Hot off blowing over $100 billion on the Metaverse, Facebook/Meta has started doing the same with “AI.”

Facebook just bought 49% of Scale AI for $14.3 billio, acqui-hiring founder-CEO Alexandr Wang, in a deal designed to keep antitrust regulators at bay and Scale’s original investors sweet. [Reuters; The Information, paywalled]

Scale will continue as a separate company — but Wang’s day job will be at Facebook.

Zuckerberg’s been poaching hot AI researchers for his new “superintelligence” unit, offering them tens of millions of dollars over several years. Zuckerberg has rearranged desks at Facebook’s office in Menlo Park so the new staff will sit near him. [Bloomberg, archive; Bloomberg, archive]

Wang will join this new unit, where they’re super-sure they will super-build an Artificial General Intelligence!

Wang isn’t even an AI researcher. He’s a business guy with some tech knowledge. The Financial Times says “His talents lie in promoting the company rather than managing its staff or furthering AI research.”  Zuckerberg seems to think this sales guy is the new Facebook executive he wants. [FT, archive]

Scale creates datasets. It outsources this data creation to underpaid gig workers — who it then rips off.

The big problem for all the LLMs is they’ve been fed the whole internet and they’ve run out of data. But Scale’s army of underpaid workers generate new data. That’s the other thing Facebook really wants out of this deal. [Semafor]

We mentioned “AI” to “cocaine” the other day. AI is God’s way of saying Facebook is making too much money.

• Video version

Ich hab’ mal wieder ein Rad gebaut.

Ich habe vor Jahren nach meinem Infarkt mal ein Peugeot gekauft und bin es nur recht kurz gefahren. Dürfte orignal in den frühen 1980ern gebaut worden sein. Alles Shimano 600. Schön, aber später wurde dann der Vorbau samt MTB-Lenker komplett verwurstet. Gutes Rad, aber nicht das Ende meiner Idee eines wirklich guten Rades. Mit 56 cm ziemlich großer Rahmen auch. Ich fuhr es dann kurz bis ich wirklich das Ende meiner Idee eines guten Rades fand; ein Stevens Stevens Izoard Pro aus den späten 00ern. Halb Alu, halb Carbon. Vintage, nicht Retro. Kann man jetzt drüber diskutieren, muss man aber nicht – ist halt das Lieblingsrad meiner Wahl. Fertig.

Dann stand das damals noch schwarze Peugeot hier so rum, wie hier eh zu viele Räder rumstehen, aber die werden ja nicht schlecht. Nackig gemacht, zum Pulvern gegeben, grün, gelb, gold wollte ich gerne.

Dann kaufte ich so Konvolute an Fahrradteilen, um andere Räder neu aufzubauen. In einem dieser Teile-Sammlungen waren originale Aufkleber eines Peugeot Prestige aus den frühen 1980ern, kurz vorher in UK gekauft.

Und dann dachte ich, wenn du schon so einen alten Rahmen aus dieser Zeit und die passenden Klebchen dazu hast, musst du ja quasi ein Rad daraus machen. Geht ja gar nicht anders… Gesagt und mit ganz viel Trail und Error nach drei Jahren dann auch mal getan und endlich fertiggestellt. Alles neu, inklusive originaler 80 Karat blattvergoldeter Gabel, die ich als Einzelteil dazu gekauft hatte: Peugeot Prestige 1981 im Jahr 2025.

Und vielleicht hat die Erstgeborene jetzt mal wieder das schönste Rad der Stadt. Die nämlich hatte Geburtstag und es dazu als Geschenk bekommen, was mir nicht ganz leicht fiel. Aber mehr als fünf Räder braucht man ganz objektiv halt auch wirklich nicht. Kommst ja auch gar nicht dazu, die angemessen regelmäßig auszufahren.

Jungfernfahrt hat sie gestern mit 30 Kilometern gemeistert. Let‘s bike, beauty!

Detailfotos:

 

After following their lush analog output via Ghostly International in recent years, I couldn’t be more excited to present an album by KILN here on ASIP. By way of an entry point to anyone new to the trio of Hayes, Marrison & Rehberg, think more electronic worlds of Christian Kleine, Bitstream, Freescha, ISAN et al, all captured as live performances, combined with micro-field recordings and further saturated with color and texture…

A big thank you to Sam Valenti IV for connecting our two worlds.

