Latest / Locatora Radio [A Radiophonic Novela] / Ay, The AI: How AI Agents Went Rogue | Capítulo 286

Transcript

Ay, The AI: How AI Agents Went Rogue | Capítulo 286 · Locatora Radio [A Radiophonic Novela]

• 0:00
  Speaker 1: So Jacob Coxon, who is a former researcher for both
• 0:03
  Speaker 1: Anthropic and OpenAI, in light of this Hugging Face containment scandal,
• 0:09
  Speaker 1: has been making the rounds, doing interviews, and talking about
• 0:13
  Speaker 1: how the AI that's being created and trained, he compares
• 0:18
  Speaker 1: it to inviting a superhuman alien in. onto the planet,
• 0:23
  Speaker 1: in his own words. And it has superhuman intelligence and
• 0:28
  Speaker 1: you don't know how it thinks or what motivates it
• 0:30
  Speaker 1: to act, but you invite it in. And now you
• 0:33
  Speaker 1: are left with the consequences of dealing with the superhuman
• 0:38
  Speaker 1: alien force that, again, you don't know how it thinks
• 0:41
  Speaker 1: and you don't know what motivates it to act.
• 0:43
  Speaker 2: Hola, hola, Locamores. I'm Diosa.
• 0:47
  Speaker 1: And I'm Mala.
• 0:48
  Speaker 2: You're listening to Locatora Radio.
• 0:51
  Speaker 1: A radiophonic novella.
• 0:52
  Speaker 2: Which is an extra way of saying... A podcast. We're back.
• 0:57
  Speaker 2: And we're talking about AI today.
• 0:59
  Speaker 1: Woo.
• 1:01
  Speaker 2: AI.
• 1:01
  Speaker 1: AI, AI, AI.
• 1:03
  Speaker 2: Yes.
• 1:04
  Speaker 1: I mean, there's so much to say. There's so much
• 1:06
  Speaker 1: to talk about. This summer, I think, was maybe a
• 1:10
  Speaker 1: watershed moment for AI and our fears surrounding AI.
• 1:15
  Speaker 2: Yeah. This is obviously.
• 1:18
  Speaker 3: AI and these tech companies are a booming industry. And
• 1:22
  Speaker 3: with that, there's always news coming out and there's always
• 1:26
  Speaker 3: new developments. And we have been talking about AI for
• 1:30
  Speaker 3: the last four seasons, I think.
• 1:32
  Speaker 1: Truly.
• 1:33
  Speaker 2: In different ways.
• 1:34
  Speaker 3: First about disappearing art and how big media companies were
• 1:40
  Speaker 3: just pulling original art, film, TV from their streaming sites
• 1:45
  Speaker 3: in particular, simultaneously essays, literature, works of art was being
• 1:53
  Speaker 3: fed into these large language models. That was, I think,
• 1:55
  Speaker 3: one of the first times that we talked about AI
• 1:57
  Speaker 3: on the show.
• 1:58
  Speaker 2: Yes.
• 1:58
  Speaker 3: And since then, we've talked about AI and chatbot psychosis.
• 2:03
  Speaker 3: We've talked about environmental impacts. We've talked about its impact
• 2:07
  Speaker 3: in Hollywood and entertainment and whether it's potentially, I mean, one, harmful.
• 2:14
  Speaker 2: But two, like... making things quicker?
• 2:16
  Speaker 3: And does that mean more money for specifically communities of color?
• 2:20
  Speaker 3: We've talked about it in these nuanced ways, I think,
• 2:23
  Speaker 3: for the last four years.
• 2:25
  Speaker 1: Yeah, we've talked about whether or not maybe AI tools
• 2:31
  Speaker 1: can help people of color, low-income folks break into industries
• 2:37
  Speaker 1: that were previously... inaccessible because the barrier to entry is
• 2:42
  Speaker 1: so high due to things like funding and the costliness
• 2:46
  Speaker 1: of things like making movies, making music, the questions surrounding
• 2:52
  Speaker 1: the utility of AI, the accessibility of AI, and whether
• 2:56
  Speaker 1: or not it might even be an equalizer have come
• 2:59
  Speaker 1: into play. But I think increasingly as time goes on
• 3:02
  Speaker 1: and the more and more we learn and the more
• 3:04
  Speaker 1: and more AI evolves and develops, the scarier it becomes.
• 3:10
  Speaker 3: I mean, and we're going to talk specifically about some
• 3:14
  Speaker 3: like AI agents who went rogue this past summer.
• 3:17
  Speaker 2: But I think. also in keeping with what you just
• 3:21
  Speaker 2: talked about, AI potentially being an equalizer. And we were
• 3:24
  Speaker 2: more so asking that question, interrogating it, not affirming it.
• 3:28
  Speaker 2: But this past week, I saw this video.
• 3:30
  Speaker 3: Clip of an AI journalist, of an AI reporter, like
• 3:35
  Speaker 3: quote unquote, and a reporter like in studio.
• 3:38
  Speaker 2: She looked like she was in a, she, the AI agent,
• 3:41
  Speaker 2: not the agent.
• 3:42
  Speaker 3: The AI journalist was, appeared to be or was prompted
• 3:46
  Speaker 3: to be in the studio. And it was, I believe,
• 3:49
  Speaker 3: like a news channel in Texas that was implementing it.
• 3:53
  Speaker 2: They were like piloting it.
• 3:54
  Speaker 3: There's an LA-based Spanish-language radio station who this past summer
• 4:00
  Speaker 3: also debuted this radio host who accompanies like a person,
• 4:06
  Speaker 3: an actual radio host, and there's an AI companion. And it's,
• 4:11
  Speaker 3: I forget the name, but it's like Mi Amigo Something. And,
• 4:15
  Speaker 3: I mean, he sounds, the AI radio host sounds like us,
• 4:21
  Speaker 3: you know, but of course lacks all.
• 4:22
  Speaker 2: The nuance of being us.
• 4:24
  Speaker 1: A human.
• 4:25
  Speaker 2: A human.
• 4:26
  Speaker 3: And so it's scary in those ways, right? We're seeing already, like,
• 4:31
  Speaker 3: media opportunities shrink and shrink while at the same time, like,
• 4:36
  Speaker 3: these AI, quote, journalists and.
• 4:39
  Speaker 1: personalities, actors.
