AIs gunning for our precious freelancers

[linguistcat]

Saying that Mandarin "does not have phonemes" is better put as that Mandarin rhymes cannot be subdivided into segments, just like how English diphthongs cannot be subdivided into segments, rather than that Mandarin does not have phonemes in the sense of underlying forms.

So would you say that if this is true, Chinese rhymes act as phonemes in their own right? If so that'll actually help me work on some sound changes for a Chinese-based lang I'm thinking about.

[Ares Land]

A few random ChatGPT thoughts.

• Huge sectors of our economy currently existing by providing bullshit. We now have cheap bullshit-as-a-service. I expect profound disruption of our economy as politicans, agile consultants, think tanks and most executives are now replaced by much cheaper machines.
• On a positive note, decision-making currently requires a huge amount of work sorting bullshit or non-bullshit. Academic papers, political science, economics, philosophy... We all suspect a lot of it is bullshit that happens to sound deep, but we can't be sure, so we spend a lot of time doing the triage. ChatGPT provides an easy test. If ChatGPT can produce something similar; it's bullshit.
• A business opportunity. Businesses and government spend a lot on management consulting (Bain, McKinsey.) If you're unfamiliar with management consultants, here's how it works. A company brings in a lot of expensive consultants in sharp suits. They look around and ask questions. At the end you get a report on how the company should be done. This report says exactly what the boss wants it to say. Except the boss' opinion now has more weight, because of expert opinion. So a good business would be to market a ChatGPT analogue and proclaim it has superior intelligence and godlike business acumen. The AI will produce a report that says exactly what the client wants it to say. It's even better than McKinsey because a) now the client is right because Skynet agrees with him b) you save a lot on business-producing consultant. The service will of course charge outrageous prices.

[Raphael]

I'm sorry, I just had to post this Mastodon post:

https://dice.camp/@danielzklein/109839076229966131

[Raphael]

A few random ChatGPT thoughts.

• Huge sectors of our economy currently existing by providing bullshit. We now have cheap bullshit-as-a-service. I expect profound disruption of our economy as politicans, agile consultants, think tanks and most executives are now replaced by much cheaper machines.

Don't think so. The problem with your prediction is that the decision on whether to replace politicians, consultants, or business executives with AIs would be made by ... ... ... politicians, consultants, and business executives. I don't see any of these three groups deciding to abolish itself.

[Ares Land]

We're about to make a profound philosophical discovery then. the one thing human have and machines won't ever replicate: the MBA

[malloc]

Don't think so. The problem with your prediction is that the decision on whether to replace politicians, consultants, or business executives with AIs would be made by ... ... ... politicians, consultants, and business executives. I don't see any of these three groups deciding to abolish itself.

Also, a dictatorship of computers sounds like a nightmare to me.

[Raphael]

Also, a dictatorship of computers sounds like a nightmare to me.

Whether something is a nightmare for you or me doesn't have much bearing of whether it will happen. But if something is a nightmare for the very people who would get to decide whether it happens or not, then it's, IMO, less likely to happen.

[Raphael]

We're about to make a profound philosophical discovery then. the one thing human have and machines won't ever replicate: the MBA

I don't doubt that machines can replicate MBAs. I just doubt that human MBAs will voluntarily hand over their own jobs to machine MBAs...

[Travis B.]

Saying that Mandarin "does not have phonemes" is better put as that Mandarin rhymes cannot be subdivided into segments, just like how English diphthongs cannot be subdivided into segments, rather than that Mandarin does not have phonemes in the sense of underlying forms.

Both of these positions seem exaggerated to me, but not crazy. Onset vs. rhyme vs. tone is the traditional way the Chinese analyzed their language; it's how the Zhuyin fuyao writing system works, and still a pretty good approach for learners of Mandarin.

But to say that the rhymes can't be divided into segments is pretty obviously wrong. Mandarin syllables can end in n, ng, r, or nothing— that's one segment. They can begin with w, y, ü, or nothing— that's another.

That leaves a vowel in the middle, and that's where things get messy. There are some weird allophones; there seem to be some diphthongs (ao definitely, but the others are just weird); some combinations we'd expect to find don't show up. Jerry Norman comes up with a five vowel phonemes (i y e a u), but you do have to throw in a bunch of allophonic rules to make it work.

