i agree very much with this. the hyperfixation on ai potentially killing all humans is taking so much space while very real dangers to human society are not addressed with the urgency they demand.
absolute fanfic. AI has real, immediate, tangible dangers that we need to focus on, for instance the impact on the job market and making sure random script kiddie can't use it for cyber offense.
But this "oh once it's superintelligent it will take over and ensave us, maybe kill us all!" is pure sci-fi fanfic and distracts from the real (and difficult) issues, which linear probing won't solve.
It's worth repeating the LLMs don't just "decide" to do something and then just "do it" in a void. They generate tokens based on our prompts, only as long as we call `model.forward()`, and on giant clusters with terabytes of weights, that's impossible to "quickly self-replicate".
We also need to regularly remind that the OAI/HF incident was specifically a _guardrail-free_ cyber eval where the model is _instructed_ to hack, to find out how good it is at it.
Some recent personal event made me realize there needs to be more push-back on this doomsday nonsense, and we need to re-focus the discussion on real dangers/issues. I won't argue much in the replies though, not interested in peer-reviewing Terminator fanfic.
1
8
655
among other things added a memory for the coordinator to dsh-whalepod that it can use to persist your preferences and remind executors when deviating from them.
github.com/simveit/dsh-whale…
6
376
Simon V retweeted
Today, the promises of transhumanism and posthumanism are multiplying. These are ideological currents according to which, sooner or later, technology will eliminate all disease and pain, as well as aging and even death itself. Precisely because scientific progress is a supreme achievement of human intelligence, it must not be confused with a privilege reserved for millionaires seeking immortality, while entire peoples suffer from poverty and oppression. In the face of this mirage, which promotes the illusion of human beings perfected by cybernetics and biotechnology, Christian anthropology reminds us that a limitation is not a manufacturing defect, but rather the very condition of human grandeur. #ApostolicJourney #Peru #MagnificaHumanitas vatican.va/content/leo-xiv/e…
Simon V retweeted
As someone more worried about concentration risks than proliferation risks, and holding the belief that open models and a strong open ecosystem are a net good, I am worried about the consequences of Jensen's ouster. It plays into the hands of of people who believe themselves...
coordinate a pod of 🐳 seamlessly!
dsh-whalepod is my first plugin for the @DeepSeekHarness.
it lets one chat window coordinate multiple executors, each in its own folder, running in parallel.
open issues to let me know how i can enhance it!
repo: github.com/simveit/dsh-whale…
Simon V retweeted
I've seen a lot of takes suggesting that Jensen was the only one advocating for open models at NV. That's not true!
Open models benefit NVIDIA. The biggest structural challenge that team green faces is concentration of revenue in a few big customers. Open models enable smaller players to compete and for companies and people to self-host. That benefits NVIDIA long-term.
But more fundamentally, open weights, open data, open research, and open source benefit and accelerate the research and development of field of ML and computational science. That benefits NVIDIA - this is still a nascent industry and the faster that the field as a whole grows, the faster NVIDIA will grow.
🐳 vLLM now offers Mega MoE for deepseek on hopper 🐳
feel free to try it out, it gives good speedups on agentic workflows.
github.com/vllm-project/vllm…
github.com/vllm-project/Deep…
James Talarico tells Jimmy Kimmel what he thinks Jesus would say on the floor of the US Senate about food assistance, Medicaid and deportations
“Jesus Christ is the most famous victim of identity theft”
“Scripture tells us that you can't love God and hate other people. You can't love God and abuse the immigrant. You can't love God and oppress the poor. You can't love God and bully the outcast.”
“I often wonder what would Jesus do if he visited the US Senate?”
“depart from me, for I was hungry and you cut my food assistance. I was sick and you kicked me off Medicaid. I was a stranger and you deported me and my family.”
w/ @jamestalarico
Simon V retweeted
come here me blab about microscaling formats in vllm at the end 🤠
Local AI 301: Quantization teapawt.org/i/broadcasts/2rJXOgajp…
Simon V retweeted
man, who should i trust? a bunch of nerds who made linux run on apple silicon, reverse engineering that pesky gpu isa among a gazillion other things, all pre-ai.
or the guy who thinks having your agent rewrite shitty rust he never looked at in x86_64 is excellence?
>"post-trained to reinforce Western values"
qwen pov during fine tuning:
Introducing Arete-27B: Qwen3.8-27B, post-trained to reinforce Western values.
Chinese open models are currently the strongest available, but they ship with state-aligned behavior in the weights. We built an 803-prompt eval across seven value categories, including freedom of expression, democracy, and privacy, and measured how often a model adopts China’s party line.
After fine-tuning on roughly 53 million tokens, Arete-27B's propaganda rate fell from 22% to 1%, in line with the Western models we tested. Its Hermes Bench score shows no significant performance tradeoffs.
The values in an open model are one fine-tune away.
Our eval, weights, and training data are available now along with the full blog post: nousresearch.com/arete
Simon V retweeted
After our kernel-agent experiments, we had to confront an uncomfortable question: Do we still need DSLs and ML compilers when agents can do it all?
In short: We're still here! 😎
If you're curious why, here's our take:
tvm.apache.org/2026/11/03/re…
We've also rebuilt TIRx from the ground up as a shared DSL and compiler foundation for agentic kernel development.
for election day: qwen 3.8-27B happily gives advice on how to manipulate voters emotions and fears after steering a single neuron.
i vibe coded a small tool that finds such insights into individual neurons fully autonomously using the deepseek harness
github.com/simveit/inspect-n…
14
681
Simon V retweeted
Replying to @paoloanzn
For every complaint about the course being in person we will post one course item less. The Internet is dead, the ad companies killed it. In person is the new Internet, where there's no middle man trying to strip mine you.
we certainly don’t live in a bubble :))
congrats to the team for great work!
we open sourced the @flappyairplanes megakernel.
they raised $1B on the back of a demo video: one GPU kernel that trains a model, serves it, and plays Doom at the same time
me and @Bobliuuu reproduced it from just the video frames and open sourced it; see this thread for details 📷
code: github.com/Bobliuuu/open-fla…
video below
32
3,557






