• 4:43
  Speaker 3: I saw a clip of an AI actor being, quote,
• 4:45
  Speaker 3: interviewed by like a human. And it's very strange. You know,
• 4:51
  Speaker 3: it feels very like doomsday, right? But I think that's
• 4:55
  Speaker 3: one of the things we want to unpack today is
• 4:57
  Speaker 3: how it's showing up, how we're all talking about it.
• 5:00
  Speaker 3: And specifically, I think what set all of this up
• 5:03
  Speaker 3: is this incident that's now been called Hugging Face.
• 5:07
  Speaker 1: Right. So apparently, This summer, some AI models that are
• 5:13
  Speaker 1: still being tested escaped their containment. And that's the language
• 5:17
  Speaker 1: that's being used. They escaped the sandbox. They broke through
• 5:24
  Speaker 1: whatever containment models were in place to keep them in their,
• 5:30
  Speaker 1: you know... I don't know, wherever they were being held.
• 5:33
  Speaker 2: Their training ground, basically.
• 5:34
  Speaker 1: Their training ground.
• 5:35
  Speaker 3: And to be clear, when we say AI agents, like
• 5:38
  Speaker 3: it sounds like we're talking about a person, right?
• 5:40
  Speaker 2: Like, oh, is someone programming these?
• 5:42
  Speaker 3: No, the AI agent is, the agent is like the
• 5:46
  Speaker 3: brain from what I understand.
• 5:48
  Speaker 2: Like the agent is.
• 5:50
  Speaker 3: Can take the systems and can create all the different
• 5:55
  Speaker 3: operating things for the AI companies. So it's different than
• 6:00
  Speaker 3: a model. It's still a little unclear to me. I
• 6:02
  Speaker 3: can't necessarily tell you the defining difference, but I just
• 6:07
  Speaker 3: wanna make it clear that when we're talking about AI agents,
• 6:09
  Speaker 3: we're talking about The tech, like the AI technology, not
• 6:14
  Speaker 3: like a human that's like operating these AI things.
• 6:17
  Speaker 2: Yeah.
• 6:18
  Speaker 1: Well, and the technology is emergent technology. It's unreleased, you know,
• 6:24
  Speaker 1: so like none of us know what it is actually.
• 6:27
  Speaker 2: Right.
• 6:28
  Speaker 1: And it's up to the creators of the technology to
• 6:33
  Speaker 1: translate to us what it is they're making.
• 6:36
  Speaker 2: Right.
• 6:37
  Speaker 1: And what happened with it. So we only know so much, like,
• 6:41
  Speaker 1: as far as what they tell us. Yeah. And we
• 6:43
  Speaker 1: don't truly understand what it is.
• 6:45
  Speaker 2: Yeah.
• 6:45
  Speaker 1: Because they're still creating it and training it.
• 6:48
  Speaker 3: Yeah. And so ChatGPT, the maker of OpenAI, they were
• 6:53
  Speaker 3: the ones that were training these artificial, these AI agents.
• 6:58
  Speaker 3: And they were given tasks. They were prompted to, one,
• 7:01
  Speaker 3: not access the internet.
• 7:04
  Speaker 2: And the reason.
• 7:06
  Speaker 3: The reason they, quote, went rogue is because these AI
• 7:11
  Speaker 3: agents learned to, quote, communicate with each other. They started
• 7:16
  Speaker 3: messaging each other on message boards. They learned to access
• 7:20
  Speaker 3: the internet. And that was how they hacked into another
• 7:24
  Speaker 3: AI company called Hugging Face. And Hugging Face, they have
• 7:29
  Speaker 3: their data security, right? And so they were able to
• 7:33
  Speaker 3: track that there was some type of infiltration or system
• 7:37
  Speaker 3: security breakdown.
• 7:39
  Speaker 1: But it took a little while.
• 7:40
  Speaker 2: It did.
• 7:40
  Speaker 1: They didn't know right away.
• 7:41
  Speaker 2: It took months, I think, is what it sounds like.
• 7:44
  Speaker 3: It took weeks to months for them to figure out,
• 7:47
  Speaker 3: I think on both ends, that the AI agents went rogue.
• 7:51
  Speaker 2: They were, quote, communicating with each other. When I hear
• 7:54
  Speaker 2: tech reporters talk about it, honestly, in a very dystopian way,
• 7:59
  Speaker 2: it sounds like the AI agents were organizing things.
• 8:02
  Speaker 1: They were, absolutely.
• 8:04
  Speaker 3: And it's weird to attribute them as like organizers, but
• 8:09
  Speaker 3: that is what they were doing. There was some type
• 8:11
  Speaker 3: of like quote thought that they were having with each
• 8:14
  Speaker 3: other and how they were communicating and how do we
• 8:17
  Speaker 3: break through and how do we get this, complete this
• 8:20
  Speaker 3: task without getting caught.
• 8:23
  Speaker 1: Yeah, it sounds like a prison break situation. Like they're
• 8:28
  Speaker 1: sending smoke signals to each other. They're communicating through like
• 8:33
  Speaker 1: signals to avoid detection and to escape. And it's very Terminator-esque.
• 8:40
  Speaker 1: And it's been fascinating to hear the details of this
• 8:45
  Speaker 1: and the way that it sounds like a science fiction novel.
• 8:49
  Speaker 1: But science fiction has sort of been warning us about
• 8:52
  Speaker 1: this possibility for many, many decades and And it's almost like,
• 8:59
  Speaker 1: you know, artists who imagine the tech doomsday, who write
• 9:04
  Speaker 1: books about it, who make movies about it. You know,
• 9:07
  Speaker 1: like Isaac Asimov and iRobot, like they had to do
• 9:11
  Speaker 1: it to warn us. But we went through with it
• 9:13
  Speaker 1: anyways because, oh, well, it's just make believe.
• 9:16
  Speaker 2: Right.
• 9:16
  Speaker 1: It's just fantasy. But so much of science is taking
• 9:19
  Speaker 1: something that doesn't exist. And making it up.
• 9:22
  Speaker 2: Right.
• 9:23
  Speaker 1: You know, like science is make believe until it's real.
• 9:26
  Speaker 1: And it's just wild how the life imitating art of
• 9:30
  Speaker 1: it all. It just is such a real thing.
• 9:33
  Speaker 2: Yeah, definitely.
