When people ask me what the phrase "artificial intelligence" means, my short answer is, it means "venture capitalists, give me some money".
But OUP asked me for a longer answer, given here and summarized in the section headings:
Preprint available from my publications page:
@emilymbender
Hide, devalue and steal human labour.
@jlperuyero My dude, I am literally posting the encyclopedia entry that I wrote on this topic -- each of those headings is expanded. And yet, you feel the need to correct what I am saying? I'd be astonished at the hubris but then I remember this is Mansplainodon, so yeah.
@emilymbender
I've been following you for some time, and that should be a hint on my according with your work.
But if you take a minimal comment, intending only agreement with your message, as mansplaining, maybe you'll find useful adding some "NO INTERACTING PLEASE" banner to your account.
So sad. So many barriers, trenches, walls and barbed wires set up between real people. No wonder more and more are turning towards AI.
@jlperuyero I have no way of knowing how long anyone has been following me so can't take that as a hint, no.
I am not putting up any barriers. I am asking for you to NOT contribute to the absolute chorus of mansplaining that is mastodon.
If your intent is agreement, then acknowledge what I said rather than correcting!
@emilymbender
Adding is not correcting. Nothing that you have said sounds wrong to me. Reinforcing should be interpreted as agreement, at least in my environment. And consider that,
- english is not my mother tongue
- I've never travelled to an english speaking country
- I have to struggle with a spanish autocorrect to write this 😃
Greetings from the not-so-privileged side of the world.
Adding reads as correcting: you are saying I didn't say enough/didn't say the right thing (and saying that without reading what I posted).
I don't think the issue here is L1/L2 so much as male privilege.
You are replying to an expert on the topic and not acknowledging that expertise.
@emilymbender
(bows silently)
@jlperuyero @emilymbender
Oh come on! Stop agreeing with her!
@emilymbender @jlperuyero
Tone is hard to pick up sometimes, and I don't want to mansplain mansplaining... but I'd just like to echo what these blokes said here:
https://youtu.be/im1amtaN7MM?si=CdYb3C0w3eTc604S
@jlperuyero stop being a reply guy. If you don't want to be sad about people being prickly and asking you to interact differently, just apologise and stop. What after your first toot and maybe the clarification was really necessary? Certainly not telling her to change her profile. Just take a step back.
@emilymbender @jlperuyero Purely for informational reasons, I already have bro tagged in my notes as "yet another 'akshually' guy." Apparently it's a behavior pattern, and I'm sorry as an authoritative woman in a high-visibility field you have to deal with this shit relentlessly.
At least now I have an excuse to block him.
Jul 8, 2026, 6:56 PM@emilymbender @jlperuyero being a woman and having opinions about tech is an incredible experience. some random ass dude will literally come out of nowhere to "uhm, actually" you on everything from the mildest take to YOUR FUCKING ACADEMIC PAPER on the topic. every time.
@emilymbender I'd love to read the pre-print. I don't know if the link is live (yet), but I tried it now and it gave me a 404. https://faculty.washington.edu/ebender/papers/Bender-AI-2026.pdf was the location the pre-print linked me to.
@gregdosh Thank you -- should be fixed now.
Jul 8, 2026, 4:56 PM@emilymbender This looks great. I am eager to read your preprint, but am getting a 404 for this URL:
https://faculty.washington.edu/ebender/papers/Bender-AI-2026.pdf
@aarbrk Should be fixed now. Please try again.
Jul 8, 2026, 5:03 PM@emilymbender All better! Thanks so much.
@emilymbender Looks great, thank you for sharing the preprint!
Jul 9, 2026, 1:08 AM@emilymbender thank you for making this, and most of your other publications, easily and freely available. I've bookmarked the page for future delving.
Jul 8, 2026, 5:13 PM@emilymbender been looking forward to this thank you for sharing
@emilymbender don't forget: AI is automated bias and discrimination for usual or unusual reasons.
This comment is here only as a break from the the mansplaining.
(Also, ty for posting here)
Jul 8, 2026, 5:31 PM@emilymbender Artificial data collection. Intelligence is data collection in English. So is the CIA not a very smart US agency, but the data collection agency for the US. Intelligenz is in German beyond that. Therefore AI is not smart.
Jul 8, 2026, 5:33 PM@emilymbender
As the old joke goes:
- If you're broke, you do statistics.
- If you have a research grant, you do machine learning.
- If you're a venture capitalist, you do #AI.
And given how the term “AI” has inflated like a third-world currency – what we have now does not merit the term, it's just “generative models” – one may add:
- If you're a Silicon Valley startup, you do #AGI.
Jul 8, 2026, 5:40 PM@emilymbender@dair-community.social It really is the blockchain of software buzzwords.
@emilymbender Thank you for continuing to clear up our understanding of the expression "AI", and esp. in important reference sources like OUP. I find I often have to reach for very authoritative sources to back up perspectives that goes against the mainstream infatuation with chatbots.
when people ask me what AI means, I say
A senior PhD with some programming skills said: thanks to AI, I was able to do in a few hours what normally takes a week
or
Good programmer: I'm not that strong in python, so when I need a python "snippet" I find AI saves me a lot of time
or
Senior and IMO very good engineer overseeing a staff of 25 people:
If you don't have AI skills, or are unwilling to learn, we are not gonna hire you - and that is not next year, it is today
or
65 yro female neighbor, college educated: I use AI all the time, it is so helpful
shrugs
@emilymbender I find myself wondering, can we create AI that enhances human labour, that strengthens accountability, that decentralises power? Are @DAIR or anyone else you know working on that? And if so, how can I get involved?
