Latest / Best of Business / Karen Hao: journalist and author on her new book exploring the impact of AI

Transcript

Karen Hao: journalist and author on her new book exploring the impact of AI · Best of Business

• 0:09
  Speaker 1: You're listening to a podcast from News Talk sed B.
• 0:12
  Speaker 1: Follow this and our wide range of podcasts now on iHeartRadio.
• 0:16
  Speaker 2: So just what does the future of AI have in
• 0:19
  Speaker 2: store for us all? None of us really know, do we?
• 0:21
  Speaker 2: Seven years ago, when investigative journalist Karen Howe started looking
• 0:24
  Speaker 2: into Sam Oltman's Open AI, she thought they were one
• 0:27
  Speaker 2: of the good guys. Extensive research and unparalleled access to
• 0:31
  Speaker 2: those closest to the AI race have left Karen with
• 0:34
  Speaker 2: a much different view of tech companies and those leading
• 0:37
  Speaker 2: the charge who work in this space has made her
• 0:39
  Speaker 2: one of the most foremost tech journals covering AI. She's
• 0:42
  Speaker 2: been listed in Time Magazine's one hundred Most Influential People,
• 0:46
  Speaker 2: and Karen is heading to New Zealand for the Auckland
• 0:48
  Speaker 2: Writers Festival with her book Empire of AI and Karen
• 0:52
  Speaker 2: Howe's with me now.
• 0:53
  Speaker 3: Good morning, Hi Francesca, thank you so much for having me.
• 0:57
  Speaker 2: Firstly, and I'm sure my audience knows this, but can
• 1:00
  Speaker 2: we justifine the term AI and AGI?
• 1:04
  Speaker 3: Yes? So, I always like to say that the word
• 1:07
  Speaker 3: AI is like the transportation. It refers to such a
• 1:11
  Speaker 3: wide collection of technologies that can be as different as
• 1:14
  Speaker 3: different from a bicycle to a rocket, and really these
• 1:19
  Speaker 3: technologies are just loosely tied together based on the fact
• 1:22
  Speaker 3: that they are each trying to mimic a certain aspect
• 1:27
  Speaker 3: of human capability. So there are systems that mimic our
• 1:32
  Speaker 3: language abilities, there are systems that mimic our listening abilities,
• 1:36
  Speaker 3: are visual recognition and it ultimately each of these these
• 1:45
  Speaker 3: different types of AI technologies can actually be vastly different
• 1:48
  Speaker 3: and have very different cost benefit trade offs. AGI is
• 1:53
  Speaker 3: a term that refers to artificial general intelligence and is
• 1:56
  Speaker 3: meant to evoke this idea that somehow the ultimate quest
• 2:02
  Speaker 3: for building AI technologies is to replicate all human care
• 2:08
  Speaker 3: capabilities in a single system. But one of the problems
• 2:12
  Speaker 3: with AGI is that it's very poorly defined, in that
• 2:16
  Speaker 3: we don't really have good ways of defining what human
• 2:19
  Speaker 3: intelligence is and whether or not we've actually reached the
• 2:23
  Speaker 3: goalposts of AGI. And so my criticism of companies that
• 2:28
  Speaker 3: use this kind of rhetoric is that they are essentially
• 2:31
  Speaker 3: just leveraging the lack of a shared definition as a
• 2:34
  Speaker 3: way to just hype up their technologies.
• 2:37
  Speaker 2: Bringant, thank you for starting us off with it. That's excellent.
• 2:41
  Speaker 2: Is AGI about science? The betament of humankind, power or
• 2:45
  Speaker 2: money or all of the above.
• 2:48
  Speaker 3: Primarily it's about power and money. AGI at this point
• 2:54
  Speaker 3: is you could say, almost a quasi religious quest, and
• 2:59
  Speaker 3: just like other religions, there is this idea that it
• 3:05
  Speaker 3: is ultimately meant to bring people who believe in AGI
• 3:10
  Speaker 3: to heaven, and it could also be a tool that
• 3:14
  Speaker 3: damns people to hell. But the reason why this is
• 3:19
  Speaker 3: such a potent narrative and is used so often is
• 3:22
  Speaker 3: because companies that are saying that they're building AGI are
• 3:26
  Speaker 3: ultimately able to use this to justify the accruval of
• 3:29
  Speaker 3: an enormous amount of money and power. So all of
• 3:32
  Speaker 3: the tech companies that are developing AI and using the
• 3:38
  Speaker 3: AGI narrative, like open Ai, Anthropic, Xai, Google, they now
• 3:45
  Speaker 3: are the most valuable companies in the world, and they
• 3:50
  Speaker 3: still want to continue accruing yet more capital to build
• 3:54
  Speaker 3: their technologies. They are deploying more capital than ever any
• 4:00
  Speaker 3: other industry ever before to build massive infrastructure projects all
• 4:04
  Speaker 3: around the world. So yeah, at the heart of this
• 4:07
  Speaker 3: conversation is indeed money and power, and.
