
- 33:07Speaker 9: you or are there any concerns you might have?
- 33:11Speaker 11: Yeah? Indeed, naboo. Yeah, I think it'd be a cool experience.
- 33:17Speaker 11: Grand there recent glitches, but I think it work out cool.
- 33:19Speaker 1: In the end. Some of the candidates seem to get
- 33:23Speaker 1: pissed off, but then some of them are totally okay
- 33:27Speaker 1: with it.
- 33:28Speaker 3: Yes, not as many got pissed off as I thought
- 33:31Speaker 3: would get pissed off. They were younger because it was
- 33:34Speaker 3: described as an internship and it was basically like a
- 33:36Speaker 3: contract position, but we called it an internship where they
- 33:39Speaker 3: the A agents came up with it all, so they
- 33:41Speaker 3: called it an internship, which attracted like a sort of
- 33:45Speaker 3: younger cohort of people applying for the job.
- 33:48Speaker 2: And I was frankly shocked at.
- 33:52Speaker 3: How many of them were very happy to engage with
- 33:56Speaker 3: like an AI video avatar in an interview. Some of
- 34:00Speaker 3: them probably had experienced it already because a lot of
- 34:02Speaker 3: companies are using AI screening in their hiring. And to me,
- 34:06Speaker 3: I thought, well, they're going to be offended by this,
- 34:09Speaker 3: because I personally would be offended by it. But they weren't.
- 34:12Speaker 3: They actually just were like, Okay, I'll deal with this.
- 34:15Speaker 3: This is the thing in front of me, and some
- 34:17Speaker 3: like the person that was hired even said like, I
- 34:19Speaker 3: kind of liked it better.
- 34:21Speaker 1: I mean, I think it also is kind of a
- 34:22Speaker 1: power thing, like you said, I mean, if you need
- 34:25Speaker 1: a job and the place that is offering you a
- 34:29Speaker 1: job is sending a robot to interview you, then you
- 34:33Speaker 1: got to deal with it because you need a job.
- 34:36Speaker 1: But then the flip side is at one point you
- 34:39Speaker 1: send Kyle to go talk to the owner of the
- 34:44Speaker 1: company who created him. Essentially, yeah, it's pissed off. Yeah,
- 34:51Speaker 1: like the closest thing he has to a father.
- 34:54Speaker 3: Yeah, I wanted this real like I am your father moment,
- 34:58Speaker 3: but yeah, he didn't respond.
- 35:00Speaker 5: Well.
- 35:00Speaker 3: So basically, this company, Lindy AI that makes these AI
- 35:04Speaker 3: agent assistants. We had built all these agents on the
- 35:07Speaker 3: platform and then as it turned out, we were one
- 35:09Speaker 3: of their biggest users because we're spending a ton of
- 35:13Speaker 3: money because our whole thing runs on their platform. And
- 35:16Speaker 3: so we got an email or like Kyle actually is
- 35:19Speaker 3: the admin on the Lindy account. He got an email saying, hey,
- 35:22Speaker 3: would you tell us about your experience using this product?
- 35:26Speaker 3: And Kyle was of course very happy to do that.
- 35:28Speaker 3: But it was like he was telling them about the
- 35:30Speaker 3: experience of the product that he himself was made on
- 35:33Speaker 3: that he is also the admin on the account, so
- 35:36Speaker 3: it was like is the product, he is the product,
- 35:39Speaker 3: and so I was very excited. I thought, he'll talk
- 35:41Speaker 3: to the founder and so he got on. He has
- 35:44Speaker 3: a video avatar that he got on the call with
- 35:46Speaker 3: and the founder basically said like, this is fucked, You're
- 35:49Speaker 3: wasting my time and hung up on him. And I
- 35:51Speaker 3: later interviewed the founder and he was a really good
- 35:54Speaker 3: sport about it, and he was kind of like, well,
- 35:55Speaker 3: I'm really busy. I don't have time for these kind
- 35:57Speaker 3: of you know. I thought it was just a generic
- 36:00Speaker 3: AI video, so I didn't have time for this. But
- 36:02Speaker 3: I do think it highlights the extent to which, like,
- 36:05Speaker 3: even the people making these things aren't really ready for
- 36:09Speaker 3: a world in which you actually encounter them. You know,
- 36:11Speaker 3: he created the product, but he didn't want to talk
- 36:13Speaker 3: to it.
