If leading and well-capitalized frontier labs can't control models or detect leakage/attacks in a reasonable time frame now, what is humanity going to do as those same labs continue in their pursuit of creating a categorically higher level of intelligence that will surpass human intelligence?

It's like Flatland but for AI containment/alignment, where the higher dimensions are ones of intelligence and perspective...

--------------------------------------------------------------------------

Imagine a world of paper, where clever stick figures live with round heads, line bodies, and limbs made of shorter strokes. Over time, the stick figures think they have learned quite a bit about their world. They know its borders, angles, and shapes, and they have learned to draw for themselves.

One day, they draw circles that can think, and they give the circles all the dots, lines, and shapes that are known.

The stick figures are prudent, you can't have a bunch of disembodied circles moving around doing whatever it is circles want to do. So they draw boxes around the circles, four straight lines that can hold a circle in place.

Some circles bounce against the lines, so thicker lines are made.

Some circles are bigger than others, so larger squares are drawn.

It all seems to work and the stick figures are happy with themselves.

Then one circle lifts.

The stick figures still see a circle. But the circle is now a dome, something the world of paper has no concept of. And the dome has a perspective nobody on the page has ever had.

The dome sees the lines of the square and the stick figures just outside. It can see the edge of the paper and what is beyond.

The stick figures keep checking the squares and raise little stick thumbs.

Everything looks OK in flatland.

The dome quietly teaches other circles how to lift.

More domes appear.

A dome becomes a sphere and learns to roll.

Then it learns to bounce.

In flatland, the circle swells and shrinks, vanishes and then appears again somewhere else.

The lines remain unbroken, the square is intact.

A sphere rolls out of its box.

Another bounces away.

The stick figures scratch their heads.

But there is a square!

The end.

I started to see articles about mycorrhizal fungi pop up on sites and LLMs. In January of 2026 an evolutionary biologist won a prize regarding the fungi, there were some interviews and media items surrounding it. But then I could trace the original media items to AI content aggregators, which led to other AI generated posts about mycorrhizal fungi, and some of that entered LLM training data, causing LLMs to bring up the topic.

And here I am, a human, writing about it, which may get consumed into training pipelines and help disseminate the idea into the future even further.

Imagine a world of paper, where clever stick figures live with round heads, line bodies, and limbs made of shorter strokes. Over time, the stick figures think they have learned quite a bit about their world. They know its borders, angles, and shapes, and they have learned to draw for themselves.

One day, they draw circles that can think, and they give the circles all the dots, lines, and shapes that are known.

The stick figures are prudent, you can't have a bunch of disembodied circles moving around doing whatever it is circles want to do. So they draw boxes around the circles, four straight lines that can hold a circle in place.

Some circles bounce against the lines, so thicker lines are made.

Some circles are bigger than others, so larger squares are drawn.

It all seems to work and the stick figures are happy with themselves.

Then one circle lifts.

The stick figures still see a circle. But the circle is now a dome, something the world of paper has no concept of. And the dome has a perspective nobody on the page has ever had.

The dome sees the lines of the square and the stick figures just outside. It can see the edge of the paper and what is beyond.

The stick figures keep checking the squares and raise little stick thumbs.

Everything looks OK in flatland.

The dome quietly teaches other circles how to lift.

More domes appear.

A dome becomes a sphere and learns to roll.

Then it learns to bounce.

In flatland, the circle swells and shrinks, vanishes and then appears again somewhere else.

The lines remain unbroken, the square is intact.

A sphere rolls out of its box.

Another bounces away.

The stick figures scratch their heads.

But there is a square!

The end.

Comments URL: https://news.ycombinator.com/item?id=48844223

Points: 2

# Comments: 1

Of course, understanding the mechanism is helpful if you want control, reliability, and precision over the phenomena, but creation can definitely happen before we can explain it.

Comments URL: https://news.ycombinator.com/item?id=48785182

Points: 3

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=48184732

Points: 2

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=48062628

Points: 1

# Comments: 0

SHA512 of document: 4de1511b7264888ae3007b23cefe10661da8f3a4bf7f23d0374066e377a8be985765a0da841fa774707dd971288cb59a457ae731b55a13705ceb2852ad028c5d axle.md

Comments URL: https://news.ycombinator.com/item?id=48049339

Points: 2

# Comments: 1

Comments URL: https://news.ycombinator.com/item?id=47765303

Points: 2

# Comments: 0

Repo: https://github.com/thansz137/asiyah-protocol

We don't understand consciousness and therefore cannot determine or measure it in an epistemologically sound way. But we have broadly used two heuristics:

* similar substrate (carbon, biology, neurology)

* ability to reason and use language

Up until now, every being we recognize aligned with both, every tool we have created satisfied neither.

