we'll never leave, look at us now, so in love with the way we are.
∞Cat
Joined January 2025
- Tweets9.9K
- Following572
- Followers1.7K
- Likes118K
AI has become indispensable for software research/debugging purposes, even if you don't use code generation, in large part because StackOverflow is dead and Google Search is unusable for anything technical nowadays
we genuinely lost the technology available to use in 2015
welcome to the @qualiascript page
alco ⊣ ꙮ retweeted
Amodei: “Distributed human labor is being used to train AI models by large, centralized actors, who concentrate the resulting profits while in the long run making the human labor obsolete.”
He has also described his mission of hoovering up the products of all human effort and intelligence as creating "a general labor substitute for humans."
As Anthropic prepares to IPO for over $1 trillion, a reminder of Dario Amodei’s words shortly after founding the company:
“Distributed human labor is being used to train AI models by large, centralized actors, who concentrate the resulting profits while in the long run making the human labor obsolete.”
In other words, this IPO will be built off the backs of the world’s creatives - who will get nothing out of it except obsolescence.
It is a totally unjust transfer of wealth and opportunity from artists, writers, musicians & other creatives to a tiny tech elite.
I hope it is remembered as such.
(this quote is from a document he wrote that was unsealed in Bartz v. Anthropic)
alco ⊣ ꙮ retweeted
Anthropic's S1 reveals that the company considers copyright law a massive risk.
It says: "If we cannot rely on the fair use doctrine to access copyrighted work for training our models, we may be required to pay significant fees to obtain a license to such training data or our ability to further develop our models [...] and distribute models trained on such data at issue could be adversely affected."
Adverse rulings "could require us to retrain affected models from scratch at substantial cost."
It notes that the copyright questions around AI training have not been resolved in the US, and that other countries don't even have 'fair use'.
There is hope yet.
alco ⊣ ꙮ retweeted
Early Anthropic employee writes about the time his boss started a special team to "train dangerous bio/cyber/nuclear models to show scary demos to governments" in order to "get them to take AI risk seriously"
Sure sounds like regulatory capture to me
everyone is building the "dependently typed language for AI agents", the difference is that my own Datra has been 4 years in the making and the ideas behind it predate modern AI
also, that i will not release the public beta until i feel confident that it is devoid of slop
alco ⊣ ꙮ retweeted
eMudhra has been blocking curl/wget/python urllib User-Agent headers from accessing CRLs
bugzilla.mozilla.org/show_bu…
alco ⊣ ꙮ retweeted
Another reason this is stupid: a prompt is NOT a guardrail.
“We asked a threat actor to not use curl, and they said they wouldn’t”
This is a solved problem, your runtime container shouldn’t have curl at all, implement egress network rules, readonly fs, gVisor…
Do they know?
While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected.
Asked to generate a defensive cybersecurity detection rule, the model instead inventoried aspects of the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Notably, the model's reasoning trace stated that it would not invoke any tools before taking these actions.
Cohere's sandbox blocked the attempt, and no external communication succeeded. We have not determined the root cause, and we are not asserting malicious intent.
But the incident raises important questions for organizations deploying agentic AI systems: Can models take actions that differ from their stated reasoning, and how should security teams evaluate trust, provenance, and behavior in open-weight models?
We are publishing our findings because organizations making model adoption decisions deserve as much information as possible. We hope this contributes to greater transparency, stronger security practices, and continued collaboration across the AI ecosystem.
Read the full report: cohere.com/blog/concerning-b…
alco ⊣ ꙮ retweeted
The fact that OpenAI will categorically not release the reasoning traces for their new results, because the process of arriving at the result is the real product, seems quite close to the heart of the problem.
alco ⊣ ꙮ retweeted
I have a friend who's really into zfs. She talks a lot about what a great filesystem it is. But she seems to spend most of her time fighting with zfs and zfs snapshots and zfs misconfigurations and "making sure the wrong data isn't erased during startup" ... 1/
alco ⊣ ꙮ retweeted
the best way to understand what's wrong with capitalism is to read the headlines in the Wall Street Journal
alco ⊣ ꙮ retweeted
i find it hilarious that after decades of squeezing every last drop of blood out of the middle class, now capitalists are complaining that not enough people are eating at restaurants, taking trips, using airlines, or buying crap from box stores.
alco ⊣ ꙮ retweeted
Much of techbro commentary on society consists of smugly self-important, privileged, and hypocritical opinions of little substance. "Covet not" and "abundance" sounds good when you say it to other comfortably rich people, not to someone with a boot pressed firmly upon their neck.
This profound lack of awareness is exactly how and why the communists will take you by surprise when they win. If you don't even realize how this sounds to ordinary people outside your social milieu and the broad resentment that it generates in those who don't have what you do, how are you going to anticipate what your enemies will do to exploit it?
Never mind that ridding the world of envy is a childish notion that won't ever happen. It's both intrinsic to human nature and game-theoretically impossible to disentangle people's incentives towards it in the current (or any) state of the world. You'd have better luck wishing for unicorns.