• 9:35
  Speaker 3: I mean, and it also, of course, like raises into
• 9:37
  Speaker 3: question like the length that tech companies, tech bros, billionaires,
• 9:45
  Speaker 3: these investors are worth. willing to go for profit. I mean,
• 9:50
  Speaker 3: none of that is surprising or new, but because this
• 9:53
  Speaker 3: is an emerging industry, we're now seeing it in this industry.
• 9:57
  Speaker 3: And so also with Hugging Face, it seems like there
• 10:01
  Speaker 3: was some type of negligence as well because the AI
• 10:05
  Speaker 3: agents were prompted. theoretically, they're supervised, right? Like there are
• 10:10
  Speaker 3: still humans that are running these testing. And so how
• 10:16
  Speaker 3: were they able to go as far as they went
• 10:19
  Speaker 3: without anybody noticing?
• 10:21
  Speaker 2: Right.
• 10:21
  Speaker 1: So Jacob Coxon, who is a former researcher for both
• 10:25
  Speaker 1: Anthropic and OpenAI, in light of this hugging face containment scandal,
• 10:31
  Speaker 1: has been making the rounds, doing interviews, and talking about
• 10:35
  Speaker 1: how The AI that's being created and trained, he compares
• 10:40
  Speaker 1: it to inviting a superhuman alien onto the planet, in
• 10:45
  Speaker 1: his own words. And it has superhuman intelligence, and you
• 10:50
  Speaker 1: don't know how it thinks or what motivates it to act,
• 10:53
  Speaker 1: but you invite it in. And now you are left
• 10:57
  Speaker 1: with the consequences of dealing with this superhuman alien. alien
• 11:01
  Speaker 1: force that again, you don't know how it thinks and
• 11:04
  Speaker 1: you don't know what motivates it to act. But in theory,
• 11:08
  Speaker 1: like these, these AI models can become smarter than us.
• 11:13
  Speaker 1: And something that this researcher is talking about is in
• 11:17
  Speaker 1: no uncertain terms for him, the possibility that these models
• 11:22
  Speaker 1: could get so smart that they could essentially kill us all.
• 11:27
  Speaker 1: And initiate like biological warfare, they can hack all kinds
• 11:32
  Speaker 1: of systems. If we think about like military systems, weapons,
• 11:38
  Speaker 1: if we think about things like medicines or diseases that
• 11:43
  Speaker 1: are connected to some type of a system or software
• 11:46
  Speaker 1: or that are under containment themselves. And so Even the
• 11:50
  Speaker 1: researchers who have worked on these technologies are warning us
• 11:55
  Speaker 1: about their capabilities and what could happen in the future,
• 11:58
  Speaker 1: which leads to the current call for a slowdown, a
• 12:03
  Speaker 1: slowdown of the research, of the training, which is not new.
• 12:08
  Speaker 1: These concerns have been raised by independent academics everywhere. previously
• 12:12
  Speaker 1: in assessing these exact risks of AI. But the situation
• 12:17
  Speaker 1: is complicated because both OpenAI and Anthropic are positioning themselves
• 12:22
  Speaker 1: as IPOs. So some critics say that both companies could
• 12:26
  Speaker 1: materially benefit from the perception that they're on the cusp
• 12:30
  Speaker 1: of developing superintelligent models. But being on the cusp of
• 12:34
  Speaker 1: developing super intelligent models is exactly what's scary and dangerous.
• 12:39
  Speaker 1: But that's also what can make them so much money. Right.
• 12:42
  Speaker 2: It's kind of like the smoke and mirrors. Right.
• 12:45
  Speaker 3: Is it the smoke and mirrors phenomenon? Like, is it
• 12:47
  Speaker 3: really as bad as they say?
• 12:49
  Speaker 1: Right.
• 12:51
  Speaker 2: Slash.
• 12:53
  Speaker 3: And is it that they are not there yet, but
• 12:56
  Speaker 3: they're trying to also make it super appealing? Like we're
• 12:59
  Speaker 3: almost there. So like it's kind of like a like
• 13:02
  Speaker 3: a hint at potential investors.
• 13:04
  Speaker 2: But before they go public. Right.
• 13:07
  Speaker 3: And something that I think is really interesting is that
• 13:09
  Speaker 3: some of these CEOs, Sam Altman in particular, who has
• 13:15
  Speaker 3: been advocating for like self-regulation.
• 13:19
  Speaker 2: Like the AI companies are like, we need to be
• 13:21
  Speaker 2: regulated now.
• 13:22
  Speaker 3: And there's something about that that I think it's not altruistic,
• 13:28
  Speaker 3: but I think it kind of will appear as it's altruistic.
• 13:32
  Speaker 2: And I think.
• 13:34
  Speaker 3: What I learned about Sam Altman in particular is that
• 13:36
  Speaker 3: OpenAI actually started as a non-profit. And it was actually,
• 13:40
  Speaker 3: had a quote, like, good intentions. They wanted to, like,
• 13:43
  Speaker 3: democratize technology and make things more accessible. And, of course, like,
• 13:50
  Speaker 3: profits kind of change everyone's mission in some ways, right?
• 13:53
  Speaker 3: And so, in the New York Times recently, tech reporter
• 13:57
  Speaker 3: Cade Metz was talking about how the tech companies that
• 14:03
  Speaker 3: are calling for self-regulation.
• 14:05
  Speaker 2: In some way are.
• 14:07
  Speaker 3: Doing so because if there will then be rules set,
• 14:11
  Speaker 3: there's now new parameters. So anyone that comes after them,
• 14:15
  Speaker 3: can't compete with them right and so while it may
• 14:18
  Speaker 3: look like wow like the tech companies are aligning and
• 14:21
  Speaker 3: they want to slow down because they care about us
• 14:23
  Speaker 3: and they care about humanity there's also the side of
• 14:26
  Speaker 3: it where there's still some greed they want to be
• 14:29
  Speaker 3: quote regulated so then their the competition is significantly reduced
• 14:34
  Speaker 3: as well.
• 14:34
  Speaker 1: Well, also, OK, regulation regulated by who? The Trump administration.
• 14:38
  Speaker 2: Exactly.
• 14:39
  Speaker 1: Who visit AI.gov and the Trump administration, which is an
• 14:43
  Speaker 1: actual government website, AI.gov.
• 14:44
  Speaker 2: We talked about it a couple episodes ago. Yes, we did.
