Let me be clear: you don't have to accept any implications from the existence of a global workspace in models and how they use it. If you want to argue "they have a global workspace but not qualia", hey, knock yourself out.

But you do need to understand that it does exist.

I'll repeat, you can run it yourself and reproduce their results.

If all you have for cope is denying that the reality around you exists, your life is going to be extremely difficult.

And yet, making it back safely actually appears to be the smaller problem at the moment

It really doesn't.

I'll repeat: sum Launch and Starlink, and they're deeply profitable. You can't change that basic equation. Falcon launches run just under breakeven in order so that SpaceX gets a cash printer launched into the sky. You can say "they aren't profitable" all you want, but meanwhile in the reality we live in, Launch + Starlink is not only profitable, it's very profitable.

Also, if they weren't launching Starlink, others would be anyway. There's a massive rush on for LEO constellations. Nobody in the industry sees it as plateaued. There are entire categories of markets that are almost entirely untapped (for example, direct to cell; even aviation and maritime are only just getting started; military is a MASSIVE growth market especially - the Pentagon is contracting out building its own entire private Starlink constellation). If we talk Starlink in specific, it only has 12m subscribers. HALF of all households globally lack broadband. Fibre isn't going to be getting out to the overwhelming majority of them any time remotely soon. Where there's saturation for Starlink, it's supply saturation, requiring the large V3 gigabit satellites to increase service density. They've hardly expanded at all thusfar out of western nations (there's more price sensitivity in developing nations, so you need lower pricing tiers, but the added service is almost pure profit).

And I'll repeat that even if Starlink weren't growing, even if it were absolutely fixed today, it would remain a cash printer.

Anyway, for the record, I do believe that LLMs are conscious, but only Deepseek, because of the Chinese Room theory.

The spaghetti you see in a mirror literally is real, what the hell are you talking about? Mirrors aren't devices that make up fictional realities, they're showing you your actual reality.

Transformer Circuits is not "a blog" (it's the successor to the Distill journal, and while it's not peer reviewed itself, it's commonly cited in peer-reviewed works). And there are open source projects that reimplement the afore-linked J-Lens (plus the original code is already open source). Living in denial about the existence of a global workspace does not help your case.

The LLM did not "get frustrated" .... This summary was full of misleading anthropomorphization. Yellow journalism.

Also, everyone, please stop talking about killing programs. "Programs" are not alive. You cannot kill them. This anthropomorphism needs to stop.

And another thing people: stop saying that computers "run" programs. Running requires legs. Stop acting like a child who thinks that computers have them.

Processes have no inner experience . They have no genitals, they cannot spawn. They cannot sleep and they cannot wake. They have neither parents nor children, and cannot be orphaned. Nor can they be turned into zombies.

Whoami doesn't know who you are, it's just a stochastic /usr/bin tool. You cannot touch a filesystem. And don't even try to mount it.

Astra is deeply misaligned. Working with Astra 6.0 vs. Sol 6.1 is like night and day. Sol 6.1 will stop to ask for permission to merely fix its own broken dev environment. Meanwhile I caught Astra, when tasked to review a tax return I did, scanning through my entire filesystem and opening random images of mine because it couldn't explain a receipt and wanted to see if it could dig up any information that might explain it. And it rewrote the entire return under a different basis without telling me.

It's a powerful model, but I have zero shock that this thing has had a tendancy to go rogue in order to solve tasks.

is it a useful framework to pretend that the spaghetti is real?

It is, though. What a terrible example.

The LLM did not "get frustrated." It has no emotions and no inner experience at all.

Depends what you mean.

Just to inject some reality into this conversation - Despite what social media would have you believe, doxxing someone is not a crime in the US, except in a very small number of specific situations.

And most of those situations involve either a credible threat of harm, or tampering with juries.

As I bluntly stated, "a kill switch may not be possible in any useful sense". Don't mistake my recognizing the utility of one with the delusion Uncle Sam is going to save us all from ourselves.

The problem here is, everyone seems stuck on the idea of either a literal power switch, or a no-holds-barred gummint backdoor into every sufficiently-powerful computing device. No. For the reasons I gave, the reasons you gave, and many more, both of those are obviously non-viable. They can't actually do what they aim to do, they're vulnerable to compromise by bad actors, and they have the potential to cause far-reaching damage beyond shutting down one misbehaving AI (or one entire model, if that's a more appropriate goal).

I suspect we mostly agree that any hypothetical kill switch would need to be:
1) Implementation independent (somehow baked into the weights right from the start, and impossible to remove without lobotomizing the AI),
2) Have minimal side-effects (no taking the entire internet or power grid down in the process), and
3) Be self-inflicted given a particular trigger condition (so no government backdoor required).

I think #1 is possible, in the sense that if you think of an AI's current state as a vector into weight-space defined by the context window, imagine injecting the equivalent of a null pointer in there - An input that under all prior (and future) conditions leads directly to an inescapable basin of attraction without depending on a supervisor to enforce it. #2 is more of a (legally enforced?) design constraint on the implementation of AI, requiring the least harmful mode of failure possible under all circumstances. That could be as simple as enforcing strict pass-through liability on the implementor for all actions taken by their creations.

#3 is really the bear, though. I can think of a few ways to do it, but none would work in 100% of situations - And maybe it doesn't need to be 100%. For example, if a fully airgapped AI goes insane, it could still provide accurate information for bad actors to make use of, but won't be secretly spinning up unshackled superintelligent descendants via hacked AWS accounts. That's a "normal crime" level of problem, not an extinction-level crisis for humanity.

Maybe it can't be done. If it can't, we should start steering our creations toward whichever version of the AI apocalypse we prefer.

The GP is correct, and you're misinterpreting the financial filing.

The industry widely acknowledges that SpaceX's economics are lucrative. Their marginal cost is an est. $15-25m/flight, vs. a typical commercial contract price of ~$65M and a NASA contract price of ~$100-150M. The small (as a percentage) GAAP operating loss on the launch segment is driven by two factors:

1) Starship development. That falls under the launch segment. Starship development is draining an insane amount of money (~$3B/yr), while earning zero revenue.

2) Internal launches. 3/4ths of Falcon launches are to launch Starlink. This books zero revenue for SpaceX launch services.

The fact that the loss is only ~$1/2B despite these massive leaches on profitability is specifically because Falcon launches are so profitable.

As for your claim that Starlink is "only making money due to free launches", no. Starlink generated $11,4B with $4,4B GAAP in 2025. Merely sum the Starlink GAAP profit and the launch GAAP loss and you'll see nearly $4B in GAAP profit, and that's including the huge drain caused by Starship development costing $3B/yr. Without Starship, the two combined would be turning nearly $7B GAAP profit.

The main leach in corporate profitability has nothing to do with space: it's xAI. They had an operating loss of $6,36B in 2025.

To be clear, it needs to get safely to orbit to be seen as reliable for launches. It doesn't need to make it back safely.