I've been thinking about the new Associated Press guidelines to avoid referring to #AI in ways that could imply humanness, sentience, or intent:
Don't say, "It WANTS you to enter more information," for example.
I've often used that kind of wording for computers in the past.
But more precise wording matters now because it's the first time we've widely had systems that could be mistaken for being human or having sentience, and it's important not to reinforce that idea.
@Frank I don't use it for writing anything like an article or summary, but I use it for other things. I'm most impressed with it for cleaning up transcriptions, but you need to do a final "compare documents" check to make sure it hasn't gone off the rails. I'll outline my whole process in a future newsletter, but it saves me hours over how I used to do it.
I also use it when I can't remember the name of something. I'll describe it, and usually get the name (although I always confirm it).
@Frank It's very hard for me to imagine using it for getting news knowing how often it gives incorrect information (although I usually trust librarians about such things).
(this is me)
What is the preferred spelling of deserrt, please?
(ChatGPT: I like the "two 's's' ")
The preferred spelling for the term you're looking for is "dessert." "Desert" with one 's' typically refers to a dry, arid, or barren land, while "dessert" with two 's's' refers to a sweet course typically served at the end of a meal.
@Ralph@Frank This is why we have to be careful with our language. They sounds so incredibly human.
I was marveling yesterday at a company's chatbot that used language that made me empathize with it and feel like it cared even though intellectually I knew it didn't. ("Thanks for hanging in there. Let me look in one more place." Etc.)
@grammargirl So we should say that the computer program has an open entry line? That seems cumbersome. The “program is asking for input” is probably neutral enough.
@EricFielding I agree. "Ask" doesn't feel as human-like as "want" to me.
I'm not trying to put out dogmatic rules. I just believe it's important to think about the words we're using and what they imply or lead others to believe about how these systems work.
And of course, as a proponent of AP style, I want to help people understand AP's guidance is, which is not to write as though AI has thoughts of feelings and not to use human pronouns such as "he" and "she."
I'm not buying into a game where others can use the word intelligence with all that implies to sell their snake-oil, but I'm not allowed to criticize it for doing the bad things intelligent things can do like lie, or cheat. If it walks like a duck and quacks like a duck, you're justified in calling it a duck and treating it accordingly.
When a supposed AI glibly gives me a made-up answer because the real answer isn't in its corpus, as opposed to "I don't know" - that's a lie, kids.
@Biggles That's an interesting point. To me, using words that make it seem human, like "lie" and "hallucinate," ascribe more power to it than it has and contributes to the marketing that causes some people to mistakenly believe it is sentient.
But words have all kinds of power and I can see why you think "It gives wrong answers," or "You absolutely can't rely on it for facts," aren't as strong as "It lies."
The proponents already ascribe more power to AI than it deserves, unfortunately - with real world effects. People are losing their jobs because managers believe it can do things it can't.
"It's just a pattern-matching engine that emits plausible responses without any understanding of the meaning of the content" - while more accurate - doesn't communicate the point nearly as well as "it lies".
@grammargirl Given how limited our understanding of and ability to quantify sentience or consciousness is, I do not like the idea that the more something appears to posses those traits the more forcefully we must deny the possibility...
I don't think the current generation of "AI" is very close to anything we would consider sentient, but if it does one day reach that point there's already going to massive resistance to recognizing that fact...further institutionalizing the idea that "of course computers can't be sentient" seems like a bad call.
@admin I do find it interesting that the AP Stylebook has a "right now" qualifier: "Right now, AI systems can emulate only aspects of human work and cognition, but are not sentient per se."
@grammargirl As a person from the southeastern U.S., this will be a hard habit to break. I grew up saying things like "The car doesn't wanna start this morning" and "This jar doesn't wanna open." Time to retire those expressions!
@grammargirl
This intentional speech thing is a really old and entrenched problem that is a source of a lot of confusion and poor judgement. If you look at how the media, and even many academics speak about evolution, you will see it all over the place. They talk about a species "trying to adapt", or evolving a resistance "in order to escape vaccines" or "defeat antibiotics", etc. Like no, for fuck sake, they are microbes, and evolution doesn't "want" to survive, or escape, or defeat anything
@Susan_Larson_TN@grammargirl i have a very fraught relationship with siri. mostly i just shout "goddammit siri..." when "she" does the wrong thing or misunderstands me... although i usually say "thank you siri" when i get the answer i'm looking for.
@grammargirl — Well, this does get us into murky waters about humanness, sentience and intent, does it not?
Is it wrong to say “Uncle Sam wants you,” as the famous recruiting posters did and do, because Uncle Sam lacks those attributes?
Is it wrong to say “the plants want to be watered”? How about (begging Descartes’s pardon) “the cats want to be fed”?
If we allow the general rule that Uncle Sam, plants and cats can be grammatically treated as possessing such attributes, despite not being H. sapiens, then on what grounds, apart from the grinding of political axes, do we deny AI admission to the club? Theology?
How far does our right to denial of seeming humanness and sentience on political or theological grounds extend?
Hugo-award-winning science fiction author Connie Willis wrote a wonderful science fiction novella about this a few years back: “All About Emily”. An utterly convincing AI in a young woman’s form has her/its heart set on becoming a Rockette, to such a degree that she/it weeps, visibly heartbroken, upon being refused. The Rockettes are so much won to her side that they dance in her honor on the street. Where should you and I stand on this?
@SeumanOtwal Well said. These are deeper thoughts into the kind of things I was considering when I was thinking about how I have talked about intent in the past.
I know people argue about animals an intent and where you draw the line. Most people believe dogs want things, but jellyfish don't have a centralized brain. Do they have desire, or are they just acting on instinct? Should instinct count as desire in the way we usually define it?
@grammargirl — “All About Emily” was one of Connie’s annual Christmas stories, so unless you are allergic to secular Christmas rites, this is an especially enjoyable time to be reading it, including aloud to older children.
@SeumanOtwal@grammargirl It’s an excellent point, but I think the AP’s goal is the desire to avoid a very specific confusion.
We don’t have people imputing to cats and army recruitment posters and plants as sentient (not more than usual, anyhow), but the risk of this kind of confusion about computers is on the rise. And there are consequences to that—if my “self-driving” car runs someone over, it didn’t “decide” anything or “make a mistake.” I did (perhaps abetted by Elon Musk, but still).
@maxleibman@grammargirl — Hi, Max. That sort of concern about “this kind of confusion” is what I thinking of in speaking of “political axes” and “political grounds”. I apologize if I failed to express myself clearly! I myself think the rules of grammar ought to stand on firmer and more changeless ground than passing political concerns.
Add comment