In “The AI Con,” Emily Bender and Alex Hanna document the long history of AI hype and how it’s harming everyday people
As billions of dollars have flowed into artificial intelligence startups, particularly those based in San Francisco, a growing number of people have raised concern about that investment creating a massive financial bubble that’s bound to pop.
In their upcoming book, “The AI Con: How to Fight Big Tech’s Hype and Create the Future We Want,” researchers Emily Bender and Alex Hanna argue the industry is surrounded by a parallel bubble of hype. With their book, the pair hope to pierce that bubble and help everyday people see how AI really works, the harms it’s causing and who’s actually benefiting from it.A sociologist by background, Hanna was a member of Google’s Ethical AI team before leaving in frustration in 2022 at what she saw as the company’s failure to address the harms its technology was causing. She then joined the Distributed AI Research Institute, which Timnit Gebru — her former manager at Google — founded. Bender, a linguistics professor at the University of Washington who’s also on the faculty of the institution’s engineering and information schools, has been writing for years about AI and machine-learning systems work and the harms they can cause. She’s perhaps most famous for a paper in which she and her collaborators, including Gebru, dubbed large-language models like OpenAI’s GPT-3 “stochastic parrots” for their ability to mimic human speech in a way that convinces people there’s actual meaning behind it. In “AI Con,” which goes on sale next Tuesday, Bender and Hanna start essentially from the beginning, questioning even the term “artificial intelligence.” “AI,” they argue, is just a marketing name for a disparate collection of technologies like language models, recommendation engines and automatic classification systems that helps obscure what those technologies do and how they work. Whenever possible, rather than referring to those systems as AI, they try to specifically and descriptively name them. They like to refer to generative AI chatbots like ChatGPT, for example, as “text extrusion machines.” But they don’t stop there. They dive into the history of the technology and its hype. In succeeding chapters, they argue AI in its various forms is undermining labor, being used as a cheap and poor replacement for health care and other social services, and built on stolen works from artists and other creators and is sabotaging an assortment of creative work. They also try to puncture the notions that this technology is inevitable and that it is leading to some kind of potentially humanity-threatening superintelligence. That’s just more hype, they argue. Per the book’s title, they offer ways people can fight back and fight for a fairer, more just world — starting with just refusing to use AI services. In an interview last week, Bender and Hanna talked about the harms they see from AI, why they think AI is in a bubble that will burst and what they hope readers get out of their book. This interview has been edited for length and clarity.Bender: A goal, at least, is for people to be able to spot the hype and push back against it and push back especially against these narratives of inevitability. It's really frustrating to me to hear people who say, “Well, it's here to stay. We just have to learn to live with it." And to that, we're like,"Well, no, actually." One of the examples I like to point to is the hole in the ozone layer. That was a global-scale problem, and we addressed it by coming together with some regulation, and now we have many problems, but that's not one of them. My hope is that with the book, as we pull back the curtain on what's actually working behind these systems, we will empower people who feel a little bit off about this but don't know quite how to say why, to be able to be confident that they're seeing it the right way around and learn how to articulate that to people around them. Hanna: The nice thing about it is I think the book does have that power. A lot of the folks we've talked with on this, they're often librarians or artists or other creatives, and they're like, “Yeah, my sense is that this is icky. But you're actually putting your finger on a few core elements of it —” have no access to meaning. They are in some cases used as a tool to bludgeon workers, especially in creative industries. And they are doing nothing but polluting the information ecosystem and making it hard for people like librarians and writers and artists to do their job effectively with respect to maintaining and producing knowledge and culture.Bender: I think you can divide the harms into a few buckets. One is the harms that come about in the production of the technology. So you've got the data theft, you've got the labor exploitation, you've got the environmental impacts, and all of that is justified by the hype. Ex // Top Stories SF taking its time to hire city's first inspector general As scandals persist, SF says it is making progress on forming an anti-corruption office Trans org out nearly $1M after SF axes Dream Keeper grants The organization, which supports The City’s trans community, is struggling to keep its services going after two of its grants were cancelled Doubt California’s clout? Congress’ push to end EV mandate proves it As Republicans expressed alarm that the rest of the country might follow the Golden State’s push to promote electric vehicles, California’s power was laid bare Then you've got these situations where the hype allows people in power to put in a so-called AI system as a Band-Aid for a place where social services are severely lacking. And a lot of harms can come from that. You've got places where you have automatic decision systems being deployed. And every time someone says,"Oh, it's an AI and it's so smart. Surely it's making fair, objective decisions, because they’re in the computer,” that provides cover for really horrible reproductions of our systemic racism in things like welfare allocation and pretrial sentencing and so on. Hanna: