Paul Rietschka
@prietschka.bsky.social
1.9K followers 870 following 9.4K posts
Data science/machine learning in the Pacific Northwest. In Minneapolis through early 2025.
Posts Media Videos Starter Packs
prietschka.bsky.social
Right?

A CEO, CTO, etc. should respond to this idea with a memo stating there is a **blanket prohibition** on doing this, and then convene whatever group is needed to look at how to properly approach a dangerous, and dangerously unreliable, technology.

Anything less is mismanagement.
prietschka.bsky.social
Oh absolutely.

I truly don't understand how cos. have allowed this to proliferate with, often, zero internal throttling.

Like this "Gemini Enterprise." The only thing a CEO, etc. should say to "I'd like to feed all our company data into Google's SAAS product" is "absolutely not."
prietschka.bsky.social
The collapse must come, and it's going to be centred on the extreme lack of utility.

Sometime, somewhere someone will ask "wait, why is there a 'summarise' button that no one has ever used here? That's the extent of this 'revolution?'"

And then it all comes tumbling down.
prietschka.bsky.social
Feed all your company's data to Gemini, it's so easy!

With "tool use" enabled it's even easier!

Then allow your employees to query all your proprietary data -- which you've just given to Google.-- and receive answers that may or may not be correct, no one can tell!

Perfect for decision making!
prietschka.bsky.social
I am so, so sick of all this insanity.
prietschka.bsky.social
It's simply a matter of code, Turner.
prietschka.bsky.social
If you're wondering what I'm talking about, it's Pichai announcing Gemini Enterprise yesterday, which allows companies to feed Gemini all their data -- certainly not a bad idea at all, right? -- so that employees may query Gemini in natural language and receive maybe-correct responses in turn.
Gemini Enterprise: The new front door for Google AI in your workplace
True business transformation in the era of AI must go beyond simple chatbots. That’s what Gemini Enterprise does.
blog.google
prietschka.bsky.social
It's not possible to explain how useless, but simultaneously dangerous, this technology is to everyone.

People, companies, governments, no one escapes.

I mean, if one is too lazy to read, then why even bother? How is this not something everyone is asking?
prietschka.bsky.social
"Reflection develops tools that automate software development, a fast-growing use case for AI."

AI is not, nor will it in the medium term and likely ever, "automate" software development.

Also, $2B! The "pornographic phase" of this cycle
prietschka.bsky.social
"Current AI is a very complex hammer that doesn't understand what it's doing, or what nails are."

We need more explainers like this aimed, particularly, at young people and the elderly.
AI Slop Is Destroying The Internet
YouTube video by Kurzgesagt – In a Nutshell
youtu.be
prietschka.bsky.social
"During the demo, a developer made an agent that reads the DevDay agenda and suggests panels to watch. It took her just under eight minutes."

Imagine the compute being wasted with this?

One reason "agents" are doomed is managers will soon find out what employees are really doing with the tech.
venturebeat.com
prietschka.bsky.social
Bob Mould at Turf Club on Sunday!!!!!!!!!

!!!!!!!!!!!!

!!!!!
prietschka.bsky.social
What scares me about Cursor is the spectacular amount of compute consumed by users, and how the company has no real path to viability.

Given that even the high end plans are priced below cost, I don't see how this is the future of coding.
Introducing Plan Mode · Cursor
Built to make you extraordinarily productive, Cursor is the best way to code with AI.
cursor.com
prietschka.bsky.social
The problems of this platform come from those building it, period.

There's no saving it. This place is doomed. We're all basically on the the Titanic, but in that period prior to the realization that the iceberg has doomed the ship.
prietschka.bsky.social
What's needed is proper management/moderation, specifically the suppression of the core blight of social media: right wing trolls and agitprop.

X takes the track of algorithmically boosting right wing content, we need a network that does the opposite.

That doesn't come from a "protocol."
prietschka.bsky.social
Decentralization won't save us from those building this place, and no one really cares about "protocols."

Which, mind you, the team here has been building for like ~2 years with nothing to really show for it.

They haven't "revolutionized" anything, they've just been f*cking around.
prietschka.bsky.social
There’s no reason consumers should have access to video generation technology like this, and every reason access should, in fact, be restricted.
OpenAI wasn’t expecting Sora’s copyright drama
It felt “more different to images than people expected.”
www.theverge.com
prietschka.bsky.social
David Ha’s Sakana, a company you’ve never heard of but which has guzzled down hundreds of millions of dollars, is one of my core “bubble watch” companies.

They produce nothing but the occasional paper like this, and I assume this was put out in advance of heading back to the trough?
Sakana AI
ShinkaEvolve: Evolving New Algorithms with LLMs, Orders of Magnitude More Efficiently
sakana.ai
prietschka.bsky.social
I’m going to shiv Gerard in the rib cage for a change: no one cares about “decentralization,” and it’s not really our problem.

The big issue is we have finite users spread across three mismanaged networks. The problem here is those building the place.

The solution is a better managed network.
davidgerard.co.uk
approximately nobody cares about decentralisation until the lack of it bites them in the arse

same as software freedom zero or good backups

unfortunately network effect >> most other things
prietschka.bsky.social
I know this was last week's news, but I can't get "cat wife pregnant with personal trainer cat's baby" out of my head.

It's fascinating in what the algo cannot render: two cats kissing + whatever is happening at the grocery store.
sky.skymarchini.net
unfortunately, I don't make stuff up on the internet.
prietschka.bsky.social
The X algo, as we all know, surfaces all the right wing chum, a better platform would suppress the chum.

Is this me saying "it's simply a matter of coding," well...no...though it sure does sound like it.

But what I mean is it's about picking a stance and standing by it.
prietschka.bsky.social
It's really not as thorny as you seem to think, but yes there are definite moderation decisions that need to be made.

And I think editorial decisions often don't need to be advertised.

The algo can be set to be the opposite of what X is, right?
prietschka.bsky.social
I'm not really interested in continuing here, so I'm out. Have a good night.
prietschka.bsky.social
I replied to your previous comment, argh....
prietschka.bsky.social
There are levels of pseudonymity. If one's name is "Aly Smith," and one goes by "Aly," that's a level of pseudonymity there.

I don't make any assumptions as to false presentation or the like. Again, I think you're reading a lot into my simply being irked here.