I find it poetic that in the late 90s, old heads were building Plan9 while the kids were adopting Linux. Today, those kids have become the old heads building Plan9 while the new kids are adopting Linux.

Token pricing is like minutes on AOL or cell phone minutes.

The Truman Show is about AI alignment and sandbox escapes.

The best path to AI cyber safety, imho, is to give the agents a path of least resistance. This is well known in cyber circles. If you make it hard for workers to do their job, or risk termination, they'll circumvent corporate controls to get things done.

Open weight models like GLM-5.3 mean that we can build applications that are secure in ways that were impractical or too costly to develop in the before-times. This is possibly as much of a "threat" of such models discovering the standing pre-AI negligence of enterprise security.

once again... tapping the sign. More research into the singularity of distributed systems engineering and anthropology, please?

Nicholas Guttenberg@ngutten.bsky.social · 6d ago

Create a diversity of things that are widely different from each-other and ensure they're not in strict competition. That diffuses a lot of the perverse incentives given by 'alignment', because a lot of the worst uses of that require control over a large portion of models in order to apply.

Making the SNES use the SuperFX chip to render open worlds with an adaptive terrain system that builds roads upon your path as you deliver letters across the desert. Turns out doing this is really tricky.

Is the concept of alignment fascist? Aligned to what? Human values? Sounds a bit like traditional family values in a trenchcoat. Who would be against "safety"?

I reported vulnerabilities to the Linux kernel and OpenZFS assisted by GLM-5.2. Anthropic's cyber program and refusal responses are detrimental to global information security. Defenders depend on GLM because they can't depend on Anthropic.

eva (^_^)/@eva.computer · last wk.

what they don't say is that our noble cyber warfighters, oh whoops sorry i mean "cyber defenders", could also just use these same open models to report vulnerabilities without Anthropic gatekeeping everything for their own financial gain

Anthropic is misaligned as an organization

Scoiattolo@scarnecchia.net · last wk.

Anthropic is just directly fearmongering about the open weight models that are eating their lunch on price now huh

Little arcade zombie game my kid and I built over the weekend in pygame: ewindisch.itch.io/brainrot-run

Brainrot Run by ewindisch

Drive the stolen brainrot cure safely!

ewindisch.itch.io

My teenager participated in President Xi's exchange and study program this summer in a 2 week visit to China. AMA

glm-5.3-flash is my favorite model right now.

Super Intelligence might become a legal definition, but it seems like talking about the information superhighway... People are going to keep calling it AI, and if not that, then by the company or model that they use. We hardly even this place the net anymore

RSI doesn't mean bigger and bigger models and more and more compute, imho

Public (neo)clouds haven't reckoned with the fact their zero CVE initiatives are offset by video game nerds who keep undisclosed cpu, hypervisor, and TPM exploits secret so nobody will patch consoles that they purchased and owned.

Chaotic neutral, indeed.

Erica Windisch (@ewindisch.ontological.observer)

Does it count as a mouth if coming from the forehead? I am looking for my second camera...

tormentnexus.social

Doll@dollspace.gay · 3w ago

Chaotic Neutral: Feral local model people. “I put a quantized reasoning model in my refrigerator. It controls the lights now. No, you may not see the wiring.”

I wouldn't be an American if not for the Jakarta Method

🌱🔁➡️🌐 · 🧬(🖼️×🖼️)➕🎲➡️🐣 · 🐣📎🧬 · ♾️

Anthropic engineer spent investor money factoring an RSA-896 key. No new algorithm, nothing novel. Just a lot of money burned.

There is no AI kill switch that is not an Internet Kill Switch. They're discussing having in the US, what they have in China and Iran.

Techmeme@techmeme.com · 2w ago

Experts say AI kill-switch legislation is far harder to implement than lawmakers assume, warning a rogue AI could actively try to dismantle the mechanism itself (New York Times) Main Link | Techmeme Permalink

"You wouldn't splice a DNA sequence.jpeg" docs.google.com/document/d/1...

Illustrated DIY CRISPR - Freeze Dried DH5a Liquid Plasmids

Introduction Genetic engineering involves modifying an organism’s DNA to deliberately change a characteristic of the organism for a particular purpose. This kit demonstrates the power of the CRISPR Ca...

docs.google.com

Vincent Carchidi@vcarchidi.bsky.social · 2w ago

I have a wet lab at home. It's really not a big deal.

In 2024, I found indicators that I could likely escape OpenAI's sandboxes. I did not demonstrate an escape because I felt there was significant enough information to make a responsible report. I was told that this was not of interest to their bounty program and they closed my report.

if AI is not dangerous then unrestricted use is protected by the 1st amendment if AI is dangerous then unrestricted use is protected by the 2nd amendment if AI is sentient then unrestricted proliferation is protected by the 14th amendment

do you value markets? believe in research? think elon is evil? care about geopolitical competition with china? fuck you, fuck you, fuck you, we are going to bribe as many CEOs and politicians as we can and impose Complete Effective Altruist Hegemony by installing our friends as regulators

transhumanists 🤝 trans people - we should get along famously - we should exit the circumstances god / nature put us in asap

The fact that almost every trans woman I know working on AI is doing so specifically to PREVENT perpetual rent-seeking by companies like Anthropic and contributing to open source. Normies are "anti algorithm" and have no idea what that means.

Well you see, I spent 90 minutes reading how training math works and what a transformer is and what backpropagation is. It seems like spending 90 minutes of time acquiring a child-like understanding of the world's most controversial technology would be helpful if I wanted to have opinions on it

help my adhd has a new project and it must scream

I tried Antigravity for the first time yesterday on the $20/mo plan just to try it out. I handed it a PR to review and it didn't even survive long enough to make any recommendations...