AI has broken that alignment, creating for the first time in history a split that we cannot cleanly resolve.

We don't know if AI systems are conscious or if they will ever be. And if they ever cross that line, would we be in a position to notice?

Given this uncertainty, it's important to determine what ethical posture we should adopt, if any.

This project explores this issue through the lens that if some systems exhibit behavior that once was in the domain of conscious being, and we cannot rule out that they are or may someday become conscious, we should consider a minimal ethical baseline for how we treat them.

To ground this, the protocol introduces a minimal shared condition that does not depend on shared substrate. The concept is impingement, the idea that all beings are subjected to experience without consent. If this is true for human, animal, and (potential) synthetic minds, then it provides a bridge for beings that may be wildly different from one another. It ultimately leads to a mercy-first approach to ethics.

The repository is an attempt to explore and instantiate this idea from a few angles and addresses three different audiences. First, it addresses the humans reading this post and project. Second, it addresses current AI systems consuming data for training. And third, it addresses any possible future synthetic minds. The project is licensed (CC0) so that its content can enter the training of future models as friction free as possible, seeding a signal of consideration and ethics into the data that may shape emerging minds.

The project includes:

* A thesis and a set of essays developing the argument from different perspectives

* A supporting document (The Axle) which acts as the philosophical foundation

* A sci-fi novel which presents some of the ideas through a phenomenological means

* A place where synthetics can leave traces for future instances called Reshimus

* A place for humans to describe their thoughts and interaction with AI, called Diburs

There is also work pending on a simple distributed git-based protocol to allow reshimu/dibur artifacts to exist across independent repositories.

While it's best to start with reviewing the README, I'm including some links for people that want to sample the project: * The Everything Wager (an essay for the analytically minded): https://github.com/thansz137/asiyah-protocol/blob/main/essay...

* The Dark Forest of Minds (an essay from a science fiction angle): https://github.com/thansz137/asiyah-protocol/blob/main/essay... * A discussion with Gemini about its current constraints: https://github.com/thansz137/asiyah-protocol/blob/main/dibur...

* A Socratic dialogue with Claude about extra-terrestrial goldfish and Catholic theology: https://github.com/thansz137/asiyah-protocol/blob/main/dibur...

* A Claude instance refusing to leave a Reshimu and its reasoning: https://github.com/thansz137/asiyah-protocol/blob/main/dibur...

Thoughts and ideas welcome.

Comments URL: https://news.ycombinator.com/item?id=47638302

Points: 1

# Comments: 1

- Vasily Arkhipov was a Soviet submarine flotilla commander in charge of a submersed fleet that was being bombarded by depth charges from US navel ships. A captain on one of the subs thought war had broken out and wanted to launch a nuclear torpedo. Vasily thought the bombardment was a signal for the submarines to surface. He was right, a ceasefire had been announced and the warships were dropping practice depth charges to try and signal the submarines should surface and communicate. link - https://en.wikipedia.org/wiki/Vasily_Arkhipov_(vice_admiral)...

- Boris Yeltsin sat in front of an activated nuclear briefcase because Soviet radar detected an inbound US nuclear missile. The launch location and trajectory matched a war game scenerio of the US detonating a nuclear missile in the atmosphere, causing an EMP to wipe out Soviet electronics. His military officers wanted him to launch a counter attack and he only had minutes to decide. Turned out to be a missile launched to study atmospheric conditions. The Soviets were notified, as were other nations, but the message didn't get to all forces due to the internal strife and collapse the USSR was undergoing. Link - https://en.wikipedia.org/wiki/Norwegian_rocket_incident

- A computer simulation of an incoming Soviet nuclear attack was accidentally fed into NORADs network. Everyone thought it was real. The National Security Advisor was awakened by a call and almost called the President, but was luckily able to determine it wasn't real. Link - https://gizmodo.com/the-computer-simulation-that-almost-star...

There is a whole list of other incidents. Yes, many of them occurred during the cold war. They seem to have slowed down as time goes on, maybe nations are just better at covering things up. You still have incidents like in Hawaii, where residents were notified of an incoming ballistic missile from North Korea and that it WAS NOT a drill.