• 4:11
  Speaker 2: We'll dig into that shortly. But AI open AI is
• 4:17
  Speaker 2: sort of the center of the book that you've written
• 4:19
  Speaker 2: and open ai began as an ultruistic organization. It was
• 4:22
  Speaker 2: a nonprofit open organization is implied by the title, It
• 4:26
  Speaker 2: had this idealistic governance structure. What was its original objective?
• 4:32
  Speaker 3: So I have a different answer for this now versus
• 4:35
  Speaker 3: when I first started covering open AI back then, I
• 4:39
  Speaker 3: thought they had this kind of pure hearted, mission driven
• 4:43
  Speaker 3: objective of trying to ensure that AI advancement ultimately goes
• 4:48
  Speaker 3: well and benefits all of humanity, and that then there
• 4:52
  Speaker 3: was some kind of profit motive that slowly corrupted the
• 4:56
  Speaker 3: organization over time. But in hindsight, I think they're real objective.
• 5:03
  Speaker 3: Their objective that wasn't actually explicitly stated when public messaging
• 5:09
  Speaker 3: was a desire to become the dominant AI lab in
• 5:14
  Speaker 3: the world or the dominant force shaping AI technology development
• 5:18
  Speaker 3: in the world, and you could say that that has
• 5:23
  Speaker 3: never changed, and what we're seeing now is just a
• 5:26
  Speaker 3: continued manifestation of this desire for dominance.
• 5:31
  Speaker 2: Because the hugely competitive these II companies. This is what
• 5:33
  Speaker 2: you learn through the book. There is this drive to
• 5:36
  Speaker 2: be first, these ideological clashes which makes each player think
• 5:40
  Speaker 2: that they need to be first to deliver AI for
• 5:42
  Speaker 2: their purpose because it's the right purpose, and yet that
• 5:46
  Speaker 2: ambition doesn't quite sit with the ethically minded AI that
• 5:50
  Speaker 2: they claim to sort of start out with, does it.
• 5:55
  Speaker 3: That's exactly right. I mean, the thing that's so interesting
• 6:00
  Speaker 3: is I often use this analogy that the AI world
• 6:03
  Speaker 3: is a little bit like Doune the Science Fiction Network
• 6:06
  Speaker 3: by Frank Herbert. And the reason is because in Dune,
• 6:11
  Speaker 3: the main character poly treatise he knows that there are
• 6:15
  Speaker 3: these myths that are seated on different planets, including his
• 6:18
  Speaker 3: own Oracus, and that that myth help portrays himself as
• 6:23
  Speaker 3: the coming of the Messiah, and that if he steps
• 6:25
  Speaker 3: into the myth, he can leverage it to control the people.
• 6:28
  Speaker 3: And so he steps into it, but he genuinely believes
• 6:33
  Speaker 3: that this is to ultimately achieve a good purpose, like
• 6:37
  Speaker 3: he is doing it in a morally righteous way. And
• 6:41
  Speaker 3: then he also begins to lose himself in the myth
• 6:44
  Speaker 3: because every day he's living and embodying this myth, and
• 6:48
  Speaker 3: the myth and reality start to blur for him, and
• 6:51
  Speaker 3: he starts to wonder whether or not actually, in fact,
• 6:53
  Speaker 3: he is the Messiah and the myth is real. And
• 6:56
  Speaker 3: this is just such an applicable metaphor for how the
• 7:01
  Speaker 3: AI world works, because there's all this myth making about
• 7:05
  Speaker 3: AGI that many people know is myth making. They're leveraging
• 7:10
  Speaker 3: it as this marketing rhetoric to accrue more money and
• 7:13
  Speaker 3: power to the company. But they also exist in this
• 7:18
  Speaker 3: space where myth and reality are blurred. They breathe this
• 7:22
  Speaker 3: myth day in and day out, and they begin to
• 7:24
  Speaker 3: wonder whether or not it's true, And at the end
• 7:27
  Speaker 3: of the day they're like, irrespective of whether or not
• 7:29
  Speaker 3: it's true, both myth making and this quest is ultimately
• 7:34
  Speaker 3: for good. We are in the moral we are we
• 7:38
  Speaker 3: are doing something morally righteous. And so there is this
• 7:43
  Speaker 3: degree of like to say it in a in like
• 7:49
  Speaker 3: a not too flattering way, self delusion that happens within
• 7:53
  Speaker 3: the AI world, where people are the revealed pursuit is
• 7:57
  Speaker 3: of money and power, but their stated pursuit continues to
• 8:01
  Speaker 3: be this really these like high minded ideals where if
• 8:06
  Speaker 3: you talk with most people in the AIR world, they
• 8:08
  Speaker 3: genuinely believe that they are on the quest to bringing
• 8:12
  Speaker 3: benefit to all of humanity.