It's really hard to rank it. This is why the book isn't necessarily organized around the types of technology. We arrange it by the area in which these interventions are being made. The turn to AI is a move that has been consistent with worsening conditions in different industries — in journalism and higher ed and in science, in creativity or in graphic design. Any place where human discretion or even knowledge needs to be reliable and consistent, that is being taken away and supplanted with these tools. One of the things you do with the book is to center people, talking about how these technologies are built on exploiting people or cause real harms to people, and that, ideally, these technologies would be used to assist people. Why did you guys feel the need to do that? Bender: One of the ways I often put it is to say, “We’ve got to keep the people in the frame.” AI is always people. And a lot of this, the parlor trick involves hiding the actual inputs that were done by people, so that it looks like the people who created this technology have actually either been a god or created a god. Hanna: The notion of keeping the human in the frame dispels this notion that there is this ability of things to be completely autonomous. This is kind of what we talk about when we talk about the “Doomers,” who say there's going to be these machines that are going to become autonomous, run themselves, self-replicate, etc., which is patently absurd. If anyone's worked in tech support, they know that's really laughable. But also the fact that there are people who are both controlling or providing data for these systems behind the curtain. There's also a set of incentive structures and people who are orchestrating these things. The hype doesn't perpetuate independently of individuals. It is certain people couched in certain types of institutions perpetuate hypes for their own gains and that's important to point to. You talk toward the end about the bubble seeming to have started to burst. But there’s more money than ever flowing into this sector. Do you still see this as a bubble that's going to pop? Bender: This can't be other than a bubble, because there's no there there. Basically what they've got in the large-language models is a technology that can mimic human language use in many, many contexts. But human language use is not the same thing as actually providing the services that they’re . Is it surprising that the bubble is getting bigger before popping? Sure. Does that mean it's not a bubble? No. Hanna: I'm happy to put the stake in the ground.These big cash flows are to me kind of like, “OK, this is the year that this is really going to make or break it, right?” We saw a few hints of that earlier this year when DeepSeek was released. There was a sharp decline in Nvidia’s stock and associated stocks. Is this going to be the same kind of thing? I don't know. It's hard to say. But there's no there there. The path to monetization is very narrow.Bender: These systems are consistently sold as reasoning machines. If we had something that could actually do what they say it can do, then there'd be tremendous possibility for disruption in all kinds of sectors. But there's not. It's not there. So, that's what I mean.Bender: Not only are they not as sophisticated as they’re being sold, but there's no pathway from a synthetic text extruding machine to something that can do what they're claiming it can do. The Sam Altmans and the Amodeis at Anthropic would like us to believe that something called artificial general intelligence is right around the corner and that these were just mere steps away from it on this particular path. It's inevitable, it's close. It's just a question of who gets there fastest and first. And none of that is proven, and there's actually no good reason to believe any of it. But that functions as some of the marketing that's keeping the hype bubble inflated. If you have a tip about tech, startups or the venture industry, contact Troy Wolverton at twolverton@sfexaminer.com or via text or Signal at 415.515.5594.
Artificial Intelligence Openai Anthropic Emily Bender Alex Hanna Ai Harms
United States Latest News, United States Headlines
Similar News:You can also read news stories similar to this one that we have collected from other news sources.
Trump responde con un arancel del 125 % en la creciente guerra comercial con ChinaChina no parece interesada en negociar, tal como algunos otros países han empezado a hacerlo.
Read more »
‘Book by book': Michigan community takes novel approach moving local bookstore Serendipity Books in Chelsea, Michigan, had its entire inventory transferred — in alphabetical order, no less — in less than two hours.
Read more »
Trump asiste al funeral de Francisco, con quien que discrepó públicamente, y se reúne con ZelenskyyROMA (AP) — El presidente de Estados Unidos, Donald Trump, era uno de los más de 50 jefes de Estado y otros dignatarios que asistían el sábado al funeral del papa Francisco, donde presentó sus respetos al líder católico con el que discrepó abiertamente sobre varias cuestiones.
Read more »
Trump asiste al funeral de Francisco, con quien que discrepó públicamente, y se reúne con ZelenskyyEl presidente de Estados Unidos, Donald Trump, era uno de los más de 50 jefes de Estado y otros dignatarios que asistían el sábado al funeral del papa Francisco, donde presentó sus respetos al líder católico con el que discrepó abiertamente sobre varias cuestiones.
Read more »
Scorsese producirá documental realizado con el papa Francisco sobre su trabajo con el cineMartin Scorsese está produciendo un documental realizado con el papa Francisco que narrará el trabajo del difunto pontífice con el cine en el movimiento educativo global que fundó antes de su muerte.
Read more »
Ataque con drones alcanza un barco con ayuda para Gaza, según grupo humanitarioTEL AVIV, Israel (AP) — Un dron alcanzó el viernes una embarcación que transportaba ayuda a Gaza frente a la costa de Malta, explicó en un comunicado Codepink, un movimiento de base por la paz y la justicia social.
Read more »