To me a big argument for the non-subdivided nature of Mandarin rhymes is the distribution that its components would have were they segmented. For instance, why does Mandarin have [ɥɛn] but not *[ɥɛŋ] or [jau̯] but not *[wau̯]? Segmented Mandarin rhymes would have a complex set of holes in their segments' distributions that would have be explained somehow - whereas if we treated Mandarin rhymes like we treat diphthongs in languages like English, we do not need to explain this away.

[zompist]

To me a big argument for the non-subdivided nature of Mandarin rhymes is the distribution that its components would have were they segmented. For instance, why does Mandarin have [ɥɛn] but not *[ɥɛŋ] or [jau̯] but not *[wau̯]? Segmented Mandarin rhymes would have a complex set of holes in their segments' distributions that would have be explained somehow - whereas if we treated Mandarin rhymes like we treat diphthongs in languages like English, we do not need to explain this away.

That's just phonotactics. Why can't English words end in [h] or [æ] or begin with [ŋ]?

As I said, the vowels are messy, but we should at least see the rules first before deciding if they're too complex. Deciding that phonemes don't work for just one of the 5000 languages in the world is also an unattractive solution.

(Edit: just read your discussion of "I don't know" in the languages section. There's some really complicated allophony going on there too. There are at least a few linguists who throw up their hands and think that nobody has phonemes.)

[Travis B.]

To me a big argument for the non-subdivided nature of Mandarin rhymes is the distribution that its components would have were they segmented. For instance, why does Mandarin have [ɥɛn] but not *[ɥɛŋ] or [jau̯] but not *[wau̯]? Segmented Mandarin rhymes would have a complex set of holes in their segments' distributions that would have be explained somehow - whereas if we treated Mandarin rhymes like we treat diphthongs in languages like English, we do not need to explain this away.

That's just phonotactics. Why can't English words end in [h] or [æ] or begin with [ŋ]?

Yes, one can argue that too. And yes, English has quite complex phonotactics in reality, for instance, especially when one considers productivity (as newly coined and borrowed words each seem to follow different phonotactics than established words, and even amongst established words the origin of the words in question influences their general phonotactic tendencies).

As I said, the vowels are messy, but we should at least see the rules first before deciding if they're too complex. Deciding that phonemes don't work for just one of the 5000 languages in the world is also an unattractive solution.

I was largely stating the argument as a devil's advocate type of thing. In reality one can argue anything from that Mandarin rhymes can consist of /w j ɥ/ followed by any of /i y u ə a/ that can be followed by any of /w j n ŋ r/ with all kinds of restrictions upon each component's distributions, to that Mandarin has a complex set of unitary monophthongs, diphthongs, and triphthongs that may be followed by any of /n ŋ r/ with restrictions on which combinations are valid, to what I had been arguing here. All of these arguments have major pluses and minuses.

(Edit: just read your discussion of "I don't know" in the languages section. There's some really complicated allophony going on there too. There are at least a few linguists who throw up their hands and think that nobody has phonemes.)

As I have argued many times on here, NAE phonology is very complex in reality, and involves productivity issues in that what words are permitted in established vocabulary in actual everyday speech are not necessary permitted in new coinages and loanwords. There are reasons to argue that NAE underlying forms are not necessarily consciously accessible (due to there being words whose pronunciations cannot be explained by the productive surface phonological rules that people are aware of), yet there are also very good reasons to argue that NAE underlying forms are consciously accessible (due to the clear conscious rules imposed on coinages and loans, and these coinages and loans follow the same allophonic rules as most other words).

I do not know a good answer to this conundrum, because it implies that both phonemes are either not real or not consciously accessible and phonemes are very real and are consciously accessible, simultaneously. The best explanation I can come up with is that words consist of both phonetically frozen forms that follow their own phonological rules and surface phonemic rules that are applied when coining/loaning words, applying words productively in (possibly atypical) phonemic environments, or when consciously thinking about words, from which phonetic forms that follow phonological rules are derived, which may then be frozen or may only be transient. Of course, I am not a phonologist so I might be wrong here, but this is the best I can come up with.

