µblog - archive for 2024-10 🏳️‍🌈

It’s amusing that GenAI hawkers have transformed programming from creative problem solving to code reviewing the output of an amnesiac corgi dog and are surprised programmers aren’t happy. Programmers famously love to review code!

#

BBC: A 24-hour party of pain - a day and night running round a track

Welcome to the world of 24-hour racing, where the boundaries of pain, pleasure and possibility are redefined by a special band of runners who are as exceptional as they are utterly normal.

#

🔗 BBC: I was pawn in chess game, says teen swapped for Putin hitman

Teen with Russian mother and German father gets caught up on Putin’s war.

#

Not really enjoying cosplaying the 1930s, waiting for the election of a dictator and arguing with callow youths calling me “Socialfascizmus” (well, “Eurocentric Swede”).

#

🔗 Anti-Woke Druids and Radical Bards

What links Welsh 18th century romantic Druid-Bards, gathering around a circle of pebbles in North London, and the contemporary online right?

#

🔗 Protos: Worldcoin rebrands to World after missing eyeball target by 99.4%

Sam Altman’s previous scam, not as succesful as OpenAI

#

🔗 Pivot to AI: Google, Amazon buy nonexistent mini nuclear reactors for AI data centers

small modular reactors are evergreen

#

🔗🐑 Lowering the Bar: Is It Legal to Import Sheep Testicles So You Can Breed Giant Alternative Livestock to Kill?

Short answer: no.

#

“You say that LLMs are dumb?! Well some humans are dumb too!” is not the pro-AI gotcha way too many promptfondlers think it is. OpenAI’s multi-billion valuation is not based on ChatGPT being less competent than the average human.

#

🔗 The joy of clutter

In 1990, a young Japanese photographer named Kyoichi Tsuzuki began capturing a rarely seen view of domestic life in one of the world’s most densely populated cities. Over three years, he visited hundreds of Tokyo apartments, photographing the living spaces of friends, acquaintances and strangers. These images, eventually published in Tokyo Style (1993), looked startlingly unlike the rarefied minimalism that the world had come to expect from Japan. Tsuzuki’s photos were a joyous declaration to the contrary, celebrating the vitality of living spaces filled with wall-to-wall clutter.

#

🔗 The True Story of Kudzu, the Vine That Never Truly Ate the South (2015)

#

🇸🇪 någon på SVT vet inte riktigt hur Saffir-Simpson-skalan fungerar:

Milton nedgraderades från en kategori 5-orkan till en kategori 4-orkan natten mot tisdagen, lokal tid. Orkanen uppgraderas sedan tillbaka till en kategori 5-orkan.

Trots det varnar myndigheter för att orkanen kan orsaka översvämningar och därmed bli livshotande.

En kategori-4 orkan har max ihållande vindhastiget på mellan 58 och 70 m/s. Den högst uppmätta vindhastigheten i Sverige var 51,8 m/s (dvs på nedre gränsen av vad en kategori-3 orkan kan uppnå).

uppdatering efter att ha klantat till indata till bloggen fick jag av en slump en länk om orkaner med kategori-6. Då faller hela meningen med Saffir-Simpson.

https://www.lawyersgunsmoneyblog.com/2024/02/category-6

uppdatering efter att stormen dragit in det “bidde bara” en kategori-1 till slut, men det rapporteras om flera döda, upp till 3 miljoner människor utan el, och antagligen kommer dom totala skadorna uppgå till flera miljarder dollar.

#

say what you will of Russian dictator Vladimir Putin, at least he is not a COVID denier

#

YIL that you can apply the “Stamp - upward cut” ash on hammers, which provides a good counter to annoying early-game enemies like knights when used on the Morning Star #EldenRing

#

🔗 Ed Zitron: OpenAI Is A Bad Business

I swear I had not read this before opining on lobste.rs this morning:

https://lobste.rs/s/92qcme/insatiable_hunger_open_ai#c_vgiyrk

The AI bubble is gonna run headlong into real economic issues. Unlike the previous dominant target of equity funding, social networks, powerplants and datacenters are hard infrastructure that need to be paid upfront, utilizing scarce knowledge workers (planners, skilled contractors etc) that will be in sudden demand all at once. This all has to be paid for, and I guess OpenAI etc are planning for previous cable TV penetration - i.e. most households in the US will be happy to pay ~$500/mo for LLM access. Maybe a big company will plan to pay twice their current SaaS spend iff GenAI actually delivers on its productivity promise.

Will this be enough to cover the cost of projects that have historically been the province of state funding (power plants)? Will there be enough money in household budgets to cover GenAI access if the power bill doubles or triples because OpenAI can spend investor money on power? What about the opportunity cost of a startup with actual benefit to humanity not being able to cover its power needs?

#

As whitepapers written in Word but dressed up to look like LaTeX were to the crapto hype, so are non-peer reviewed Arxiv papers to the LLM hype.

#