• 8:13
  Speaker 2: And Karen, who is shaping AI is important, isn't it?
• 8:17
  Speaker 2: Is it being shaped by too small a pull of people?
• 8:21
  Speaker 3: Absolutely. I mean, I think that is the central problem
• 8:25
  Speaker 3: of the A industry today. And the reason why I
• 8:29
  Speaker 3: call these companies empires of AI is because they essentially
• 8:35
  Speaker 3: are these entities that have accrued an enormous amount of power,
• 8:41
  Speaker 3: and ultimately they are these small number of people are
• 8:46
  Speaker 3: dictating to billions of people around the world how this
• 8:50
  Speaker 3: technology that is fundamentally consequential to the future is going
• 8:56
  Speaker 3: to go, and therefore how everyone in the future is
• 8:59
  Speaker 3: meant to live and relate to each other and to
• 9:02
  Speaker 3: work without any input from those billions of people. And
• 9:07
  Speaker 3: so it doesn't really matter like who is at the
• 9:10
  Speaker 3: top in this kind of scenario, as long as there's
• 9:14
  Speaker 3: this power structure that enables a small group of people
• 9:16
  Speaker 3: to dictate these things to billions of people without any accountability,
• 9:21
  Speaker 3: it really doesn't like just switching out the cast of
• 9:25
  Speaker 3: characters at the top doesn't really solve the problem.
• 9:30
  Speaker 2: It's a very broad question, and I know that it
• 9:33
  Speaker 2: can't be a quick answer, but you know, maybe you
• 9:35
  Speaker 2: could just touch on the social and environmental costs of AGI.
• 9:40
  Speaker 3: So I don't think people fully understand that AI as
• 9:45
  Speaker 3: a digital technology has actually a very physical and involved
• 9:52
  Speaker 3: supply chain that also includes an enormous amount of human labor.
• 9:57
  Speaker 3: So there are two aspects of AI development that are
• 10:02
  Speaker 3: often overlooked because of this lack of awareness about that
• 10:06
  Speaker 3: supply chain. One is that there's an enormous amount of
• 10:09
  Speaker 3: labor exploitation that happens as part of the supply chain,
• 10:13
  Speaker 3: and the other is that there's this massive physical manifestation
• 10:16
  Speaker 3: of AI development which comes in the form of data
• 10:20
  Speaker 3: centers and supercomputers that are popping up all around the world.
• 10:25
  Speaker 3: When it comes to the labor exploitation, I talk in
• 10:29
  Speaker 3: my book about these content moderators that Opening Eye contracted
• 10:33
  Speaker 3: in Kenya to build a content moderation filter that they
• 10:38
  Speaker 3: ultimately put on chat gbt to ensure that users would
• 10:43
  Speaker 3: never be exposed to the horrific toxic content that they
• 10:48
  Speaker 3: were scraping from Reddit and other social media forums to
• 10:51
  Speaker 3: train their AI models. And in the same way as
• 10:55
  Speaker 3: content moderators in the social media are these content moderators,
• 10:59
  Speaker 3: their psychological state of mind just completely devolved. And not
• 11:05
  Speaker 3: only did they break down as individuals, their families broke down,
• 11:08
  Speaker 3: their communities broke down because they are part of a
• 11:11
  Speaker 3: fabric in society and when they can't support themselves. There's
• 11:15
  Speaker 3: suddenly a hole in that fabric, and that fabric begins
• 11:18
  Speaker 3: to unwind. The other aspect, the environmental aspect. People. You know,
• 11:25
  Speaker 3: people are familiar with data centers. Generally speaking, people understand
• 11:30
  Speaker 3: data centers to be essential to our modern digitally mediated lives.