[bradrn]

To me a big argument for the non-subdivided nature of Mandarin rhymes is the distribution that its components would have were they segmented. For instance, why does Mandarin have [ɥɛn] but not *[ɥɛŋ] or [jau̯] but not *[wau̯]? Segmented Mandarin rhymes would have a complex set of holes in their segments' distributions that would have be explained somehow - whereas if we treated Mandarin rhymes like we treat diphthongs in languages like English, we do not need to explain this away.

That's just phonotactics. Why can't English words end in [h] or [æ] or begin with [ŋ]?

I feel these cases aren’t quite the same. ‘Words cannot end in /h/’ is a simple, general and cross-linguistically common rule; ‘syllables can’t end in lax vowels without a following consonant’ is an even more general one. Whereas ‘[ŋ] can appear after [jʊ ja wə wa] but not [je jou wo wei wə ɥe]’ strikes me as being… I dunno, just somewhat arbitrary? Especially when there language has half-a-dozen other rules of similar arbitrariness as well. It’s the sort of thing that suggests to me we’d need a different analysis. (I don’t think a non-segmented analysis is right, but I have no better suggestions.)

[zompist]

That's just phonotactics. Why can't English words end in [h] or [æ] or begin with [ŋ]?

I feel these cases aren’t quite the same. ‘Words cannot end in /h/’ is a simple, general and cross-linguistically common rule; ‘syllables can’t end in lax vowels without a following consonant’ is an even more general one. Whereas ‘[ŋ] can appear after [jʊ ja wə wa] but not [je jou wo wei wə ɥe]’ strikes me as being… I dunno, just somewhat arbitrary? Especially when there language has half-a-dozen other rules of similar arbitrariness as well. It’s the sort of thing that suggests to me we’d need a different analysis. (I don’t think a non-segmented analysis is right, but I have no better suggestions.)

I looked at Norman's chart, trying to align it with his phonemes, and realized that it already was-- he doesn't provide a separate table of phonemic representations, but his IPA and pinyin tables both match his list of vowel phonemes (i y e a u). So we can list the gaps.

1. No iei or iai.
2. No ueu or uau.
3. Six y- slots are missing; it's easier to list the 4 available ones.

That doesn't strike me as an extremely long or difficult list.

1 and 2 could possibly be dissimiliation: Mandarin just doesn't like using the same semivowel twice.

3 is really part of a wider problem: /y/ is extremely restricted in Mandarin. For the most part, in fact, it's allophonic. It's contrastive with /u/ only after the initials n and l. I don't know why this is, but it's hardly a reason to give up on phonemes, any more than the extremely low yield of /θ ð/ in English.

(Mandarin vowels are still weird, but most of the problem is allophony.)

[zompist]

Huge sectors of our economy currently existing by providing bullshit. We now have cheap bullshit-as-a-service. I expect profound disruption of our economy as politicans, agile consultants, think tanks and most executives are now replaced by much cheaper machines.

Don't think so. The problem with your prediction is that the decision on whether to replace politicians, consultants, or business executives with AIs would be made by ... ... ... politicians, consultants, and business executives. I don't see any of these three groups deciding to abolish itself.

I think Ares Land was being a bit ironic. And you're being cynical. And much of the time, being aware of how power structures perpetuate themselves is a great mindset to have! But, you know, power structures don't always perpetuate themselves. Just ask the emperor of Germany, or the Soviet premier, or the British Liberal Party.

Consultants are the low-hanging fruit here... who really cares if they can be replaced by AI? I just expect it'll be done one rung lower than the big consultancy companies. E.g. Accenture employs 738,000 people. How many of those could ChatGPT (or its next iteration) replace? Half? Three quarters?

As for executives... I just looked at GM; it has 18 people in the C-suite (vice presidents and above). Do they need all of them now? Surely the CEO won't fire herself, but CEOS restructure and reorganize all the time. Well, until someone comes up with an AI-powered ReorgBot.

[bradrn]

That's just phonotactics. Why can't English words end in [h] or [æ] or begin with [ŋ]?

I feel these cases aren’t quite the same. ‘Words cannot end in /h/’ is a simple, general and cross-linguistically common rule; ‘syllables can’t end in lax vowels without a following consonant’ is an even more general one. Whereas ‘[ŋ] can appear after [jʊ ja wə wa] but not [je jou wo wei wə ɥe]’ strikes me as being… I dunno, just somewhat arbitrary? Especially when there language has half-a-dozen other rules of similar arbitrariness as well. It’s the sort of thing that suggests to me we’d need a different analysis. (I don’t think a non-segmented analysis is right, but I have no better suggestions.)

