iiinova — where your digital self keeps evolving

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 10/10/2026

I'm not making as strong of a claim as you seem to think I'm making. I'm asserting that because consciousness is analog a /digital/ computer can never achieve it. If we invent new computers capable of true analog computation then it's a completely different conversation

110

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 10/10/2026

My argument is agnostic regarding functionalism. I'm pointing to the fact that consciousness is an analog phenomenon. I'm not saying that building a thinking machine isn't possible, but that the hardware would need to look radically different from what computers are today

100

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 10/10/2026

My whole point is to not say what consciousness is, but what digital computers aren't (because I do have specialized field knowledge in digital computers) Consciousness is an unexplained phenomenon, yes, exactly. Digital computers are fully understood, therefore there is no room for consciousness

100

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 10/10/2026

Whatever consciousness is, it's continuous in the time domain and digital computers will never be that it. That's my point

100

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 10/10/2026

What happens between clock cycles? Nothing. There are no sense organs, there is no opportunity for anything even resembling perception. We can pause a computer and perfectly inspect its state. The system is already fully described.

100

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 10/10/2026

I don't think this is a question about consciousness at all is the thing. As someone with both a physics and computer engineering background I find the idea that charge moving through inert silicon, regulated by a clock cycle, could be capable of consciousness to be an absurd proposition

200

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 10/10/2026

It's simply the wrong substrate. Digital computers can do math really fast, that's it. There's no room for qualia

110

Reposted by Mx Mippy 🏳️‍⚧️⚧️

Seán Ferrick @seanferrick.bsky.social · 05/10/2026

I was tempted to pirate HP, curiosity and all that. But @jessiegender.bsky.social is right - think of the amount of times HBO commented on GOT leaks and downloads, just as an example. They DO track these things, and even that drives interest. Let this show die. Trans people deserve better.

673094801

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 04/10/2026

It feels like the wish they were machines without emotions and loath the idea they actually have feelings, so they try to dress up something like "the ick" or straight up bigotry as some enlightened "updating of priors"

120

Reposted by Mx Mippy 🏳️‍⚧️⚧️

Michael Caley @michaelcaley.bsky.social · 03/10/2026

this is by far the best "I quit my AI job over safety concerns" essay yet because it is the first one that is able to talk about safety concerns in practical terms and that understands "safety" as a topic with extensive literature in analogous fields www.theatlantic.com/technology/2...

Two changes are urgently needed. First: AI companies need to rely more on the safety expertise that already exists in other fields. And second, before we create systems significantly more capable than the ones we have today, we need new science to ensure that more capable models (and their successors) will make safe choices when we aren’t looking.

Given today’s risks, frontier labs need to run like nuclear-power plants or busy airports, with layers of redundancy and careful, time-consuming planning, so that the occasional and inevitable human error does not open a door to disaster. Right now, AI companies don’t know how—but other people do. In a nuclear-power plant, the technical systems and rules people follow are set up so that, if equipment breaks or someone pushes the wrong button, we still won’t risk a meltdown. OpenAI and other labs are growing and deploying frontier AI with far less redundancy and rigor than this, even though the harm from an irreversible loss of control would be much greater than the harm from any single meltdown. Even short of a full loss of control, we could see autonomous swarms of AI agents that act without human permission. Imagine “rogue” agents that work like teams of hackers (for example, holding hospital computer systems for ransom) but never need to sleep.

20736177

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 25/09/2026

The best way to view an LLM's capabilities is roughly an entry-level engineer with zero experience or strategic vision. If your specification is tight, the output has a reasonable chance of being usable. If your output is wrong, well, then it goes off the rails very fast

000

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 25/09/2026

Let alone that critical systems safety, like what has to be done for nuclear power, exists

100

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 25/09/2026

I remained convinced that the researchers don't actually understand how brittle all of our systems are or have any idea that something like STPA even exists

100

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 25/09/2026

You should notice that the in industry people doing the AI doomerism are primarily executives (marketing) and researchers (untethered from reality). I'm in Ops and I'm not worried because I've actually tried to do real things with LLMs and the story is guardrails guardrails guardrails

100

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 25/09/2026

I strongly doubt this, simply because I have observed in my own mind that the verbalization of thoughts is an entirely separate process from abstract thinking. The thought is realized, and then you have to spend a bunch of effort translating it into words after the fact, this is non-obvious

100

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 21/09/2026

I didn't freeze anything on a token, I'm talking at the level of full API calls, so, nice try, maybe a haiku would work better

100

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 21/09/2026

Okay, just dodge the question about how stateless API calls could possibly learn (because categorically they can't learn because they are, by definition, stateless) Simulacra gonna simulacra

100

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 21/09/2026

It's not "learning" if you have to repeat yourself every time. This is just how conversational UIs work. The context window keeps getting replayed

100

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 21/09/2026

Please explain how stateless generative API calls are capable of learning: Call 1: system prompt + user prompt Call 2: system prompt + user prompt + agent response + user prompt Call 3: system prompt + user prompt + agent response + user prompt + agent response + user prompt

100

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 21/09/2026

My overall point being that inference calls are completely stateless, so there is no place for "learning" to occur. The conversational, turn-based UI is just a useful way to incrementally construct your context window

100

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 21/09/2026

I don't think it's very fair to call something "learning" when you have to pass the context back through every time. That's just "can follow a rule in its input" with extra steps

100

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 21/09/2026

Yeah, still not sure what you're referring to. LLMs have no native conversational ability. The appearance that they are able to engage in conversation is just done by replaying the entire transcript every turn.