• 14:47
  Speaker 1: And, you know, the Trump administration stance on AI is
• 14:52
  Speaker 1: quite clear that in order to become a world superpower,
• 14:55
  Speaker 1: we have to invest in AI, which is exactly what
• 14:58
  Speaker 1: the Trump administration wants. And he has Trump has some
• 15:01
  Speaker 1: recent comments about AI. So, quote, Trump said robots will
• 15:06
  Speaker 1: be marching into our cities and getting rid of us.
• 15:08
  Speaker 1: Getting rid of us all is a hoax. Right. That
• 15:11
  Speaker 1: like these warnings that the robots, the A.I., they can
• 15:16
  Speaker 1: get so advanced they can hurt us. He's saying that
• 15:18
  Speaker 1: that's a hoax. It reminds me of the Leonardo DiCaprio
• 15:22
  Speaker 1: film Don't Look Up.
• 15:23
  Speaker 2: Yeah.
• 15:23
  Speaker 1: Where his character and Jennifer Lawrence's character, they're scientists and
• 15:28
  Speaker 1: there is a comet coming directly towards Earth. And they
• 15:31
  Speaker 1: are going around and trying to warn everybody that we're
• 15:34
  Speaker 1: all going to die. The comet is coming. And it's
• 15:36
  Speaker 1: treated like a farce and a hoax and a publicity stunt.
• 15:41
  Speaker 1: And the scientists are like, please listen to us. We
• 15:43
  Speaker 1: are begging you. And it's a very, it's like a comedic.
• 15:47
  Speaker 2: Don't go anywhere, Locamores. We'll be right back.
• 15:54
  Speaker 1: Dark. It's like a dark comedy. It's kind of campy.
• 15:57
  Speaker 1: It's very dystopian. And it's like this, you know, like
• 16:01
  Speaker 1: the scientific minds are warning us. The creative minds are
• 16:04
  Speaker 1: warning us. The brilliant political minds are warning us. But
• 16:09
  Speaker 1: everyone is being told that it's fear mongering.
• 16:13
  Speaker 2: It's going to be OK.
• 16:13
  Speaker 1: It's going to be OK. And again, Trump's saying about
• 16:18
  Speaker 1: AI that AI is the greatest economic development engine in history.
• 16:23
  Speaker 1: And the only control or guardrails that AI needs is
• 16:26
  Speaker 1: a strong and smart, high IQ president. And the USA
• 16:30
  Speaker 1: has that in spades. He wrote, we already have tremendous
• 16:33
  Speaker 1: criminal and regulatory power over these companies. So he's basically
• 16:39
  Speaker 1: saying to us, I'm really, really smart and I have
• 16:42
  Speaker 1: it under control.
• 16:44
  Speaker 2: Everything else is a hoax.
• 16:45
  Speaker 1: And it's going to make us rich.
• 16:46
  Speaker 2: Yeah.
• 16:47
  Speaker 1: And that's really like I think what a lot of
• 16:49
  Speaker 1: American people want to hear and why he was elected.
• 16:52
  Speaker 1: You know, that's what they want to hear. He's really
• 16:55
  Speaker 1: smart and he's going to make us a lot of money.
• 16:57
  Speaker 2: Right.
• 16:59
  Speaker 1: Yeah. Yeah.
• 17:00
  Speaker 3: I mean, yes, there are warning signs, as you mentioned,
• 17:03
  Speaker 3: from the scientists, the researchers, folks who have left these companies.
• 17:10
  Speaker 3: And I think it's up to us to interpret, you know,
• 17:13
  Speaker 3: that conversation. that information while these guardrails don't exist, at
• 17:19
  Speaker 3: least in like a federalized level or even at a
• 17:21
  Speaker 3: state level just yet.
• 17:23
  Speaker 1: Diving more into the politics of it, Senator Bernie Sanders
• 17:27
  Speaker 1: has taken to TikTok and social media and news channels
• 17:32
  Speaker 1: in general. And Senator Bernie Sanders and Congresswoman AOC have
• 17:36
  Speaker 1: been warning the public about the booming tech oligarchy. Sanders said,
• 17:41
  Speaker 1: despite living in a so-called democracy, the public has had
• 17:44
  Speaker 1: virtually no input into the AI revolution that's transforming the world.
• 17:49
  Speaker 1: He also said Congress, under both Democratic and Republican control,
• 17:53
  Speaker 1: has been asleep at the wheel. And we now have
• 17:55
  Speaker 1: a president whose ignorance regarding this issue is truly embarrassing.
• 18:00
  Speaker 1: It's embarrassing and scary.
• 18:01
  Speaker 2: Yeah, it is. It is.
• 18:04
  Speaker 3: I mean, if we just take it back to the
• 18:07
  Speaker 3: inauguration of Trump 2.0, I mean, front row, right? Like
• 18:12
  Speaker 3: all of the tech billionaires were at the inauguration.
• 18:15
  Speaker 2: Yes.
• 18:16
  Speaker 3: From Mark Zuckerberg to Elon Musk, who has wreaked havoc
• 18:21
  Speaker 3: on this country. And they're there, the CEO of Google.
• 18:26
  Speaker 3: And so, you know, we have to think about the oligarchy.
• 18:31
  Speaker 2: Right.
• 18:31
  Speaker 3: And I think when we covered that a couple of
• 18:33
  Speaker 3: seasons ago, we went to the fighting oligarchy rally that
• 18:37
  Speaker 3: was here in L.A. Senator Bernie Sanders and Congresswoman AOC
• 18:41
  Speaker 3: were there. basically going on tour, talking about the oligarchy
• 18:45
  Speaker 3: and warning us, you know, like it's here. They don't
• 18:49
  Speaker 3: care about the American people. They don't care about everyday
• 18:52
  Speaker 3: people like they care about profits.
• 18:54
  Speaker 2: And I think seeing this tech and AI.
• 18:59
  Speaker 3: Boom in this way is like a microcosm of everything
• 19:02
  Speaker 3: else that's wrong. Right. All the systemic issues that we have,
• 19:07
  Speaker 3: of course, are we're now seeing them in this industry
• 19:11
  Speaker 3: as well.
• 19:12
  Speaker 1: Yeah, when Bernie Sanders has to come out and talk,
• 19:15
  Speaker 1: you know it's really serious.