Jul 8, 2026, 9:10 PM@PeteBleackley
When you read the paper and understand what “AI” is and does, you will also understand that the answer is no.
@Moss From the point of view of a data scientist who would like to work on something worthwhile, that's a very pessimistic answer. I don't want my field to be an ethical dead end!
Jul 8, 2026, 9:56 PM@PeteBleackley the thing is, I was not being the least bit pessimistic. My statement was based on Bender, Gebru, et. al.’s many reports—not on my own feelings, reactions, or opinions.
I can certainly appreciate what you say about not wanting your field to be an ethical dead end. I do honestly have a lot of pessimism about that. Perhaps look to the work of Adrianna Tan, who is doing her best to steer the industry towards less harm.
@emilymbender
AI is and always will be the villagers in Age of Empires
Jul 8, 2026, 6:13 PM@emilymbender "Shift accountability" is always the first thing any psychopath notices about science.
@emilymbender Just started reading the preprint, I had absolutely no idea how catch-all the term AI has always been, nor its origin.
Thanks for sharing here!
@emilymbender A fascinating paper, which I need to read again more slowly to fully appreciate its theses.
How can I respond to your article without appearing to "mansplain"? (FWIW, I've been adjacent to the "AI" R&D field for decades, and have worked on several attempts to product-ize research prototypes, in machine learning, natural languages processing, deductive databases; but I'm far from being an expert in any of these (although I've worked with experts), let alone in something as general as "AI", and my non-expertise will no doubt be indistinguishable from "mansplaining".)
That is, if you wish to have a conversation with a semi-expert.
@emilymbender@dair-community.social Mind sending me a pdf? My university's shib is misbehaving.
@emilymbender what if you run an open source AI locally on your machine? Then it wouldn’t centralise power. Right?
Jul 8, 2026, 10:39 PM
@emilymbender Read your preprint. Great work - I definitely need to go through some of the papers you cited.
(You would really think the word "Contents" would convey to people that the words in the image are not the entirety of what you wrote, but alas)
@emilymbender REAL AI (artificial intelligence) is the ultimate goal of machine learning. What is marketed as AI is Automated Incompetence based upon very smart and valuable language prediction modeling using stolen data. It is rather important to differentiate between the two. Your writing is a excellent description of what is currently being marketed.
@wbpeckham Something tells me you didn't actually read the article.
@emilymbender If it requires clicking on a link, and if I didn't request it, I never do.
@wbpeckham Well then, you are awfully confident in replying to something you didn't read. (And yes, you look like a mansplaining fool.)
@emilymbender And yet, what I did see of your writing I found compelling and worth complimenting. I guess that makes me worth criticizing.
@wbpeckham for this interaction, yes
Professor Bender,
I read your prepublication article with great interest. One thought it stirred for me is that the harms of the “AI” frame may be strengthened by the older “arms race” frame.
I was at Harvey Mudd when Edward Teller visited and argued, somewhat surprisingly, that excessive secrecy had damaged American science. His view was that classified work drove away some of the best researchers, slowed open correction, and still failed to prevent leakage.
That seems relevant to AI now. “Arms race” language does not merely describe policy. It changes who participates, what work becomes respectable, where accountability goes, and which institutions accumulate power.
My own experience across early Internet development, classified systems work, venture-funded technology, and institutional exclusion has made me increasingly suspicious of any technical framing that hides labor, concentrates authority, or prevents open critique.
Your article gave me a sharper vocabulary for that concern.
Terri Gilbert
@emilymbender This is going to be a really good reference to use for a presentation I will make next week for arguments against AI in game development, and this breakdown of the word itself in this paper will be super helpful to get a good footing and to be very specific about wlhat "AI" we are talking about in different contexts.
Thanks a lot for sharing! (Btw, can I refer to this paper and cite you in the presentation for quotes and paraphrasing?)
@glitchypixel yes please do so!
We going to wind up slow walking whether we come to it through wisdom or folly. Ethics is not a grass root, it is a cornerstone and they left it out on purpose.
I'm not sure enough of what passes for leadership remains to fight for the future. They all seem rather infected with a lethal combo of nihilism and fatalism.
I no longer expect the courts to help, sadly.
Does your paper present a solve? Is there a pre-print?
@emilymbender Wish we could find a way to standing for generational harm and class action. What's happening in education right now certainly qualifies and it seems no one is listening to those teachers.
#Retrolanguage now confirmed by a report out of The Guardian, too. Still looking for first reports of linguistic competency loss (not cognitive offloading, actual flattening of thought>word translation) but likely no one paying for that let alone double blind.
History will not be kind, IMO.🪔
@emilymbender thanks for sharing your preprint. Weizenbaum's ELIZA and the participants, "very human tendency to attribute understanding" makes me think how kind most humans are. We tend toward accepting speech as genuine which is unfortunately a tendency being abused by modern LLMs.
@tregeagle Emily's book contains a section on ELIZA.
I don't think it's kindness any more than seeing faces, animals and objects in clouds (or seeing faces in anything, pareidolia is just how our brains work) is