If GPT systems capture some sense of "usual" context based on trillions of internet tokens, and that corpus is regarded approximately "global context", then it seems not so objectionable to call "unusual", new training items that contribute to fine-tuning loss.
It seems reasonable to worry that ubiquitous GPT systems reduce social entropy by encouraging copying instead of new thinking, but it could also have the reverse effect: If I am immediately aware that an idea is not novel, I may avoid attractors that agents that wrongly believe they are "independent" will gravitate toward. -----Original Message----- From: Friam <friam-boun...@redfish.com> On Behalf Of glen Sent: Thursday, March 21, 2024 7:49 AM To: friam@redfish.com Subject: Re: [FRIAM] the inequities of uniquity A friend of mine constantly reminds me that language is dynamic, not fixed in stone from a billion years ago. So, if you find others consistently using a term in a way that you think is wrong, then *you* are wrong in what you think. The older I get, the more difficult it gets. But specifically, the technical sense of "unique" is vanishingly rare ... so rare as to be merely an ideal, unverifiable, nowhere, non-existent. So if the "unique" is imaginary, unreal, and doesn't exist, why not co-opt it for a more useful, banal purpose? Nothing is actually unique. So we'll use the token "unique" to mean (relatively) rare. And "unusual" is even worse. Both tokens require one to describe the context, domain, or universe within which the discussion is happening. If you don't define your context, then the "definitions" you provide for the components of that context are not even wrong; they're nonsense. "Unusual" implies a usual. And a usual implies a perspective ... a mechanism of action for your sampling technique. So "unusual" presents even more of a linguistic *burden* than "unique". On 3/20/24 13:14, Frank Wimberly wrote: > What's wrong with "unusual"? It avoids the problem. > > > On Wed, Mar 20, 2024, 1:55 PM Steve Smith <sasm...@swcp.com > <mailto:sasm...@swcp.com>> wrote: >> >> I'm hung up on the usage of qualified "uniqueness" as well, but in >> perhaps the opposite sense. >> >> I agree with the premise that "unique" in it's purest, simplest form >> does seem to be inherently singular. On the other hand, this mal(icious) >> propensity of qualifying uniqueness (uniqueish?) is so common, that I have >> to believe there is a concept there which people who use those terms are >> reaching for. They are not wrong to reach for it, just annoying in the >> label they choose? >> >> I had a round with GPT4 trying to discuss this, not because I think LLMs >> are the authority on *anything* but rather because the discussions I have >> with them can help me brainstorm my way around ideas with the LLM nominally >> representing "what a lot of people say" (if not think). Careful prompting >> seems to be able to help narrow down *all people* (in the training data) to >> different/interesting subsets of *lots of people* with certain >> characteristics. >> >> GPT4 definitely wanted to allow for a wide range of gradated, speciated, >> spectral uses of "unique" and gave me plenty of commonly used examples which >> validates my position that "for something so obviously/technically >> incorrect, it sure is used a lot!" >> >> We discussed uniqueness in the context of evolutionary biology and >> cladistics and homology and homoplasy. We discussed it in terms of cluster >> analysis. We discussed the distinction between objective and subjective, >> absolute and relative. >> >> The closest thing to a conclusion I have at the moment is: >> >> 1. Most people do and will continue to treat "uniqueness" as a >> relative/spectral/subjective qualifier. >> 2. Many people like Frank and myself (half the time) will have an >> allergic reaction to this usage. >> 3. The common (mis)usage might be attributable to conflating "unique" >> with "distinct"? -- ꙮ Mɥǝu ǝlǝdɥɐuʇs ɟᴉƃɥʇ' ʇɥǝ ƃɹɐss snɟɟǝɹs˙ ꙮ -. --- - / ...- .- .-.. .. -.. / -- --- .-. ... . / -.-. --- -.. . FRIAM Applied Complexity Group listserv Fridays 9a-12p Friday St. Johns Cafe / Thursdays 9a-12p Zoom https://bit.ly/virtualfriam to (un)subscribe http://redfish.com/mailman/listinfo/friam_redfish.com FRIAM-COMIC http://friam-comic.blogspot.com/ archives: 5/2017 thru present https://redfish.com/pipermail/friam_redfish.com/ 1/2003 thru 6/2021 http://friam.383.s1.nabble.com/
smime.p7s
Description: S/MIME cryptographic signature
-. --- - / ...- .- .-.. .. -.. / -- --- .-. ... . / -.-. --- -.. . FRIAM Applied Complexity Group listserv Fridays 9a-12p Friday St. Johns Cafe / Thursdays 9a-12p Zoom https://bit.ly/virtualfriam to (un)subscribe http://redfish.com/mailman/listinfo/friam_redfish.com FRIAM-COMIC http://friam-comic.blogspot.com/ archives: 5/2017 thru present https://redfish.com/pipermail/friam_redfish.com/ 1/2003 thru 6/2021 http://friam.383.s1.nabble.com/