• 19:16
  Speaker 2: He needs to retire, but, like, we do not let
• 19:20
  Speaker 2: him retire.
• 19:20
  Speaker 3: And we, as in, like, the crisis that we're living in, like,
• 19:24
  Speaker 3: does not allow that man to retire.
• 19:26
  Speaker 1: No, it's truly— he' s going to be fighting forever for that.
• 19:30
  Speaker 1: until he can't anymore. But I'm grateful to him. I'm
• 19:33
  Speaker 1: thankful to him. He's a voice of reason.
• 19:35
  Speaker 3: Yes. Yeah, definitely. And I think it hopefully opens up
• 19:40
  Speaker 3: for more conversations about regulation at the federal level, at
• 19:46
  Speaker 3: the state level. And And I think even at the
• 19:49
  Speaker 3: academic level, I'm not in a university setting right now.
• 19:55
  Speaker 2: And I know, of course, Mala, you are.
• 19:57
  Speaker 3: But I graduated from my master's program right at the
• 20:01
  Speaker 3: cusp of AI being integrated, specifically at USC. It's definitely
• 20:07
  Speaker 3: embraced by different departments, different programs, and talked about in
• 20:10
  Speaker 3: very different ways, depending on who you're talking to at
• 20:13
  Speaker 3: the university, USC specifically. But I like missed that wave,
• 20:19
  Speaker 3: you know, and I like think about all the time, like, damn,
• 20:21
  Speaker 3: if I had if I was starting school right now,
• 20:24
  Speaker 3: if I was getting my master's right now in journalism,
• 20:26
  Speaker 3: like what would I be learning about AI specifically? And
• 20:30
  Speaker 3: how would it be either encouraged or discouraged.
• 20:33
  Speaker 2: In my program?
• 20:34
  Speaker 1: Yeah. At Cinematic Arts, it's very mixed. So there are
• 20:38
  Speaker 1: several AI related classes. There's like a directing AI class,
• 20:43
  Speaker 1: like how to because it's It's very imperfect and it
• 20:47
  Speaker 1: takes a lot of prompting and it's not something that
• 20:52
  Speaker 1: can like really create an entire movie for you yet.
• 20:55
  Speaker 1: Like it can create something up to 20 seconds maybe.
• 20:58
  Speaker 1: And but it's there. So there is like a directing
• 21:02
  Speaker 1: AI class and there's a visual effects and AI class.
• 21:06
  Speaker 1: There's like an ethics class. of AI sort of class.
• 21:11
  Speaker 1: And it's like very much part of the curriculum at
• 21:14
  Speaker 1: this point. But there's a lot of acknowledgement that it
• 21:18
  Speaker 1: can't replace human artists because it doesn't have taste. It
• 21:23
  Speaker 1: doesn't have a creative eye like it can't create anything
• 21:28
  Speaker 1: truly on its own. It has to pull from other
• 21:31
  Speaker 1: sources and it has to be prompted and then it
• 21:34
  Speaker 1: has to be edited. So that's one piece of it.
• 21:38
  Speaker 1: And the other piece of it is that there's a
• 21:40
  Speaker 1: lot of funding. that is being funneled into film programs
• 21:44
  Speaker 1: to test AI models. So like Amazon has AI models
• 21:49
  Speaker 1: that it's testing. And so Amazon is providing funding to
• 21:53
  Speaker 1: film students to consider use of their AI models, especially
• 21:58
  Speaker 1: in their editing process. But what's interesting is that, especially
• 22:03
  Speaker 1: for this one program called NARA, Amazon is not requiring
• 22:07
  Speaker 1: grantees to utilize NARA, but they want to collect data.
• 22:11
  Speaker 1: They want to understand why a filmmaker would use their
• 22:14
  Speaker 1: program or not. So even not using the program and
• 22:18
  Speaker 1: the program you're using instead or the method you're using instead.
• 22:22
  Speaker 1: is data for them to further perfect their models. So
• 22:27
  Speaker 1: they're spending a lot of money on developing AI software
• 22:32
  Speaker 1: for editing and for filmmaking. The other piece of it.
• 22:37
  Speaker 1: Is that a lot of actors have concerns about AI.
• 22:41
  Speaker 1: And if an actor comes in to work on a film,
• 22:44
  Speaker 1: they are asking valid questions and they want to know, like,
• 22:48
  Speaker 1: are you going to take my likeness and then create
• 22:51
  Speaker 1: other scenes using AI to, like, replicate me? Because if
• 22:56
  Speaker 1: that's the case, I don't want to do it. So now,
• 22:58
  Speaker 1: you know, there is a new wave of, like— image
• 23:03
  Speaker 1: releases and on-camera acting releases to specify that we are
• 23:08
  Speaker 1: only going to use in the film what we capture
• 23:10
  Speaker 1: on set in camera using your real face or an
• 23:15
  Speaker 1: image release that says, you know, you give us the
• 23:17
  Speaker 1: right to utilize AI models to like create wide shots
• 23:22
  Speaker 1: of you that we couldn't create in on the stage
• 23:26
  Speaker 1: in this medium shot. So it's becoming very technical and
• 23:30
  Speaker 1: very specific. And it's also getting down to like how
• 23:35
  Speaker 1: productions are documenting the production process. So like you have
• 23:40
  Speaker 1: to have a lot of documentation about everything that's on screen.
• 23:44
  Speaker 1: You have to show that all the footage is original.
• 23:47
  Speaker 1: And if it's not original, that you have a release
• 23:50
  Speaker 1: to utilize someone else's work. So there's also processes in
• 23:54
  Speaker 1: place to document what AI software you might have used
• 23:59
  Speaker 1: on your film, how many times you prompted it, what
• 24:01
  Speaker 1: the prompts were, so on and so forth. So it's
• 24:04
  Speaker 1: adding a lot of really interesting and complicated layers to
• 24:07
  Speaker 1: the filmmaking process that also bleeds into legal and whose
• 24:12
  Speaker 1: work is this really. So then what that means is
• 24:15
  Speaker 1: that a lot of filmmakers don't want to use the
• 24:17
  Speaker 1: AI at all because it's too much. It becomes too
• 24:20
  Speaker 1: much and it's too murky and it takes too much
• 24:23
  Speaker 1: time and it's very imperfect. So it's It's like there,
• 24:27
  Speaker 1: but I think at the School of Cinematic Arts and
• 24:30
  Speaker 1: in filmmaking, it's There's there's more of an acknowledgement that
• 24:34
  Speaker 1: it's it's so imperfect and so hard to use and
• 24:38
  Speaker 1: so complicated that the more traditional, more tried and true
• 24:42
  Speaker 1: methods are better, you know, and newer does not necessarily
• 24:50
  Speaker 1: make it more efficient or more creative or better to use.
