3540 🏆 CHIVET
Paul Keating
Tim Bourne’s shim under an unstable chair leg (#6) took first place with 5 votes. There were three runners-up with 4 points: Mike Shefler, Judy Madnick, and new player Rey.
But Tim is on the list of dealer opt-outs at dixonary.net, and so the deal goes to Mike.
Definition #13
Diminutive of chive
is from OED3, in a definition written in 1889.
I expect non-botanists will have interpreted chive in the definition as ‘the smallest cultivated species of Allium’. I certainly did. And it is hard to see why that word would need a diminutive, which is why I chose it.
But the OED’s intended sense is clove, as in garlic, for which the general term, I learn, is bulbil (insert hobbit joke of choice here).
But we are not all non-botanists. It got a vote from Johnny (of course), from Tim Lodge, and from new player Rey.
The word chivet comes originally from John Kersey’s Dictionarium Anglo-Britannicum (1708–21), an esteemed lexicographer who was one of Samuel Johnson’s now little-remembered predecessors:
Chivets the small Parts of the Roots of Plants, by which they are propagated.
And this was repeated by Nathan Bailey, another of Johnson’s predecessors, in his popular Universal Etymological English Dictionary (1721–1802). The OED1 traced Bailey’s definition first back to Kersey, and then back to Edward Phillips (New World of English Words, 1658; 1720 edition revised by Kersey) and Elisha Coles (English Dictionary, 1676), who both give it as the meaning of chives, not chivets: so Kersey’s chivets looks to be an error. As the OED1 pointed out.
In short, the word chivet never really existed at all. It was probably originally an error on Kersey’s part. Bailey conferred an imprimatur on it, over 30 editions; Johnson did not. Then Murray took Bailey’s entry, and gave it a definition that was different from Kersey’s. He did acknowledge the probability of transmission error, but only in the small print.
Now, the part of this post that I have been dreading.
Rule 12. When a round is in progress, if any situation arises which is not disposed of by these Rules or established precedent, the dealer has authority to resolve the question in such manner as seems to him or her equitable.
As some of you will have seen from Dave Cunningham’s inexplicit post, he submitted to me two definitions, one of his own devising, and one sourced from an AI called Grok, and asked me to list both for players to choose in this round.
I refused.
About AI-generated fakes
The rules specify only that a submission be fictitious, and that it not resemble the real definition, supposing the player knows what that is (Rule 2). They do not say anything about how it is composed. You can read in the comments to that rule about how broadly it has been interpreted in the past: that included using text generators (Round 208, 2000) and LLMs (Round 3510). And, apart from Dave’s submission, in this round there were at least two others that in my opinion bore the teeth marks of an AI.
I believe it is not for the dealer to permit or not permit an AI-assisted submission, because in general it would be impossible to tell. Though there is no harm in persuading a player to fix a hallucination that they missed.
About this ruling
My objection to Dave’s two fakes was not that one was AI-generated, but that he wanted to submit two definitions in one round.
I suggested that he withdraw his own fake and submit Grok’s in its place. That he was unwilling to do.
I suggested as an alternative that he acquire group membership under another of his 3 email addresses, in that way becoming in effect two players, and submit Grok’s fake from that address. Casinos permit a client to play two hands at the same blackjack table, and there is nothing to stop a Dixonary player from doing something analogous, and also no way to tell. Nor can it be considered in any way unfair to other players.
But he recoiled at the admin that would entail.
Instead he wanted me to post his two definitions, admit at the close of voting that one was AI-generated, and disallow any votes for the the AI-generated fake, simply on the grounds that it was AI-generated, and so was novel, and so somehow didn’t count. It was to be accorded a sort of honorary status, on account of its authorship.
I could not convince him that (a) his AI fake was not nearly the “first” that he seemed to think it was; and (b) even if it were, that would be a poor excuse for having a definition that players were invited to vote for, but that didn’t really count. Still, he was keen to submit it, and I did my best to suggest regular ways to do that.
There is nothing in the rules to say a player may submit two fakes, or how that should be scored. I’d be keen (and very surprised) to learn if any player ever tried it.
The nub of the matter is that, at least for now, an AI is purely a tool used by a real-world player, and that real-world player assumes responsibility for following its advice, just as they take credit when the advice is good.
It may not be long before we have AIs that are fully autonomous agents, that can not only dream up fake definitions, but also respond before deadlines, choose real words that don’t evoke a dozen DQs, and perform the largely mechanical task of dealing. But that’s not yet.
When it happens, that won’t bother me at all. I yearn for the days when we had, not 16, but 32 definitions to choose from.
Tim Bourne
In principle I'm happy to deal, but in practice I'm away this weekend without WiFi or phone. With more holidays ahead, I'll stay on that list until the end of August, please.
Best wishes, Tim.