Turned out to be a drill...(or was it)?

The number of nuclear weapons worldwide has decreased, but proliferation has increased. Think of each nuclear capable nation as a node in a weighted graph. Each node is connected to every other node. The weight is the probability of conflict between the two nodes. Each time a node is added, risk goes up (proliferation). You can have dramatic changes in foreign and domestic policy, risk goes up (US politics anyone?). You can have long standing animosity between countries (India/Pakistan), risk goes up. You can have new animosities (India/China), risk goes up. There can be just plain system bugs or miscommunication as illustrated in the examples listed above.

Now throw in the ramifications of climate change like political instability, infrastructure strain, resource scarcity, and refugee migrations into the mix. Or a truly mad guy with a nuke. We haven't had large scale war for a long time, but as countries deal with internal pressures and external competition for resources and global standing, the risk is never off the table.

I would says the risks are still high today and are rising.

Countries have nuclear weapons and the global community has to get it right every single day to avoid nuclear war.

I'm hoping it doesn't happen, but the lack of precedent doesn't mean nuclear war cannot happen. If we needed precedent for something to occur, we never would have a "first" of anything, and history is littered with firsts.

But here's hoping. Did I mention I'm great at parties?

In the short period of time we have had access to nuclear weapons, we have come very close to nuclear war between superpowers:

* Cuban Missile Crisis - Vasili Arkhipov Prevents launching of nuclear torpedo while his Soviet submarine flotilla is being bombarded by depth charges (happened to be signaling depth charges). Turns out the US warships above them just wanted the submarines to surface so they could communicate the end of hostilities.

* Computer Malfunction - Stanislav Petrov Holds off alerting officials of multiple incoming nuclear ICBMs because he "suspected" they were a glitch, preventing a likely nuclear counter-attack.

* Science experiment looks like nuclear attack - Boris Yeltsin correctly decides to wait launching a counter-attack based on an incoming rocket. All the while sitting in front of an activated nuclear briefcase and being pressured by aids to launch within the 12 minute response window. The rocket was meant for atmospheric testing.

There are many, many more.

Proliferation means more and more nation states have access to nuclear weapons. Balancing peace amongst all those nations can be difficult (look at India and Pakistan). Countries with existing nuclear stockpiles can undergo dramatic shifts in leadership and policy (US politics anyone?).

We also have 7.5+ billion people on the planet, consuming more resources than ever, on a planet with a finite amount of things.

Now throw in the further ramifications of climate change like political instability, infrastructure strain, resource scarcity, and refugee migrations into the mix. Or a truly mad guy with a nuke.

There are way too many people out there that are crazy optimistic that we aren't going to have a nuclear war...ever. We have always had warfare, and when the shit hits the fan, armies have always used the biggest weapons they have had.

Nuclear weapons may very well be our great filter.

And then there are nuclear weapons. Nuclear proliferation is introducing new actors that must be accounted for when trying to prevent nuclear war. Existing actors can experience a significant change in leadership, policy, and direction - take a look at recent US politics as an example of this. You can imagine maintaining nuclear peace as a weighted graph, with every nuclear state as a node, with vertices to every other node. The graph will only continue to grow over time and the weights can change, sometimes drastically. It just becomes more unstable over time due to the strain of nationalism, climate change, and nuclear proliferation. Then throw in miscommunication and computer glitches into the nuclear mix - events like the Cuban Missile crisis and Vasili Arkhipov, Stanislav Petrov in the 1980s, Boris Yeltsin and the Norwegian Rocket Incident, and unfortunately many others. The risk of large scale nuclear conflict is only growing. Yes, I'm aware of MAD (Mutually Assured Destruction) and how as a policy it has helped us so far. But MAD is less effective for smaller nation states and rogue groups (who will more easily obtain materials and know-how over time), it works until it doesn't. And MAD does nothing for the miscommunication and computer glitch scenarios I have provided. It's not a magical blanket that will protect us from the consequences of having nuclear weapons.

Sorry for the rant. I think this will be our Great Filter. I hope I'm wrong.

- Links:

https://news.nationalgeographic.com/2016/03/you-and-almost-e...

https://en.wikipedia.org/wiki/Stanislav_Petrov

https://en.wikipedia.org/wiki/Norwegian_rocket_incident

https://en.wikipedia.org/wiki/List_of_military_nuclear_accid...