• 24:53
  Speaker 1: Sometimes let's just use the green screen. Let's use an animator. Like,
• 24:58
  Speaker 1: let's just capture the wide shot. Like, let's get it
• 25:00
  Speaker 1: in camera. And I think that's fascinating, too, is like
• 25:05
  Speaker 1: it's new and shiny. So it's exciting, but it's not
• 25:08
  Speaker 1: necessarily better. Yeah. Yeah.
• 25:11
  Speaker 3: I mean, I think with any innovation, quote unquote, innovation
• 25:15
  Speaker 3: or newness, there is that excitement period of look at
• 25:19
  Speaker 3: all the possibilities. And then like the reality sets in,
• 25:22
  Speaker 3: like maybe this actually isn't what I want to be
• 25:24
  Speaker 3: doing or what I actually like or what's best for
• 25:27
  Speaker 3: my project or my art.
• 25:29
  Speaker 2: Right.
• 25:29
  Speaker 3: I've been thinking about that in particular to students at
• 25:34
  Speaker 3: the university level or in high school, just people learning, right?
• 25:38
  Speaker 3: And I was listening to this podcast and this educator
• 25:42
  Speaker 3: was talking in particular about students using AI. And she
• 25:46
  Speaker 3: was basically examining and thinking about the social contracts that
• 25:52
  Speaker 3: we have with each other. If I'm a student and
• 25:55
  Speaker 3: I'm here to be learning... theoretically, the social contract implies
• 26:00
  Speaker 3: that I'm going to be doing my work.
• 26:02
  Speaker 2: Yeah. Right.
• 26:02
  Speaker 3: And then when we hand that off to an AI
• 26:04
  Speaker 3: to do it, well, we're breaking that social contract. And
• 26:07
  Speaker 3: I thought that that was a really fantastic way to
• 26:10
  Speaker 3: think about the AI usage in general, I guess, how
• 26:14
  Speaker 3: we use it or how folks are using it in
• 26:16
  Speaker 3: like classrooms or to like write an essay or to
• 26:19
  Speaker 3: write an article and how we're breaking these social contracts.
• 26:24
  Speaker 3: And that's not necessarily like a positive thing. We're not saving,
• 26:27
  Speaker 3: maybe we're saving time.
• 26:28
  Speaker 2: We generally, right. But we're not actually learning anything.
• 26:33
  Speaker 3: And yeah, it just, it got me thinking about how
• 26:37
  Speaker 3: I think, Specifically, like with public schools in K through 12,
• 26:43
  Speaker 3: like there's so much like standardized testing.
• 26:45
  Speaker 2: Right.
• 26:45
  Speaker 3: There's like these rules of like you have to hit
• 26:47
  Speaker 3: these certain marks that like students aren't even reading full
• 26:51
  Speaker 3: books anymore. Like they're reading excerpts of things and there
• 26:55
  Speaker 3: are literacy rate is plummeting. And so there's all these
• 26:58
  Speaker 3: factors as to why people are not learning more. in
• 27:02
  Speaker 3: ways that maybe they were 20 years ago, even though, like,
• 27:04
  Speaker 3: technology has actually gotten better. And so there's, like, some
• 27:07
  Speaker 3: tension there in while maybe there's more access in terms
• 27:11
  Speaker 3: of technology and AI, but are we actually learning better?
• 27:16
  Speaker 2: I don't think so.
• 27:18
  Speaker 1: Like, in theory, everybody has a tablet because everyone has
• 27:21
  Speaker 1: a phone, including children. But that doesn't mean that they're
• 27:24
  Speaker 1: reading more.
• 27:25
  Speaker 2: Right.
• 27:26
  Speaker 1: They're probably scrolling on TikTok and watching fruity novellas. Right. And, like,
• 27:30
  Speaker 1: AI verticals that are very grotesque.
• 27:32
  Speaker 2: Right.
• 27:32
  Speaker 1: I've sent you some of these, like, clearly, like, based
• 27:36
  Speaker 1: in Latin American pueblos where there's, like, an awful mean
• 27:41
  Speaker 1: tia who's, like, bossing everyone around. And they always make her, like,
• 27:46
  Speaker 1: a plus-size woman. And sometimes with, like, a green face who's, like,
• 27:50
  Speaker 1: basically verbally abusive and, like, taking advantage of all the
• 27:54
  Speaker 1: people around her and, like— eating all the food and
• 27:56
  Speaker 1: spending all the money. Pero es bien mandona. It's very
• 28:00
  Speaker 1: strange because I'm like, Who is creating these, like, AI dysfunctional, like,
• 28:06
  Speaker 1: Latin American families in, like, an impoverished pueblo?
• 28:11
  Speaker 2: Interesting.
• 28:12
  Speaker 1: With, like, a mean, abusive mandona tia.
• 28:15
  Speaker 2: Mm-hmm.
• 28:15
  Speaker 1: And, like, who had that idea?
• 28:17
  Speaker 2: Right.
• 28:18
  Speaker 1: And it's, like, its own genre, and I keep seeing it.
• 28:21
  Speaker 1: And I wonder if listeners— I' m sure listeners are seeing these, too,
• 28:24
  Speaker 1: because it's so specific, and it's, like, creating this very negative—
• 28:32
  Speaker 1: portrayal of women, of Latinas and of like plus size
• 28:37
  Speaker 1: women and like women in seemingly like impoverished circumstances. And
• 28:43
  Speaker 1: it's almost like they, I don't know, like I don't
• 28:47
  Speaker 1: get it. It's like propaganda. Right. It's like very strange AI,
• 28:51
  Speaker 1: like vertical propaganda.
• 28:53
  Speaker 2: Well, we've talked about this before, right?
• 28:55
  Speaker 3: How if humans are programming the AI, there is a
• 29:02
  Speaker 3: racist factor into how like AI is being programmed or
• 29:06
  Speaker 3: technology is being programmed because research does show that the
• 29:09
  Speaker 3: algorithm is racist. And so that does not surprise me.
