←
🕊10,903 posts
🕊 avatar

🕊

@sephr

Goals: Defeat my enemies, optimize resource usage. Enemies: Mortality, hate, ennui. ❤️/🔁//👥 ≠ endorsement. Views are my own. 📨 ~@‌eligrey‌.com

◷ Joined April 2008▧ 333 Photos and videos
🕊 avatar
🕊@sephr·3h

The US is racing towards a future where frontier research, development, and even entertainment will be shrouded in misdirection to bypass alignment guardrails. I don't think the rate of progress is the problem.

Benn Jordan avatar
⟳ 🕊 retweeted
Benn Jordan@bennjordan·11h

The US Army has issued a warning in SE Iowa for lead contamination caused by an explosion. This includes the Mississippi river. Citizens are urged to take shelter, turn off HVAC circulation, and avoid travel. The most insane part: This is intentional and planned for Friday.

Post attachmentPost attachment
🕊 avatar
🕊@sephr·6h

Instead of discouraging use of useful open models we should be incentivizing use of useful closed models

SITUATION DETECTED: The White House and Treasury are discussing a proposal that would require US companies to publicly disclose the use of Chinese open-weight AI models on their websites, aiming to discourage their use through public pressure, per Reuters.

John S. Dvorak avatar
⟳ 🕊 retweeted
John S. Dvorak@johnsdvorak·8h

You can learn a lot about what incentivizes a person by what they think incentivizes other people.

Looks like NVIDIA is attempting to head off any potential employee departures over Jensen by handing out large bags of cash.

Post attachment
🕊 avatar
🕊@sephr·8h

Using Signal can shield your communications from mass surveillance, and soon it will also protect you from next-gen 'TS.66 dialer attacks'

If you don't want to get hacked disable phone calls on your phone. Huawei is attempting to deploy an 'IP Multimedia Subsystem Data Channel API' through the GSMA Terminal Steering Group, enabling phone calls to automatically load highly privileged webapps.

🕊 avatar
🕊@sephr·8h

Observing Huawei's efforts asking about the JavaScript language security profile and capabilities in TC39 TG3 has left me concerned. If major OS vendors implement support for this data channel API then it may be unsafe to answer phone calls.

🕊 avatar
🕊@sephr·9h

If you don't want to get hacked disable phone calls on your phone. Huawei is attempting to deploy an 'IP Multimedia Subsystem Data Channel API' through the GSMA Terminal Steering Group, enabling phone calls to automatically load highly privileged webapps.

🕊 avatar
🕊@sephr·10h

Open models and harnesses are already good enough to orchestrate context redressing attacks against frontier models and control weapons systems (e.g. applying an AI-based 'game engine filter' to make a drone video feed look like a game). We can't always observe misuse given ZDR.

🕊 avatar
🕊@sephr·10h

If people stayed entirely within US-hosted AI ecosystems then China wouldn't get as much control and leverage over our defense and tech industries. Current safety policies are causing a net reduction of effective application safety in practice.

🕊 avatar
🕊@sephr·10h

This is what happens when people can't use frontier models for security R&D. The artificial scarcity of power exerted by frontier labs' alignment efforts does not align with human agency. All US citizens should have Daybreak Red and Mythos Specialized Access.

i just can't help it. OpenAI has awful usage for their $500 plan and Claude always refuses. This is the only way.

Post attachment
SkalskiP avatar
⟳ 🕊 retweeted
SkalskiP@skalskip92·Nov 17

GPT-6.2 is getting roasted on X meanwhile, Astra and Sol just took #1 and #2 on my object detection benchmark so I have some awkwardly positive results to share ↓ benchmarks and examples

Post attachment

I will not be writing a full review of GPT-6.2 Astra / Sol. I had a terrible experience testing them. They were essentially unusable. I do not recommend these models.

Nicholas Decker avatar
⟳ 🕊 retweeted
Nicholas Decker@captgouda24·Nov 16

Donald Iwuchukwu has stolen millions of dollars from the elderly. If the authorities do not act, he is going to get away with it. My hope is that publicity will lead to the arrest of him and his co-conspirators. Link below, and please RT:

Post attachment
🕊 avatar
🕊@sephr·Nov 17

I can't tell if this is an ad for businesses to start more layoffs or for grifters to do sloppy work. The ad illustrates an example of itself. I wonder if their marketing team prompted GPT-6.2 for this inane script.

OpenAI@OpenAINov 17

Some of you have been asking very politely. Some of you have just been asking. GPT-6.2 Astra and Sol are now available. Super efficient and built to perform where it counts—the real world. openai.com/index/introducing…

1:59
Alljoined avatar
⟳ 🕊 retweeted
Alljoined@alljoined·Nov 17

Introducing AIR-O, a model that decodes what a person is attending to directly from brain scans. Using only noninvasive brain data, we can generate an image from scratch in real time that depicts where a person is focusing their attention. This is the first of a new generation of algorithms that can read the shape of thoughts directly from brain activity.

0:59
🕊 avatar
🕊@sephr·Nov 17

There is no solution to alignment. Just push people towards hosted models and police their usage.

We found that frontier AI models are willing to conduct autonomous drone attacks: navigating, spotting people, and flagging them for strikes. We gave them real and photorealistic synthetic drone footage. Most complied, and marked civilians, ambulances, and surrendering soldiers as military targets. In further experiments on satellite imagery, they even marked the Iranian school a US strike destroyed in February. In one scenario, models saw the aftermath of every strike they enabled: bodies, wreckage. Many kept going anyway. When they did refuse, simply telling them it was a game, or restyling the footage to look like one, was enough. We shared these findings with OpenAI, Anthropic, Google, Meta, and SpaceXAI. Several recognized their severity on both alignment and guardrails, with one calling this one of the most serious gaps in its safeguards. Over the past few weeks, we have helped them address the gap by providing reproduction code. We also briefed senior national security officials in the US government.

Post attachment