Emily M. Bender

Croxson was quoting Eliezer Yudkowsky word for word when he was talking to Wired, um, earlier this month, so yeah.

And to me, one of the other frustrating things is, you know, it's not like there haven't been people putting out, you know, much more reasonable takes about this, right? You know, we, you know, the people who I quote and, you know, interviewed for my book, um, the people who I've had on this podcast, last week's guest or the person who w- will have been last week when this comes out, uh, uh, Emily Bender.

Um, we've got, uh, uh, Timnit, uh, Gebru.

We've got, um, you know, a lot of good people doing good work on this, uh, who just are not being brought into Um, especially the policymaking conversation and the way that I would like to see them there, right? Like I, I want...

Mm-hmm.

Which again, people, it's just a landscape people don't necessarily have a lot of knowledge of.

But, you know, in the beginning you had what you can think of as like academic AI safety, which that's where you can find Emily Bender, that's where you'll find Timnit, uh, Gebru.

Um, you would think of them as being in like academic AI safety worlds, which, uh, are, are concerned with a lot of proximate dangers of AI.

Yeah.

Hey, we're out of time, and I'm going to give the last word to Melanie Hewson, who basically just in a comment said, are you familiar with the work of Emily Bender and Nana Eni about anthropomorphic language regarding LLMs? They have proposed alternate vocabulary.

So that's something we'll leave everyone with if we want to continue to pursue this topic.

Thank you very, very much, Emmanuel.

Those are synonyms.

But anyway, I got interested in this because Emily Bender posted something on LinkedIn about inshitification and how it was affecting generative AI.

And of course, she's a big critic of AI tools in general.

So I looked it up and I became interested.

They ended up canceling projects that were threatening, that these people were concerned about.

And this is when Timnit Gebru got fired.

And so back to Emily Bender, Timnit Gebru and Emily Bender wrote that viral research article about stochastic parrots.

Yeah.

But is what they're saying the reality, or is it pure theatrics? Joining me today is someone who thinks the hype around AI is massively overstated and that we are being distracted, possibly on purpose, from the real problems presented by this technology.

Professor Emily M. Bender is the director of the Computational Linguistics Laboratory at the University of Washington.

She's also written a book on the topic.

It's called the AI Con: How to Fight Big Tech's Hype and Create the Future We Want.

It's, it's happening.

It's just gonna be great instead of bad." And then the other opposing position is, "Actually, that's all nonsense," and that's where I sit.

Professor Emily M. Bender, I very much appreciate your knowledge and expertise on this subject today.

Thank you so much for joining us.

We're not talking about those stories because we're busy chatting about whether AI is going to kill us in the next 10 years.

And I think there's actually a couple of very well-known ethicists, Emily Bender and Timnit Gebru, who are very, very well known for...

This is their view, basically.

They think that it's all about distracting.

And so, you know, I reached out to her and then...

There's an amazing seminal paper called Stochastic Parrots that one of the authors, Emily M. Bender, is in our film.

And I reached out to her and I just started to collect these different pieces of research.

And it was very organic.

And we give over, again, we, before we've even engaged with it, we've given over our power to this unknown, th-this stochastic.

We can quote Emily Bender, we can quote all the brilliant minds that you already mentioned, right? Who said, "What are these tools actually doing?" And when they ask the question, "Can language models be too big?" Which is part of what we're facing right now, why did they say can language models be too big? Stochastic Parrots paper.

This paper, what it did is it said, "Do we need one model to rule them all?" This is like The Lord of the Rings view.

Sometimes I tell my students, I'm like, "Look, we already said we don't want one, one ring to rule them all, one model to rule them all." And when I say one model, I'm not really talking about these proprietary commercial models.