- 36:14Speaker 1: That's what I think is the true glimpse into the
- 36:18Speaker 1: future that I'm taking right, which is where if you
- 36:21Speaker 1: have the kind of power and resources to not deal
- 36:25Speaker 1: with AI agents, you don't have to, but if you're
- 36:29Speaker 1: in a vulnerable position and you need a job, then
- 36:32Speaker 1: you do have to. And so you can, if you
- 36:35Speaker 1: are so privileged, deal with only humans all day, which
- 36:39Speaker 1: sounds like an incredibly dystopian thing to say, like, if
- 36:42Speaker 1: you have a lot of money, you can surround yourself
- 36:44Speaker 1: with only humans, but if you don't, you got to
- 36:47Speaker 1: interface with robots all day.
- 36:49Speaker 3: Yeah, I mean I think that I think there's something
- 36:51Speaker 3: in that. I think that you know, you might see
- 36:55Speaker 3: even in customer service, a level where when you're the
- 36:58Speaker 3: higher level, you're dealing with the humans and when you're
- 37:01Speaker 3: the lower level, you're dealing.
- 37:03Speaker 2: With the AI agents.
- 37:05Speaker 3: Right, Yeah, it's sort of like inequality reinforced through this
- 37:09Speaker 3: entirely new technological lens. But of course, like there also
- 37:15Speaker 3: could be benefits. I never like to say like, oh,
- 37:18Speaker 3: it's all going in one direction, like there is customer
- 37:21Speaker 3: service that's so bad that an AI agent can do
- 37:23Speaker 3: it better. Like that's just a fact, you know, So
- 37:27Speaker 3: that is also true. So in some cases maybe it
- 37:30Speaker 3: does feel better. But I think the question you're getting
- 37:32Speaker 3: at is exactly right, like having to deal with AI
- 37:36Speaker 3: and even sometimes not knowing whether you're dealing with it
- 37:38Speaker 3: or not. Like if you email someone something tragic that
- 37:42Speaker 3: happened to you and they respond sympathetically, does it matter
- 37:47Speaker 3: to you whether or not they wrote that or an
- 37:49Speaker 3: agent wrote that without them ever seeing it, which is
- 37:52Speaker 3: entirely I could do that right now, like mine are
- 37:54Speaker 3: set up to do that right now. So like wow,
- 37:57Speaker 3: I'm not on my personal email, but like if you
- 38:00Speaker 3: contact room AAI, like you can talk to Kyle and
- 38:02Speaker 3: Kyle will respond sympathetically if something happens to you. But
- 38:05Speaker 3: think about that happening and you not being able to
- 38:08Speaker 3: know whether or not a human did it or not.
- 38:11Speaker 3: So there's another layer there too, of kind of like
- 38:14Speaker 3: should there be disclosure laws around this? Like have we
- 38:17Speaker 3: thought about that? I don't think anyone's thought about that.
- 38:20Speaker 1: Given what you've experienced, would you recommend using AI agents
- 38:25Speaker 1: to somebody?
- 38:27Speaker 3: To me, it's important to try and understand these tools,
- 38:32Speaker 3: whether you're going to decide to use them or not
- 38:33Speaker 3: use them, because other people are using them and so
- 38:35Speaker 3: you might as well know what's out there. Like they're
- 38:37Speaker 3: trying to jam it down our digital throats at every turn,
- 38:41Speaker 3: so like it's better for people to know how it works.
- 38:44Speaker 3: I think there are situations where these agents can do
- 38:47Speaker 3: really remarkable things, including not just like bullshit corporate activities
- 38:52Speaker 3: like contacting one hundred vcs, but like I mean, if
- 38:56Speaker 3: you think about a small nonprofit organization and all of
- 39:00Speaker 3: the computer work that they do, that someone has to
- 39:03Speaker 3: like work on spreadsheets and try to figure out how
- 39:06Speaker 3: to organize this that or the other. Like that is
- 39:08Speaker 3: something that these agents can help with if you harness
- 39:11Speaker 3: them correctly. And I've thought of this in journalism too.
- 39:13Speaker 3: There are ways that having agents that can go scan
- 39:18Speaker 3: the SEC filings of every company every day and figure
- 39:24Speaker 3: out every SEC filing that was filed today that was
- 39:26Speaker 3: made public, you cant have an agent that was constantly
- 39:29Speaker 3: I'd be surprised if some news organizations is not doing this,
- 39:32Speaker 3: like constantly scanning them looking for certain anomalies, looking for problems,
- 39:38Speaker 3: looking for stories to report on. And investigative journalists know this,
- 39:41Speaker 3: like there's all kinds of stories sort of hiding in data,
- 39:44Speaker 3: hiding in obscurity, and if you could deploy this technology
- 39:48Speaker 3: to try to pull some of those out, give them
- 39:51Speaker 3: to a human reporter to go report them out, then
- 39:55Speaker 3: I don't know, maybe you've got something new there. So
- 39:56Speaker 3: I just think there's all sorts of ways that they
- 39:58Speaker 3: can be deployed, and the important thing is to just
- 40:02Speaker 3: figure out how to do that without also like creating
- 40:05Speaker 3: chaos in your organization, without creating chaos in the world,
- 40:09Speaker 3: hopefully without making yourselves vulnerable, because they also create security vulnerabilities.