I looked at Norman's chart, trying to align it with his phonemes, and realized that it already was-- he doesn't provide a separate table of phonemic representations, but his IPA and pinyin tables both match his list of vowel phonemes (i y e a u). So we can list the gaps.

1. No iei or iai.
2. No ueu or uau.
3. Six y- slots are missing; it's easier to list the 4 available ones.

That doesn't strike me as an extremely long or difficult list.

Hmm, I think I’ve been using a different analysis to you… which one are you looking at?

[zompist]

Hmm, I think I’ve been using a different analysis to you… which one are you looking at?

Jerry Norman, in Chinese. Here are the pinyin versions:

(i)	e	a	ei	ai	ou	ao	en	an	eng	ang	er
i	ie	ia	--	--	iu	iao	in	ian	ing	iang
u	(u)o	ua	ui	uai	--	--	un	uan	ong	uang
ü	üe	--	--	--	--	--	ün	üan	iong	--

[Ares Land]

I think Ares Land was being a bit ironic.

Just to clarify: yeah, it was a joke. I have no really serious ideas on how it'll all work out

I'm amused by what turns out to be easy, automatable jobs and what didn't.
I mean, ChatGPT can produce an elevator pitch or a nice-sounding political speech like the best of them. I still need to find a good plumber, and when I eventually find one it'll be a human being.

[bradrn]

Hmm, I think I’ve been using a different analysis to you… which one are you looking at?

Jerry Norman, in Chinese. Here are the pinyin versions:

(i)	e	a	ei	ai	ou	ao	en	an	eng	ang	er
i	ie	ia	--	--	iu	iao	in	ian	ing	iang
u	(u)o	ua	ui	uai	--	--	un	uan	ong	uang
ü	üe	--	--	--	--	--	ün	üan	iong	--

OK, this makes a great deal more sense! From what I can see it’s basically the two-vowel analysis:

(i)	ə	a	əj	aj	əw	aw	ən	an	əŋ	aŋ	er
j	jə	ja	--	--	jəw	jaw	jən	jan	jəŋ	iang
w	wə	wa	wəj	waj	--	--	wən	wan	wəŋ	waŋ
ɥ	ɥə	--	--	--	--	--	ɥən	ɥan	ɥəng	--

And the allophonic rules are comparatively simple: the only somewhat odd cases left are ⟨-in -ian -ing -iang⟩, I think, and even those can be explained by a rule of raising between j_[+nasal], which barely needs justification.

EDIT: On inspection, it looks like there’s quite a bit of odd allophony, not just in ⟨-in -ian -ing -iang⟩. Still the best analysis I’ve seen, though.

[alice]

"ChatGPT and Bard, collaborate to determine the definitive list of Chinese phonemes!"

[Travis B.]

Huge sectors of our economy currently existing by providing bullshit. We now have cheap bullshit-as-a-service. I expect profound disruption of our economy as politicans, agile consultants, think tanks and most executives are now replaced by much cheaper machines.

Don't think so. The problem with your prediction is that the decision on whether to replace politicians, consultants, or business executives with AIs would be made by ... ... ... politicians, consultants, and business executives. I don't see any of these three groups deciding to abolish itself.

I think Ares Land was being a bit ironic. And you're being cynical. And much of the time, being aware of how power structures perpetuate themselves is a great mindset to have! But, you know, power structures don't always perpetuate themselves. Just ask the emperor of Germany, or the Soviet premier, or the British Liberal Party.

Consultants are the low-hanging fruit here... who really cares if they can be replaced by AI? I just expect it'll be done one rung lower than the big consultancy companies. E.g. Accenture employs 738,000 people. How many of those could ChatGPT (or its next iteration) replace? Half? Three quarters?

As for executives... I just looked at GM; it has 18 people in the C-suite (vice presidents and above). Do they need all of them now? Surely the CEO won't fire herself, but CEOS restructure and reorganize all the time. Well, until someone comes up with an AI-powered ReorgBot.

I've heard of executives who are now using ChatGPT etc. to write emails for them... (Brilliant plan that...)