Stability AI decimates staff just weeks after CEO's exit

Stability AI is laying off staff in its search of steadier footing following the sudden departure of its CEO late last month. The British generative AI builder revealed the layoffs Wednesday night in a memo to staff obtained by The Register. "Following a review of the global team, we have determined the need to restructure …

[cyberdemon]

... And nothing of value was lost

The post is required, and must contain letters.

[b0llchit]

Re: ... And nothing of value was lost

Oh yes there was... Money was attracted in a gravitational pull into a dark deep pit where no light can escape and information gets destroyed.

[beast666]

Why don't they ask an 'AI' how to run a profitable business?

[AVR]

Why are you assuming they didn't do so? AI hallucinations are the opposite of reliable after all.

[Sorry that handle is already taken.]

All output from a LLM is a “hallucination” — generated from the latent space between the training data.

I recently asked Google Gemini to summarise the findings in a rape case that had just been handed down. It produced very well-written and mostly correct text that... swapped the offender and victim.

[Michael Wojcik]

This is Stability. They could ask it to draw a picture of a profitable business, but that's about it.

Diffusion models are a theoretically interesting technology, but I'm not keen on using them for their original purpose, for all the reasons I've stated in posts in other threads (encouraging intellectual laziness, learned helplessness, etc); and I don't see another obvious application.

[Plest]

winners and losers

"survival of the fittest" exactly as Darwin described, in that "fittest" should always be seen as a synonym for "most adaptable".

[Pascal Monett]

Re: winners and losers

And "with enough money"

[Michael Wojcik]

Re: winners and losers

This particular industry segment is more "survival of the richest". OpenAI, Google, Microsoft, and Meta are going to spend down their war chests and let the upstarts run out of cash. Amazon, Google, and Microsoft will make money selling shovels (GPU cycles) — which obviously will help subsidize their own "AI" investment.

Trying to get into this business without a lot of capital on hand and/or fat revenue streams is a bad, bad idea. There's little moat — even if you invent some new twist that the market notices, the big kids will duplicate it quickly. And then they'll just wait until you starve.

[Sorry that handle is already taken.]

Re: winners and losers

You could, if you felt so inclined, argue that the bankroll is what technically makes them "fittest". Like the squirrel who stored the most food for winter...

[Will Godfrey]

Congratulations El Reg

This has to be the very first time I've seen any publication use the word 'decimated' correctly!

[Anonymous Coward]

Re: Congratulations El Reg

...but not "let go"...

:-/

[Johnb89]

A sustainable business

Is not one that has direct COGS 9x revenue, that also relies on stealing critical inputs.

Then again... the CEO of GM was once quoted as saying they lose $1000 on every car but hope to make it up in volume, so what do I/we know?