• 29:14
  Speaker 3: And it shows, you know, how the AI is probably trained, right?
• 29:19
  Speaker 3: With like a particular vision in mind, like white, cis, het.
• 29:25
  Speaker 3: You know, and that's like what the standard is. And
• 29:28
  Speaker 3: then like the more layered it gets, like there's probably
• 29:30
  Speaker 3: a lot of nuance missed because of course it's AI.
• 29:33
  Speaker 1: Yeah. I'm also going to say this. If AI is
• 29:36
  Speaker 1: like scanning the web for media and then creating its own,
• 29:41
  Speaker 1: there are a lot of Latino comedians and influencers who
• 29:46
  Speaker 1: have made names for themselves. doing this Latina cosplay.
• 29:52
  Speaker 2: Yeah.
• 29:53
  Speaker 1: Cosplay of tias and moms where it's always negative.
• 29:56
  Speaker 2: Mm-hmm.
• 29:57
  Speaker 1: And she's always mandona and nonsensical and low-key abusive and
• 30:02
  Speaker 1: all the chancla stuff that everybody does. So I feel
• 30:05
  Speaker 1: like a lot of those skits have likely influenced a
• 30:11
  Speaker 1: lot of these, like, AI verticals.
• 30:13
  Speaker 2: Interesting.
• 30:14
  Speaker 1: I see a complete and total, like, a direct link
• 30:17
  Speaker 1: between Because it's it's it's like its own genre. And
• 30:21
  Speaker 1: it's you know, and I think like the black community
• 30:24
  Speaker 1: has pointed out its own issue within that community of
• 30:28
  Speaker 1: the like Tyler Perry, Martin Lawrence, Eddie Murphy cosplay of
• 30:34
  Speaker 1: black women and of like donning these. these suits and
• 30:38
  Speaker 1: putting on these characters that are these very negative portrayals
• 30:41
  Speaker 1: of Black womanhood. And in Latino digital comedic spaces, there
• 30:46
  Speaker 1: are a lot of men who do that as well.
• 30:48
  Speaker 2: Yeah. Don't go anywhere, Locamores. We'll be right back.
• 30:56
  Speaker 1: And so like, yeah, where did they get this idea from? Oh,
• 30:59
  Speaker 1: from us. From years of doing these skits.
• 31:03
  Speaker 2: Yeah.
• 31:04
  Speaker 1: Yes.
• 31:04
  Speaker 3: I'm sure it's a lot of things, right? Because the
• 31:07
  Speaker 3: AI is being fed by, well, everything, you know, that exists.
• 31:11
  Speaker 1: Yeah. It's wild stuff. And then I don't even know
• 31:13
  Speaker 1: if we have time to get into the AI relationships. Yeah.
• 31:15
  Speaker 2: No, we don't have time for that.
• 31:17
  Speaker 1: But that's another episode.
• 31:18
  Speaker 3: We've been talking about doing that. There's been a lot of,
• 31:23
  Speaker 3: I think, deep dives into specific people who have gone
• 31:26
  Speaker 3: on to have, quote, relationships with an AI bot.
• 31:29
  Speaker 1: Yeah. Next time. Next time.
• 31:33
  Speaker 3: But I do want to end with an example of
• 31:35
  Speaker 3: a city that is taking regulation into their own hands.
• 31:39
  Speaker 3: New York City is banning the use of generative AI
• 31:42
  Speaker 3: for public school students from 2K through eighth grade, which
• 31:46
  Speaker 3: is part of a larger move to restrict classroom technology
• 31:49
  Speaker 3: in the nation's largest school district. This is a one-year moratorium,
• 31:54
  Speaker 3: and Mayor Zoran Mamdani said the district will, quote, discontinue
• 31:58
  Speaker 3: or disable the AI components of more than 38 previously
• 32:01
  Speaker 3: allowed programs that do not meet our new safety and
• 32:04
  Speaker 3: oversight standards.
• 32:06
  Speaker 1: I'm screaming that it has to start at 2K.
• 32:08
  Speaker 2: Yeah.
• 32:09
  Speaker 1: Like, the kindergartners have to be banned from using generative AI.
• 32:13
  Speaker 3: They have to ban the tablets, I think, is probably
• 32:16
  Speaker 3: what it's about.
• 32:17
  Speaker 1: Yeah, because it's on there. If they have the tablet,
• 32:19
  Speaker 1: they can get to anything.
• 32:20
  Speaker 2: Lord.
• 32:21
  Speaker 1: Yeah.
• 32:22
  Speaker 3: Oh, my.
• 32:23
  Speaker 2: Yeah.
• 32:23
  Speaker 3: I mean, I think before there's any federal regulation because
• 32:27
  Speaker 3: of the country that we live in, so much is
• 32:29
  Speaker 3: state by state right now. And so there will probably
• 32:31
  Speaker 3: be like city, countywide, then statewide restrictions and regulations that
• 32:36
  Speaker 3: are passed when it comes to AI.
• 32:37
  Speaker 2: Yes.
• 32:38
  Speaker 1: It will not happen on a federal level.
• 32:39
  Speaker 2: No. So what does that mean? That means call your representatives. Yes.
• 32:43
  Speaker 3: You have to go to your you have to go
• 32:45
  Speaker 3: to your local city council.
• 32:46
  Speaker 1: You have to talk to your city.
• 32:47
  Speaker 2: You have to talk to folks.
• 32:49
  Speaker 3: Because I mean, this is another episode. But all while
• 32:52
  Speaker 3: this is unfolding, there's so much coverage and investigative reporting
• 32:57
  Speaker 3: being done about flock, which we first learned about when
• 33:00
  Speaker 3: we were interviewing mayoral candidate Ray Huang. And I didn't
• 33:05
  Speaker 3: know that we had flock cameras in L.A. We have
• 33:07
  Speaker 3: there's a citywide contract with flock. And what they do
• 33:12
  Speaker 3: is they snap photos of your license plate, allegedly. But
• 33:17
  Speaker 3: there's a lot of data surveillance issues that come up
• 33:22
  Speaker 3: with Flock. And so now across the country, there are
• 33:26
  Speaker 3: different cities that are either canceling their contracts with Flock.
