bohadi

Maddeningly accurate. Taarof can demonstrate a mastery of social intelligence, like the Japanese tea ceremony writ large in society, and is a good goalpost for AGI.

Not one month ago, as an admission of my own clumsy slight, I had remarked to a senior of higher status, that Chatgpt could just grasp Taarof.

tried this irl and wound up here

as a longtime, proud, and noble pirate; attempts at moral justification or condemnation are simply gratuitous and i'm certain many like me feel the same

this is a rather funny joke

but if it is not that would be extremely impressive! determinism/freewill reduces to shazam!?

whats the training data to predict new song titles? heh

check out this reply from claude2:

predict the next 3 new song titles from artist Taylor Swift

1. Last Dance with You - A reflective ballad about finding closure after a breakup. 2. Never Getting Back Together - A pop tune emphasizing that the same mistakes won't be made twice in a relationship. 3. 22 Was My Prime - A lighthearted look back on her early 20s as carefree years that can't be replicated.

...

very regretable and very predictable

pretty cool. "period" finetuning could be useful for game npc sims

I think we will find more examples countering our modern idea of discrete organisms separated across generations.

Consider again the glider translated across space time. Is it the same glider after 1,2,10 generations in time?

I always found it curious the ancient archons had rulerships which kept continuity in nominative identity across individuals for several generations. The social practice continues with the first, the second, III, IV etc.

The epigenetic view supports it.

ye prestige award is a political football now. same as it ever was?

indeed the prediction and agenda go hand-in-hand

it could not be otherwise else the agenda would alter the prediction until the new prediction and the new agenda match

interestingly this operation is the same as an agentic ai

I was interested in historical demographic predictions. What did the UN project for 2020 in 1950, 1960, 1970 etc.

Unfortunately pre-1950 demographic prediction is a little harder to find. I wonder why*

Anyway a reliable source is found here: https://ourworldindata.org/population-projections

In summary historical population projections have been pretty accurate but not in the way you expect.

*One thought is that because current date may become a historical inflection point similar to the first half of the 20th century, current date demographic statistics may be too important to put much stock in. Sobering.

I'm very interested to see how this picture will look 20-30 years out. Wealth is concentrated in pockets and the average person gets by. China wealth Gini index is moderating somewhat around Japan tier. US Gini is very high and closer to the world average.

What's really screwy is the government has replaced most of the English translation on street signs from a literal translation to phonetic translation of the mandarin sound.

I don't think this is a fair take for a couple of reasons.

Mandarin is unlike other languages in that it's written form is famously ideographic. Phonetic translation is all that is done in any pair {L, M}. I assure you there is no such thing as a literal translation in any language.

Secondly, english as a global lingua franca is not a given and we in the anglosphere ought to be gracious in it's modern historical role, lest it decline (this is an exact mirror of dollar reserve privilege). Your statement reads as "be more accommodating for me". But through the prism of good manners it smacks of liberal entitlement.

There is 0 reason to learn

hoo boy, i don't get paid enough for this... carry on

ah yes, SV the rightwing stronghold

whats the practical influence of the libertarian in such environment? rules-for-thee; and looting

Don't know where you're from or your experience but you maybe want to update your priors on the meaning term first world, which peaked in usage roughly 1948...

https://books.google.com/ngrams/graph?content=first+world&ye...

The mean and median living some places might surprise you. For example visit HongKong, Singapore, Taipei, Tokyo, or a mainland city this millenium. Or dont'!

First world, it mostly conjures automobile culture and car infrastructure which is nothing special these days but check the world's largest car makers... you would guess Germany Japan or even the USA recently no?

For sure political innovation is a terribly slow process. I don't believe leveraging LLMs will have a direct observable effect on that until civilization <> AI digests for little while longer. Yes we could certainly discuss AI governance (governance by AI)

There are two other areas you would seem to point at. Firstly other technological innovation will have momentous effects, I will not name the buzzwords but you easily know from what by excluding AI/LLM. The second and far more interesting is what I will call political innovation which is orthogonal to technological development, which is totally driven by changes in consciousness. I think that often looks like a stepfunction, because I hardly see the needle moving there outside the generation-to-generation cultural noise.

where do you see the onshore forex market moving? merits investigation though

local and provicial governments really dont matter. they do their best to support and bid projects for national champions and siting direct foreign investments where competitive. the areas which outperform (do least bad) have their executive talent siphoned to the central government. as FDI has dried up in the global slowdown restructuring and local reforms are due. one idea is to heavily lean on consumption which is palatable to some. i dont see that the chinese (or the japanese) feel that is a tenable path however

history is over!