• 11:33
  Speaker 3: But what is not well understood is the degree to
• 11:37
  Speaker 3: which AI data centers are of a fundamentally different breed
• 11:41
  Speaker 3: of data center. So Meta's first data center in twenty eleven,
• 11:46
  Speaker 3: compared to their current AI supercomputer that they are building
• 11:50
  Speaker 3: in Louisiana, the supercomputer is four hundred times the size
• 11:55
  Speaker 3: of that first data center, and it is on track
• 11:58
  Speaker 3: to be one fifth the size of Manhattan and to
• 12:02
  Speaker 3: eventually use the same amount of power as the average
• 12:07
  Speaker 3: power demand in all of New York City. So these
• 12:11
  Speaker 3: facilities are the size of cities and use the power
• 12:16
  Speaker 3: demand of cities, and there are like they're popping up
• 12:21
  Speaker 3: all around the world, Like this is just one facility
• 12:24
  Speaker 3: that I'm talking about. And so the data center expansion
• 12:30
  Speaker 3: is single handedly beginning to reverse the climate gains that
• 12:36
  Speaker 3: we made in the last decade, Because what are these
• 12:39
  Speaker 3: facilities actually being powered by. They're being powered by fossil fuels.
• 12:44
  Speaker 3: The technistry loves to talk this big game about how
• 12:47
  Speaker 3: they're going to use nuclear to power these facilities with
• 12:50
  Speaker 3: more clean with clean energy, but the nuclear build out
• 12:59
  Speaker 3: would not be able to happen nearly fast enough for
• 13:03
  Speaker 3: supporting this data center build out. And so what we're
• 13:07
  Speaker 3: actually seeing is coal plants that were meant to be
• 13:11
  Speaker 3: retired that are instead having their lives extended. Natural gas plants,
• 13:18
  Speaker 3: new natural gas plants that are being built, and there
• 13:21
  Speaker 3: have been many many cases in which these plants or
• 13:27
  Speaker 3: these plants end up. You know, they're not just spewing carbon,
• 13:32
  Speaker 3: they're also spewing air pollutants into communities. And there's an
• 13:37
  Speaker 3: infamous case of Elon Musk building a giant supercomputer called
• 13:41
  Speaker 3: Colossus in Memphis, Tennessee, and using thirty five methane gas
• 13:44
  Speaker 3: turbines to power this facility, and the community that exists
• 13:50
  Speaker 3: near this facility found out that these methane gas turbines
• 13:54
  Speaker 3: had popped up overnight when they started smelling what smelled
• 13:58
  Speaker 3: like a gas leak in their homes, and they suddenly
• 14:01
  Speaker 3: discovered that they know they were having their right to
• 14:06
  Speaker 3: clean air taken away from them. So there's just all
• 14:09
  Speaker 3: of these interlocking harms that are very hidden to the
• 14:16
  Speaker 3: average consumer of AI technologies. When it comes to this
• 14:20
  Speaker 3: technology's development.
• 14:22
  Speaker 2: Karen, AI is moving, it's you predict how it's going
• 14:24
  Speaker 2: to continue to develop.
• 14:26
  Speaker 3: I don't engage in predictions because predictions assume that the
• 14:30
  Speaker 3: future's already written, and all of my work is ultimately
• 14:34
  Speaker 3: about how we today, based on our actions, create the
• 14:39
  Speaker 3: future together. And so what I hope will happen with
• 14:43
  Speaker 3: the AI industry and with AI development is that everyone
• 14:48
  Speaker 3: listening to this program, everyone that's thinking about AI and
• 14:52
  Speaker 3: thinking about how they ultimately want to live their lives,
• 14:56
  Speaker 3: realize the degree of agency that they have individually and
• 15:00
  Speaker 3: collectively to apply pressure to these companies and hold them
• 15:04
  Speaker 3: accountable when they do not like what they see, and
• 15:07
  Speaker 3: ultimately upward in a completely different vision for AI development
• 15:12
  Speaker 3: where people can truly benefit from this technology.
• 15:16
  Speaker 2: Karen, it's been wonderful to talk to you. I thank
• 15:18
  Speaker 2: you so much for your time. Really appreciate it.
• 15:21
  Speaker 1: For more from News Talk sed B, listen live on
• 15:24
  Speaker 1: air or online, and keep our shows with you wherever
• 15:27
  Speaker 1: you go with our podcasts on iHeartRadio