Are you annoyed with the anthropomorphizing language being used in the "AI" discourse, but not sure how to talk about this stuff without it?

[stevefoerster@social.vivaldi.net]

@emilymbender Don't anthropomorphize LLMs. They don't like it!

[rrb@infosec.exchange]

@emilymbender Maybe instead of using the term "intelligence" a more appropriate word, like "fecal matter" should be used?

[mimesatwork@wandering.shop]

@martinvermeer @Petesmom @emilymbender I don't think 'corpus-based generation' is a feasible alternative. A good alternative is as easily understandable and pronounceable as possible, otherwise it's not going to be adopted into wider use, and most people will associate "corpus" with any biological body, because their only connection with the word is from corpus Christi, as in body of Christ

[thisalex@hachyderm.io]

@emilymbender the topic is interesting, but the choice of words is really complicated. A lot of different things are packed there and it’s way too easy either to oversimplify (vulgarise) or overgeneralise. I will be thinking about this one.

[martyfouts@mastodon.online]

@iamnotU @emilymbender People have been doing this since the first welding robot was installed at GM in 1959.

[rcosta@archaeo.social]

@JamesWidman @emilymbender We need to face LLM communication as a new category of high level programming language. We can argue that writing something like: print "$name", is also 'anthropomorphizing language'. We get annoyed. Pioneers that communicated with machines using physical switches and machine code also did, after witness kids using BASIC and keyboards. The best advice to interact with LLM I read: Talk like a caveman : "do this", "add that", etc.. Save tokens, energy and time.

[iamnotu@mastodon.social]

@MartyFouts @emilymbender Wally Welder

[megamichelle@a2mi.social]

@emilymbender

I don't usually have to describe the output of LLMs, but when I do, I like to say "it emitted some text which seemed to say".

But even "seemed to say" is too anthropomorphizing.

[megamichelle@a2mi.social]

@emilymbender

One time, I even tried "it emitted some text which, if it had been written by a human, would say..."

But maybe I could go with "it emitted some text which, if it had been written with intention, would say..."

[fishidwardrobe@mastodon.me.uk]

@emilymbender for my own amusement i've for a while now here been calling it "Al". no–one seems to have noticed, but obviously this is not helping matters

[alsweigart@mastodon.social]

@fishidwardrobe Believe me, I have noticed.

[fishidwardrobe@mastodon.me.uk]

@AlSweigart IoI

[haskellelephant@bitbang.social]

@emilymbender Good suggestions, but I think some of those names reveal a bias

[nilz@norden.social]

@emilymbender

hallucination → undesirable output

This is the way

[emilymbender@dair-community.social]

@MegaMichelle I think the long form is good: It emitted (I use "extruded") some text that the reader was able to interpret as....

[megamichelle@a2mi.social]

@emilymbender

Oooh, I like the focus on the reader's interpretation. It doesn't imply that the meaning is actually in the text.

For this reason, I might avoid "able to interpret" because that changes it back to implying that the meaning was actually in the text. Maybe "It extruded some text that a reader might interpret as saying..."

[clickhere@mastodon.ie]

@martinvermeer I'll take it!

@Mimesatwork @Petesmom @emilymbender

[ezeno@mastodon.uno]

@emilymbender
Slop
Smile-face Shoggoth
Brainrotter
Faker machine
Bullshitnator

[martinvermeer@fediscience.org]

@Mimesatwork @Petesmom @emilymbender I don't entirely agree. Corpus/corpora is widely used in science to describe bodies of text etc. put together as a basis for research. Adopting it for this use - which is already a reality among the professionals in this field - does not seem like a long shot to me.

https://en.wikipedia.org/wiki/Corpus

And BTW if you object to 'corpus', you should also have a problem with 'generation', which for lay people will refer either to successive generations of people, or to generating electricity. 'Generating' text or imagery or other output is fairly new and specialized and not yet common in the general language.

[gatesvp@mstdn.ca]

@emilymbender

Zero pronouns, just "it". I call it the Robot or the text Robot. It doesn't have a human name. I also frequently refer to it as "the BS machine".

I've definitely been using "input" terminology for a long time as well. Though I also like "input prompt", we've had a "command prompt" in computing for a long time, it's never really felt like an anthro term to me. And "input prompt" is a subtle shift on existing terminology.

I really like "unexpected results" over hallucinations... Though I am still very hesitant to acknowledge the term hallucination at all. Even the notion of errors seems kind of wrong. The output may include things that you didn't want in it, but in no way does that mean the system failed.

I think that's the weirdest thing about these Gen AI systems. Not the anthropomorphic language trying to humanize the machines, but the functional language falling to represent what they do.

I use these tools and still hate the language.

I will be a stick in the mud.