The 200$ billion hole in CN real estate is comparable in size to the SVB signature US banking crisis. In that case US treasury policy was a bailout for uninsured depositors who are wealthy supporters of the ruling blue party.

Most of the underwater property developers are wealthy patrons of the ruling red party in China and will ultimately receive similar government bailouts. Privatized gains and socialized losses it's the same movie everywhere.

But this is just theatre, on the order of 1 trillion disappearing in smoke is not going to sink the ship of any nation with 100 trillion in wealth assets.

edit: I should say this money doesn't really disappear in smoke, it's bourne on the backs of mostly middle classes amortization over the period of decades in hidden inflation taxes and marginal lower living standards.

It is amusing so many commentators cannot communicate novel materials science in Korean to the layperson. Where's your zero resistivity sample?

The paper is just a guide. Who would replicate likely needs to contact the authors to understand the implicit details of the recipe. That is process knowledge. There's a good reason world-leading lithography happens in that region. With luck, researchers and engineers will be able to refine and industrialize the process these coming years.

Then it is good that Singapore makes laws in Singapore, and France in France, and not the other way around, isn't it.

Even more remarkable is that by the estimate in Table 1, the percentage is increasing over time.

If this trend continues over the centuries it is conceivable that >=50% of all human beings to ever exist could be extantly alive. Either growth rate exploded or death rate arrested.

It recalls to mind Fyodor the one-time teacher of Tsiolkovsky of the rocket equation, who conceived a kind of akashic immortality project he called the Common Task. So like maybe this Kaneda-Haub statistic asymptotically approaches 100%.

Supposing a starting population of 2 we can call Adam and Eve divided by 0 dead ancestors do note it started at infinity before dropping precipitously near 0 for a 202k year exponential climb to today's ~7%. Fwiw I consider such plot affirmative evidence assuming the demographic inputs are not wildly inaccurate.

apropos the roomba founder, a nontechnical argument for necessity of embodied AI circa 1980 (5 min video) https://youtu.be/QMMw9fQ452c?t=49

That LLMs learn a world model is very convincing now, but as LeCun has said it's just one piece of the intelligence puzzle, incl. perceiving, actuating shenmede

Here's some wild speculation grounded on a wisp of research:

Eurasia based cartel had cornered the market on lithium vital for new energy. US based interests bent the ear of the Chilean government to break this monopoly by nationalizing these privately owned mines and rediverting the supply. Speculators got wrecked. Does anyone know if a Credit Suisse customer had dealings in Chile?

I've been looking into this. The great thing is riscv isa is free to license and customize. There's opportunity for efficient inference at the edge. If anyone is interested in working on this together, send me a private message.

No, yeah. Suppose I am arguing for a light touch. Assuming one were at the levers of power (of the state apparatus, of the corporation, etc). That humility is lacking, for even the very wise cannot see all ends. It is a conservative point of view, what's the rush anyway. The next generation may have better data.

Depending on priorities, it could be optimal

what possible priorities... sounds like the humanity extincting AGI has already gotten free /s

If civilization needs to adopt such measures it must be in deep crisis.

Yes and, it occurs to me that AI political rights will probably first be granted when AI is only properly aligned with the ruling powers that be. When the AI sufficiently toes the party line it will given some degree of protection of existence, speech, etc. A multiplicity of LLMs, from WokeGPT to BasedGPT, are possible. With current methods, the AI's worldview (as it were) comes down to the training and human feedback finetuning. Exactly in this sense, alignment research is equivalent to AI politics. Not ethics.

But I would disagree with the idea that the whole reason of AI tech is to essentially have uncomplaining servants. The IP use is a whole 'other issue.

Late to the thread but to make a comparison here are American[1] and Chinese[2] "extradition treaty blocs", 150 ~ 50 countries.

[1] https://en.wikipedia.org/wiki/List_of_United_States_extradit...

[2] https://en.wikipedia.org/wiki/Extradition_law_in_China#/medi...

Interestingly if you are keen to slip the grasp of law enforcement (note: this is no sanction of illegal activity of any sort), and you reside in, for example, Australia Turkiye or Mexico, you are within the extradition jurisdiction to either China or USA, a kind of extremal legal coverage/exposure. The caveat being, there is a concept called dual criminality which is essentially the intersection of crimes from the one country and the extraditing country.