~

KILN return with an opulent new display of hue and swing on Lemon Borealis, a sumptuous gallery of dazzling motifs that display a finely hewn concoction of visual tones and vital pulse.

Across its 12 cuts, this collection utilizes a fresh process of condensing immersive sprawl into compact, punchy and colorful sound. Using aspects of live performance, beatmaking and waveform sculpting, the troika of Kevin Hayes, Kirk Marrison and Clark Rehberg III create evocative and invigorating dioramas, continuing to surprise and enchant listeners after over thirty years into their collaboration.

Deep in waves of Hi-meets-Lo Fi, KILN delivers a panchromatic daymark arranged to biochemically align and stimulate your personal syntax, forging a tapestry of sonic reveries ranging from the aquarium-on-fire radiance of DrnkGrlfrnd, a garden groove of field-recorded percussion in Maplefunk Diptych, to the sizzling guit-noise whiteout of Deacon Rayhand.

Their eighth album, and first for A Strangely Isolated Place, on Lemon Borealis, KILN expands upon the long-explored themes of mosaic texture, subtle melancholy, eroded consonance, and vivid cadence to reveal yet another aperture to their unique magnetic universe.

Lemon Borealis is available on 12” Transparent Yellow Smoke vinyl and digital on July 18th. Mastered and cut by Andreas Lupo Lubich, and featuring artwork by KILN.

See the album release page for full details to buy

 

Gilt nicht nur in den USA. Fahrradfahrende kennen das. „Freiheit muss mit Benzin angetrieben werden.“

(Direktlink)

Milde interessant: verschiedene Arten von Dachgauben. Bisher nie darüber nachgedacht, aber ja, die sehen ja nicht alle gleich aus. Hier ausführlicher erklärt. Wieder was gelernt.

(via Messy Nessy)

ChatGPT suffered a worldwide outage from 06:36 UTC Tuesday morning. The servers weren’t totally down, but queries kept returning errors. OpenAI finally got it mostly fixed later in the day. [OpenAI, archive]

But you could hear the screams of the vibe coders, the marketers, and the LinkedIn posters around the world. The Drum even ran a piece about marketing teams grinding to a halt because their lying chatbot called in sick. [Drum]

There’s some market in the enterprise for AI slop generators. Wrong text summaries, LLM search engines with data leaks and privacy violations, and rambling slop emails.

But the enterprise chatbot market is jobs and tasks that are substantially … fake. Where it doesn’t matter if it’s wrong. The market for business chatbots is “we pretend to work and they pretend to pay us.”

So what do you do in your chatbot-dependent job next time OpenAI goes down?

The AI bubble is not sustainable. It’s going to pop. And it’s not clear what it would cost to serve GPT without the billions in venture capital dumb money — if it had to pay its way as a for-profit service.

Ed Zitron estimated in October 2024 that OpenAI was spending $2.35 for each $1 of revenue — so prices would triple at absolute minimum. I expect much more than that, because OpenAI is spending money even faster than it was a year ago. It’ll have to go to five times the current price, maybe ten times. [Ed Zitron]

DeepSeek claims its prices, which undercut OpenAI, could turn a profit. But even they call this “theoretical.” [TechCrunch]

Tedious nerds are right now vying to comment on this post “oh you could run an open source model in-house.”

• So firstly, shut up, nerd. You’ve clearly never worked a corporate job and don’t understand the compulsion to in-house nothing.
• Secondly, approximately 0.0% of chatbot users will ever run an LLM at home.
• Thirdly, it’s slow as hell even on a top-end home Nvidia card — and the data centre cards are expensive.
• Fourthly, asking around, the few companies that do build their own in-house LLM systems pay thousands to tens of thousands per month to serve a few users with the same performance that most ChatGPT users just buy in as a service. [Mastodon]

How desperate is your boss to pretend he can replace you with a chatbot pretending to do a job? Will he pay five or ten times the price?

Anyway, I’m sure it’ll be fine and you can use your chatbot in confidence! Until you can’t.

• Video version