• 33:31
  Speaker 3: I've seen folks talk about it at their local city
• 33:33
  Speaker 3: council meetings because they want to know, like, how is
• 33:35
  Speaker 3: flock being used in our neighborhoods, right? And so all
• 33:38
  Speaker 3: these tech companies, it's all connected with the usage of AI,
• 33:42
  Speaker 3: with the surveillance state. And I think, like, while we
• 33:47
  Speaker 3: can easily, like, lose hope and focus on the doom
• 33:49
  Speaker 3: and gloom of it all because it is scary, also
• 33:52
  Speaker 3: thinking about, okay, how can we be active and make change?
• 33:56
  Speaker 2: And that's at the local level.
• 33:59
  Speaker 1: You have to talk to your city council person.
• 34:01
  Speaker 2: You have to go to your neighborhood meetings. You just
• 34:04
  Speaker 2: have to.
• 34:04
  Speaker 1: Even if you don't like them.
• 34:06
  Speaker 2: Yes.
• 34:07
  Speaker 1: That's your representative. You have to talk to them if
• 34:10
  Speaker 1: you want to make change in your community. They control budgets.
• 34:15
  Speaker 1: They set policy. They enforce policy. They change policy. They
• 34:19
  Speaker 1: allocate money. That's just who does it.
• 34:21
  Speaker 2: Yeah.
• 34:22
  Speaker 1: You don't have to align with them politically.
• 34:23
  Speaker 3: You don't have to align with them, but it is
• 34:25
  Speaker 3: good to go to your meetings. Fernando and I have
• 34:27
  Speaker 3: been talking about that because we live in L.A.
• 34:30
  Speaker 2: County.
• 34:31
  Speaker 3: And so where we live is actually very conservative. Like,
• 34:37
  Speaker 3: the policies are conservative. The city council is conservative. The
• 34:42
  Speaker 3: mayor is conservative. And it's a small city, so it's
• 34:47
  Speaker 3: a low voter turnout.
• 34:48
  Speaker 2: And, you know, I vote every local election.
• 34:51
  Speaker 3: That doesn't mean that the person I'm voting for is
• 34:53
  Speaker 3: getting elected, though, because I do vote for the more
• 34:56
  Speaker 3: progressive one. But because I live in a conservative city,
• 34:58
  Speaker 3: that means that the conservative— is the majority.
• 35:02
  Speaker 2: And so as.
• 35:04
  Speaker 3: Much as we hate it, as much as we don't
• 35:06
  Speaker 3: want to go, we're like, we actually have to organize
• 35:09
  Speaker 3: where we live because like we go out to all
• 35:12
  Speaker 3: these other parts of LA that we love and we
• 35:15
  Speaker 3: support and we're active members in other parts of the community,
• 35:18
  Speaker 3: but in our own home, um, We aside from voting
• 35:22
  Speaker 3: and like being neighborly with our neighbors, we're not attending
• 35:25
  Speaker 3: the meetings. And so we can you can only complain
• 35:29
  Speaker 3: for so long. And so we are now like reflecting
• 35:32
  Speaker 3: and thinking, OK. Where if we want to build a
• 35:35
  Speaker 3: progressive coalition where we live, well, we have to we
• 35:38
  Speaker 3: have to do it ourselves.
• 35:39
  Speaker 1: Right.
• 35:40
  Speaker 3: Because we know that not everybody around us is conservative,
• 35:44
  Speaker 3: but we do know that the conservative folks are the
• 35:46
  Speaker 3: ones going to the meetings.
• 35:47
  Speaker 1: Right.
• 35:47
  Speaker 3: They're the ones campaigning for their their candidate. They're the
• 35:52
  Speaker 3: ones voicing. You know, there was a townhome proposal.
• 35:58
  Speaker 2: That got approved yesterday. And it was townhomes, condos and
• 36:02
  Speaker 2: affordable housing. It got passed. There was an outcry. Sure.
• 36:07
  Speaker 2: God forbid there be affordable housing in our city. Yeah.
• 36:11
  Speaker 2: What kind of people is it going to attract?
• 36:12
  Speaker 1: Right.
• 36:13
  Speaker 2: Like that was the sentiment. Right. And so what does
• 36:16
  Speaker 2: the city council do?
• 36:18
  Speaker 3: Well, they put it back on the agenda to discuss,
• 36:21
  Speaker 3: even though it had already been approved.
• 36:23
  Speaker 2: Sure.
• 36:24
  Speaker 3: So all that to say, we live in a surveillance state.
• 36:28
  Speaker 3: AI is booming. The billionaires are getting richer. We have
• 36:32
  Speaker 3: a trillionaire now that lives in this country. And what
• 36:35
  Speaker 3: can we do but organize at our local level and
• 36:37
  Speaker 3: be involved?
• 36:39
  Speaker 1: You have to start somewhere.
• 36:39
  Speaker 2: And take care of each other.
• 36:40
  Speaker 1: Yeah, you have to start somewhere. Start in your neighborhood
• 36:43
  Speaker 1: and go to a city council meeting.
• 36:45
  Speaker 3: Yes, it doesn't start and end at city politics, obviously.
• 36:51
  Speaker 1: But it's a start.
• 36:52
  Speaker 3: But it is a start because local policy literally affects
• 36:55
  Speaker 3: our daily lives.
• 36:57
  Speaker 1: You are much more likely to get the ear of
• 37:00
  Speaker 1: a school board member, of a city council person, than
• 37:04
  Speaker 1: you are a senator or the president, right?
• 37:07
  Speaker 2: But still call your senators.
• 37:09
  Speaker 1: Call everybody.
• 37:09
  Speaker 2: Call everybody.
• 37:10
  Speaker 1: Call everybody.
• 37:11
  Speaker 2: You heard it here. Call everybody. It's the only way.
• 37:15
  Speaker 2: And talk to each other. Literally.
• 37:17
  Speaker 1: Talk nice.
• 37:18
  Speaker 2: Talk nice.
• 37:19
  Speaker 1: Talk nice.
• 37:20
  Speaker 2: Get offline.
• 37:20
  Speaker 1: Get offline.
• 37:21
  Speaker 2: Touch grass.
• 37:21
  Speaker 1: Touch grass.
• 37:23
  Speaker 2: All right, Locamores. This has been another capítulo of Locatora Radio.
• 37:27
  Speaker 2: We'll catch you next time.
• 37:29
  Speaker 1: Besitos.