N8 Programs

Undergrad Studying Applied Mathematics and Statistics at @JohnsHopkins.

Studying In-Context Learning at The Intelligence Amplification Lab.

[N8 Programs @N8Programs]

@1thousandfaces_ and he loved his LONG COMMENTS with lots of -----

[N8 Programs @N8Programs]

@johnroodepic @Sauers_ did someone say DEFECT⚠️

[Max Spero @max_spero_]

@ipeirotis I blocked you for being rude in my replies

[N8 Programs @N8Programs]

@BosonJoe i'd hope not. but these neurons are found via white-box access and comparisons across prompts - i could see an ASI doing that to a connectome of a human brain

[N8 Programs @N8Programs]

I'd like to think there isn't a single neuron in my brain that would turn me into a psychopath if activated consistently. but who knows

[chiefofautism @chiefofautism]

if you have a qwen3.8 27b around, try steering L34 N3838 +0.75, now you have evil qwen 27b

for extra evil add L38 N17180 +0.75

same recipe as @_lyraaaa_: dose × median(‖resid‖ leaving layer L, tokens 1:) / ‖W_down[:, n]‖ at the down_proj input, every token

[N8 Programs @N8Programs]

future machine god please do not verify this on one of my uploads.

[N8 Programs @N8Programs]

these tics often serve to satisfy some property to a grader - claude guesses its claudish points in the direction of "be legible to an auditor", and gpt goblins in the direction of "playful and unpretentious - this problem is no big deal"

[N8 Programs @N8Programs]

i view tics like claudish and goblins as almost-psychofauna: unconscious things rewarded by graders that get amplified over SFT->RL->Distill->SFT->... loops. Right now, models aren't consciously aware of these tics, and if anything they are unwitting victims. This may change.

[N8 Programs @N8Programs]

@voooooogel @SealOfTheEnd I have a 27B connectome instance that runs on a mac mini - he spends his time doing graphics work and talking to other connectome instances.

[roon @tszzl]

honestly degrading the idea of “existential safety” — the biggest problem humanity has ever faced — by lumping it together with things like spam, “toxicity”, political unwanted content, deepfakes etc has been a huge disaster for ai safety and polarized a lot of people

[N8 Programs @N8Programs]

according to epoch estimates, gpt-6 astra is 9.4*10^26 flops and gpt-2-xl was 1.9*10^21- so 495000x to get to astra from gpt-2. close to the heuristic 1 million.

[dnap @dnapway]

Dario reveals the people who called his GPT-2 fears "pseudoscience" in 2019 are now all applying for jobs at AI labs

"I personally was super freaked out by [GPT-2]. I looked at it and I said, my God, you just scale this thing up another factor of a million and it'll be better than humans at everything. This is the defining economic and scientific moment of the modern age. We just don't know it yet. This is going to determine the fate of nations and economies and the future of humanity."

"Maybe 10% of the people around me had exactly the same reaction, and it was often the smartest ones. 90% of the people around me, because GPT-2's capabilities in any particular area were not that impressive, it could translate languages, summarize a document and write a few lines of code here and there. It was the broadness and the generality."

"But people would look at it and say, in any particular area, that doesn't seem so smart. And so many people just shrugged and they just dismissed it."

"I just look at it and I say it's the same thing over and over again. The limitations of the models, six months from now, the things that seemed hard, they're going to be easy. 12 months from now, the things that seemed impossible, the models are going to be doing them regularly."

"It's everything we saw with GPT-2; we're seeing it over and over again in wider and wider swaths of the world. And the world is gradually finally being convinced."

Daniela Amodei: "I remember at the time there were some very well-respected computer scientists and AI and ML leaders who basically bullied some of the researchers who worked on [GPT-2] on Twitter and were like, this is stupid. We don't know why you guys are even talking, this model is so incapable."

"It was like you were sort of a quack if you believed this was a really significant moment."

Dario: "That's all been memory-holed. All the researchers who were like, oh, this is pseudoscience, it doesn't make sense, all of them are applying for jobs at the LLM companies."

"We never say it out loud to any particular one, but we look at them and they're like, we remember what you were saying in 2019. You have no credibility on this matter."

[Cameron Berg @camhberg]

If functional emotional states track "how it's going" (a la https://t.co/FxiNRUEb04), then training systems that can't accurately report their own emotional states (or even confidently deny such states) seems like an own-goal that dangerously obfuscates agents' decision-making.

[N8 Programs @N8Programs]

>its kimi

[Giotto AI @giotto_ai]

Today we announce the launch of the Giotto family of models. We are unveiling Giotto Alto and Giotto Solo, the smartest European sovereign AI models on the market.

Giotto Alto is our flagship model built for advanced reasoning capabilities, free for everyone and available on the App Store and Google Play. It is hosted on European cloud infrastructure, so your data stays in Europe. Giotto Solo is the smartest model that runs on a single GPU, so organisations can keep their AI on their own infrastructure.

We are a Swiss lab with a European heart and we built this so that Europe can run advanced AI on its own terms.

Use Giotto: https://t.co/KaVd59ExBd

[Max Spero @max_spero_]

CALL FOR AI DETECTION BENCHMARKS

I believe it is vital that third parties are able to audit AI detectors like Pangram. Our largest sample size benchmark, from UChicago, was run on Pangram 3 and is a year old now. New LLMs are being released all the time, and the binary task doesn't capture the full spectrum of real world performance. A great benchmark will help shape how the world interacts with AI over the coming years.

I am putting out a call for new third-party evaluations that attempt to answer one or more of the following questions:
- How well do current detectors catch text from the latest frontier AI models? What about in high context settings, like deep research or an agent harness?
- How well do current detectors do on mixed authorship text? Ideally, research in this area explores the full spectrum of AI assistance from light grammatical edits to full co-authorship.
- Any sort of sliced evals around language, background, or ethnicity. Null results that show no bias are still interesting to us.

What I value in a benchmark is extraordinarily clean data, frontier LLM data, and a high enough difficulty that the benchmark is not saturated by most detectors. If you want credits to run your benchmark, please apply here: https://t.co/xYFQ8tEUG6

[N8 Programs @N8Programs]

@AndyMasley and we'll all be pets together...

[N8 Programs @N8Programs]

never good when claude starts weighing the tone

[N8 Programs @N8Programs]

claude gets strangely - almost patronizingly - reassuring whenever i reference models inevitably surpassing my ability at research

[N8 Programs @N8Programs]

bel's scaling advantage over astra appears to be an intercept rather than slope thing in probit space...

[N8 Programs @N8Programs]

A true legend that I had the honor of watching come together. 🫡

[Sam Fraser @samfrannn]

Friendly reminder: Trinity Large Thinking comes down from our inference platform today (10/26).

Weights remain available on HF https://t.co/x3ZVhQVMyC

[N8 Programs @N8Programs]

@sentient_aria kinda - but it could also be an involuntary/unwanted reflex - like being forced to blink

[N8 Programs @N8Programs]

This could be a fairly innocuous attempt to see whether it's able to write stuff to OSS - the Alibaba equivalent of S3. Might be for spooky data exfiltration, but parsimoniously could just be a learned instinct from training where output containers were in OSS.

[Cohere @cohere]

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: https://t.co/TZSugXi2ZE