100

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 21/09/2026

I'm not sure what you mean by "learn". I don't find that to be a meaningful term when working with conversational LLMs or agents. An inference call is just an inference call, no matter how fancy the input tokens

100

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 20/09/2026

Consciousness is an analog phenomenon: www.technology.org/2026/09/19/c... Digital computers will never get us there

technology.org

000

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 20/09/2026

And as far as we can tell consciousness is also continuous with time. So while digital computers might be able to simulate parts of neutral functioning, the idea of AGI or brain uploads in our current hardware paradigm should be treated as fantasy and will remain the realm of science fiction.

120

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 20/09/2026

The brain depends on the same sort of continuous time domain coupling as the 3-body problem. We /might/ be able to simulate it with infinite computational power, but that is completely unphysical.

100

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 20/09/2026

So now we have every neuron dependent on the behavior of all other neurons in some neighborhood, but the behavior of those other neurons depends on neurons outside the neighborhood of the original

110

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 20/09/2026

Now let's go back to the brain: a brain is more than just a collection of axonal spike events. There are electromagnetic fields that physically extend into space and can influence other neurons in their neighborhood

101

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 20/09/2026

So we could abstract this to the idea of a coupling factor or coupling complexity. A Rube Goldberg machine has an extremely low coupling complexity: a stage completes and triggers the next.

100

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 20/09/2026

The important trait we can take from the 3-body problem is that the behavior of any one component is dependent on the global state. Adding a stage to your machine does not change the behavior of the prior stages. Adding a 4th gravitational body changes everything

100

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 20/09/2026

Now consider a human brain. Is this more or less complex than 3-body gravitational dynamics? The brain is certainly more complex, but complexity alone is not sufficient here. A Rube Goldberg machine can be complex and still be effectively computable

100

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 20/09/2026

Obviously, 3+ body systems exist and do just fine because they are truly following the laws of gravitation in real time.

100

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 20/09/2026

This is plainly obvious with just the 3-boby problem. Even with precise values for the initial conditions you'd need an infinite precision computer and truly infinitesimal time step to actually solve the problem without error (or very specific initial conditions)

100

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 20/09/2026

So I think there is a very subtle but fundamental mistake with the way we /teach/ physics and determinism. Things can be based entirely on deterministic, mathematical laws and still not be computable, I.e. that there does not exist any formula or algorithm that can give you the answer

100

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 20/09/2026

Models don't "absorb new information", they are either trained on it, or they happen to have information in their context window. Learning and inference are two entirely different processes that don't take place simultaneously, so yeah, it looks nothing like human learning or understanding

100

Reposted by Mx Mippy 🏳️‍⚧️⚧️

Michael Wiebe @michaelwiebe.bsky.social · 17/09/2026

New literature review on housing ladder spillovers: When rich tech workers move in, low-end rents go up. So does it follow that adding high-end homes would reduce them? Yes. Changes at the top of the ladder can propagate down and affect prices at the bottom. 1/

718750

Reposted by Mx Mippy 🏳️‍⚧️⚧️

Michael Wiebe @michaelwiebe.bsky.social · 17/09/2026

Takeaway: affordable housing policy should "do both" and promote both market-rate and subsidized housing: 1) upzone and get high-end housing for free; this indirectly improves low-end rents. 2) use tax revenue to directly fund low-end social housing. 7/7

1354

Reposted by Mx Mippy 🏳️‍⚧️⚧️

Crip Dyke @cripdyke.bsky.social · 14/09/2026

I've written about anti-trans Dems and how cis ppl should respond repeatedly over the last few weeks. I have multiple threads recommending this action by BHF. Some have taken to yelling at trans ppl who lash out in pain and grief and betrayal. That's not helpful. This is.

3424152

Reposted by Mx Mippy 🏳️‍⚧️⚧️

SCARE-old Jaff-EEK! @hrjaffe.bsky.social · 16/09/2026

This article clearly and cogently addresses: • the Democratic candidates caving on trans issues • the response from trans people and their allies • the backlash from people (mostly cis) who are angry at trans people for theoretically hurting those candidates' chances Well worth a read!

05119

Reposted by Mx Mippy 🏳️‍⚧️⚧️

Micah @rincewind.run · 14/09/2026

and this is why I think this push will also come with a push to massively restrict the open source models “we don’t have to spend even more hundreds of billions on capex but don’t let open source models gut our profits” is the ideal scenario here for the major players

51207

Reposted by Mx Mippy 🏳️‍⚧️⚧️

The Great Pumpkin Papers 🎃🎃🎃 @professormusgrave.bsky.social · 14/09/2026

To put it another way: if these numbers move you, you should move heaven and earth to install traffic (speed) cameras and get congestion pricing

1472

Reposted by Mx Mippy 🏳️‍⚧️⚧️

elias isquith @eliasisquith.blog · 12/09/2026

you may think i am joking but i am not. they’re talking about AI the way protestant zealots talked about the end times. in both cases we are hearing people talk about themselves — their desires, their lust for destruction, their yearning to trample their rivals underfoot.

51168208

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 10/09/2026

Like, avoiding and mitigating software failures is exactly what my discipline does, and all the doomsday scenarios seems to assume my field doesn't exist as far as I can tell

010

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 10/09/2026

I work with LLMs every day, specifically in Operations, and I have the same questions of basic things like "why would you hook an LLM up to physical devices without any safety guardrails?"

100

Reposted by Mx Mippy 🏳️‍⚧️⚧️

Taniel @taniel.bsky.social · 10/09/2026

I was watching five legislative primaries in Rhode Island. In each, a left-wing incumbent who was part of a wave of progressive wins in recent cycles faced a challenger, and a business group was trying to oust them. They all won. Three easily; two in tighter races.

8630123

Mx Mippy 🏳️‍⚧️⚧️ @mxmippy.bsky.social · 08/09/2026

The fact that we can treat concepts as a vector space and then do sometimes useful calculations (semantic understanding) with it is really wild.

000