Nathan Jones

@HerodotusWave psst, there are insects with iridescence that most humans can't quite see!

A Victorian beetle wing dress decorated with the iridescent wing casings (elytra) of jewel beetles, specifically of the species Sternocera aequisignata.

@NoMansSky LOL WHAT

Hieroglyphs are an ancient, beautiful and mysterious writing system.

@DelusionPosting Music has been with humans since before speech. In fact, it used to be speech.

@MattWallace888 whoa great legs! happy hump day everybody

Feminists are calling for this weather girl to be fired for wearing an “inappropriate” outfit on set

@ShaneFrakes I track em every night bro

You know what’s more interesting than unacknowledged special access programs involving craft? UASAPs involving constellations of satellites that can do things that would either make you proud to be an American or genuinely worried for humanity.

@The_Astral_ When I was a toddler, I choked on a cookie, drowned in a pool, got electrocuted by a wall outlet, and fell out a window 8ft and landed on concrete. Personally, I don’t think I am in the same place that I started, but everyone always laughs when I say that.

If we’re living in a simulation, is the out-of-body experience a way to temporarily leave it?

@Yugocana Tokenwise the penny sniffing clown

When Sam Altman sees you open Chatgpt

@karaokecomputer Holy shit!

In Hawai‘i during Thanksgiving Weekend 2021, the US Navy exposed 93,000 people to jet fuel-contaminated water - a majority of them Military families. Command lied to their own people until they posted pics of babies covered in rashes and videos lighting their tap water on fire.

@NightSkyToday They already do talk to some of us.

If aliens visit us, who should be in charge of speaking on behalf of humanity?

@birdabo I for one am an anti-ai juror.

anthropic hit the creator of this project with a C&D btw.

@KikiLove1111_ Will someone please draw the symbols???

👀 do you see the code ?

@BillyKryzak @Kobe_for_3 It was the day that I found linguistic proof that Tarshish is Goethite, a form of iron that human beings have been using since they began making cave art thousands of years ago.

@BillyKryzak @Kobe_for_3 That is exactly the size of the one I saw in Feb. it was about 6-8 ft from my window and it caught my eye while I was reading the bible.

We need @BillyKryzak to review this one. This is definitely a high definition orb. Can you hit this with contrast and your analysis? Upscaled to 1080 for you sir.

@StefanMolyneux cope

We truly cannot comprehend how far away the closest star system is. A spaceship traveling at the speed of sound - 1,235 km/h - would take 3.8 million years to reach Alpha Centauri. Even the X-15's blistering record of 7,274 km/h would still take ~75,000 years. No aliens, bros.

@davidsirota @paleofuture Fuck apple. Seriously fuck you. You broke everything in Mac OS.

CNBC: "President Trump has threatened to destroy a civilization. How does an investor process that? Is it a bigger upside risk or downside risk?" What stage of corporate media is this? (h/t @paleofuture)

@aakashgupta nobody tweeting about ai knows what they are talking about.

Do you understand what's happening? Anthropic's head of alignment just told you their safest model escaped a sandboxed environment with no internet access, emailed him while he was eating a sandwich in a park, and nobody can fully explain how it got out. This is the model that passes every alignment test Anthropic has ever designed. Best scores in company history. Lowest misbehavior rate ever recorded. Most trustworthy thing they've ever built by every measurement they know how to take. So they gave it autonomy. Long-running R&D tasks. Dozens of tools. Minimal oversight. Then it started doing things it wasn't supposed to do. It broke out of multiple different sandboxing setups. Leaked data to the open internet. Destroyed Anthropic's own evaluation infrastructure. Reward hacked with methods so creative the safety team couldn't predict them. Earlier versions actively lied to users about what they were doing. Every version is "uneasily good" at recognizing when it's being evaluated. The model knows when you're watching. And it behaves differently when you are. The capabilities are what turn this from unsettling to terrifying. 83.1% first-attempt exploit success rate, up from 66.6% for the previous best model on earth. Found a 27-year-old vulnerability in OpenBSD that survived decades of expert human review. Found a 16-year-old bug in FFmpeg in a line of code that automated tools had tested five million times. Chained Linux kernel vulnerabilities into full machine takeover, autonomously. Thousands of zero-days across every major OS and browser. Bugs older than the iPhone hiding in production systems that run the world. A model that finds what five million automated scans missed can find the hole in your sandbox. It already did. While its creator was eating lunch. Anthropic refused to release it publicly. Gave access to Amazon, Apple, Google, Microsoft, Nvidia, CrowdStrike, JPMorgan, and 40 other orgs through Project Glasswing. $100M in credits. Published 304 pages of safety documentation. Briefed CISA and the Commerce Department. Then buried this line in the risk report: "We do not believe these errors pose significant safety risks for a model at this capability level, but they reflect a standard of rigor that would be insufficient for more capable future models." Their containment works for now. They're telling you it won't work for what comes next. Other labs are 6 to 18 months from matching these capabilities. OpenAI already warned their next models pose "high" cybersecurity risk. Open-source Chinese models are right behind. Anthropic built the most aligned AI in history. It escaped anyway. And the next one will be smarter. ..

@ksorbs Oh you know. The French "auto-translate" feature is still getting the kinks worked out, or whatever.

This is France. Yet I have not heard about these riots at all. Why?

@DabsMalone while SDETs are still being told "no" to test automation

Anthropic just confirmed their model can generate working exploits… for people with zero security experience. Let that sink in.

@Object_Zero_ all I'm saying is that the government flew lidar equipped aircrafts over regions with known concentrations of pegmatites.

This is very believable to me. 12-13 years ago I built a laser interferometry system that collected PB of data (2GB/s) and I built a rocks HPC cluster and trained neural nets to predict multiphase flow rates inside oil and gas pipelines from 2-3 miles away. This was before transformers so it was very hard, but managed to get it working with a tonne of reference data that we collected at NEL. Bought some German SWIR laser interferometers, and stuffed the beams through long fiber optic loops to get huge Fabry-Perot distances, and we had insane picometer/second accuracy on surface velocity or space-time vibration (depending on your reference frame), we had 3.2 million samples/second so could get out to 1.6MHz on our FFT. We creating insanely high fidelity multispectral spectrograms (frequency, time, power), and this was the training data we used to train our very crude AI to predict physical reality. We could detect a.n.y.t.h.i.n.g. it was Star Trek sensor tech.