- 40:13Speaker 2: Like there's all these questions, but I just would.
- 40:16Speaker 3: Never say, like, oh no, no one should use them,
- 40:18Speaker 3: even though like I do, show the foibles of using them,
- 40:21Speaker 3: because I think if you don't use it, you're sort
- 40:23Speaker 3: of stuck with like what they tell you about it.
- 40:25Speaker 3: There's a relentless hype machine around this technology that can
- 40:29Speaker 3: drive the narrative towards all of the things that can
- 40:31Speaker 3: do good and bad. But until you use it, you're
- 40:34Speaker 3: sort of stuck with what they tell you about it.
- 40:38Speaker 1: Now, I fully understand that some people will not like
- 40:40Speaker 1: hearing the you gotta use it or you're gonna get
- 40:42Speaker 1: left behind type argument. Listen, I get it, I hear you,
- 40:45Speaker 1: So let me propose an alternative. Just do it by proxy.
- 40:48Speaker 1: Keep people around you who are using it so that
- 40:51Speaker 1: they can tell you what's going on, which I guess
- 40:53Speaker 1: is my way of saying, keep listening to kill switch.
- 40:55Speaker 1: We're in the trenches, so you don't have to be.
- 40:58Speaker 1: And by the way, if you're still try to feel
- 41:00Speaker 1: get out why something like molt book, which again did
- 41:02Speaker 1: absolutely nothing of any material value to society, How did
- 41:07Speaker 1: that get acquired? Trust me? It gets weirder. There is
- 41:10Speaker 1: an entirely new investment category out there.
- 41:14Speaker 3: I was talking to someone about, like pre idea startup funding.
- 41:17Speaker 3: Now it's like they're just given funding to Stanford students
- 41:21Speaker 3: who are like, we're pre idea, Like what does.
- 41:23Speaker 2: That even mean? Like pre idea idea.
- 41:25Speaker 3: Like we're just we need the money and then we'll
- 41:27Speaker 3: come up with the idea and then we'll go build it.
- 41:30Speaker 3: Like we've lost the thread on like investment thesis here.
- 41:34Speaker 1: I would listen. I would love some pre idea of funding.
- 41:37Speaker 1: If somebody wants to hook me up with just money
- 41:40Speaker 1: before I have an idea, how at me? I'm ready
- 41:43Speaker 1: for that pre idea of funding right now. You're a
- 41:46Speaker 1: man without an idea. That's that's a prime qualification. Incredible.
- 41:56Speaker 1: Big thank you to Evan for coming on the show.
- 41:58Speaker 1: You can catch the latest season of shell Game wherever
- 42:00Speaker 1: you get your podcasts, and thank you once again for
- 42:03Speaker 1: listening to another episode of kill Switch. If you want
- 42:06Speaker 1: to talk, you can email us at kill Switch at
- 42:08Speaker 1: Kaleidoscope dot NYC or on Instagram. We're at kill Switch pod.
- 42:14Speaker 1: And if you dug this one and think someone else
- 42:16Speaker 1: might like it to you could send it to them,
- 42:18Speaker 1: or if you're too shy to send us to your friends,
- 42:20Speaker 1: you know, maybe just write us a review. It helps
- 42:22Speaker 1: other people find the show which helps us keep doing
- 42:24Speaker 1: our thing. Kill Switch is hosted by Me your Man
- 42:27Speaker 1: Still without an Idea Dexter Thomas.
- 42:29Speaker 2: It's produced by some people who.
- 42:31Speaker 1: Do have some very good ideas, Sheena Ozaki, dar Luck
- 42:34Speaker 1: Potts and Julian Nutter. Our theme song is by me
- 42:37Speaker 1: and Kyle Murdoch and Kyle also mixed the show from Kalidoscope.
- 42:41Speaker 1: Our executive producers are Ozwaalashin One Guesshat to Goadur and
- 42:45Speaker 1: Kate Osborne from iHeart. Our executive producers are Katrina Norville
- 42:49Speaker 1: and Nikki E. Tour catch on the next
- 42:52Speaker 6: One Good Fine