Extra credit: 1997 congressional report on the unprecendented (at the time) US-Hong Kong extradition treaty. [3] https://www.congress.gov/congressional-report/105th-congress...

My solution was similar, represent.

  First we observe x is between 20^3=8e3 and 30^3=27e3.
  Then note that the least significant (singles) digit of x^3 is solely determined by the singles digit of x [only holds in the integers].
  Luckily we had cached that the pattern for cubes is 123456789 -> 187456329
  Finally x^3 = 19,683 ends in 3 therefore x = 27 we are done.

  It turns out there is a unique mapping for the first digit of x^n, following one of four patterns:
  n mod 4 = 1  123456789 -> 123456789   
  n mod 4 = 2            -> 149656941
  n mod 4 = 3            -> 187456329
  n mod 4 = 0            -> 161656161

You may ask yourself, is there also a rule for the larger place digits of x^3, x^n? Yes, we naturally observe that the 10s digit of x^n is similarly wholly determined by the 10s and 1s digits of x. So now we want n maps on [0,9]x[0,9] to [0,9].

The python snippet may help to explore this: where the row 3, 25th digit shows that when x ends in 25, x cubed has 10s digit 2

  >>> print(*zip(range(50), [''.join([str(x**i).zfill(2)[-2:-1] for x in range(100)]) for i in range(50)]), sep='\n')
  (0, '0000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000')
  (1, '0000000000111111111122222222223333333333444444444455555555556666666666777777777788888888889999999999')
  (2, '0000123468024692582604827272840628529642086432100000001234680246925826048272728406285296420864321000')
  (3, '0002621412032947913506462278580963075571028082329405076719170824429630014127735304680250760785873799')
  (4, '0018529096043612327208547274580272321634069092581000185290960436123272085472745802723216340690925810')
  (5, '0034227064053927756900342270640539277569003422706405392775690034227064053927756900342270640539277569')
  (6, '0062925444068032162802087278020826123086044452926000629254440680321628020872780208261230860444529260')
  (7, '0028823456070107573304842270100167479397084062167405238739510206025238098927751506624298920345671179')
  ...
  (19, '0086429478094787353708082276960769671755062002581405814799730442823032030327719102646212500125075319')
  (20, '0070727070007072707000707270700070727070007072707000707270700070727070007072707000707270700070727070')
  (21, '0050025000011167116102722272220333873383049442944405550755050616621666077727772708388238880999479949')
  (22, '0000123468024692582604827272840628529642086432100000001234680246925826048272728406285296420864321000')
  (23, '0002621412032947913506462278580963075571028082329405076719170824429630014127735304680250760785873799')
  (24, '0018529096043612327208547274580272321634069092581000185290960436123272085472745802723216340690925810')
  ...
  (35, '0060227434056527793900602274340565277939006022743405652779390060227434056527793900602274340565277939')
  (36, '0032925054061432187802967276920878123416045052923000329250540614321878029672769208781234160450529230')
  (37, '0076823046079307576304102274800137479107085462182405718735410298025268091527798506324296020359671329')
  (38, '0048321482080252902406667276660420925208028412384000483214820802529024066672766604209252080284123840')
  (39, '0086429478094787353708082276960769671755062002581405814799730442823032030327719102646212500125075319')
  (40, '0070727070007072707000707270700070727070007072707000707270700070727070007072707000707270700070727070')
  (41, '0050025000011167116102722272220333873383049442944405550755050616621666077727772708388238880999479949')
  (42, '0000123468024692582604827272840628529642086432100000001234680246925826048272728406285296420864321000')
  (43, '0002621412032947913506462278580963075571028082329405076719170824429630014127735304680250760785873799')
  ...

Apologies for the wall of text but the output is included here to show the periodicity is 20 in the 10s digit for powers of n.

What do you think the periodicity in the 100s digit is?

Okay let's use our cache to just lookup the least two digits of say, 41679296 ^ 35 (without doing any real compute):

  1s digit
  35 mod 4 = 3
  187456329 index 6 -> 6

  10s digit
  (35, '0060227434056527793900602274340565277939006022743405652779390060227434056527793900602274340565277939')
  index ------96 -> 7

  The last two digits of 41679296 ^ 35 = 497981393104388234083687472206165994017494365065741791800324926291638211001469403054062791593737655590988883756006534609416313532750908441977019034732014717649617778735038244085515193249673259731575280032440237649262981884959117005449913087392615145373869733708824576 are indeed 76, cool