snthpy

Lol is this some kind of meta joke?

"I can calculate the movement of the stars, but not the madness of the people."

Isaac Newton

Exactly!

Why do we view using the level of abstraction immediately below as cheating while the levels below that are taken for granted?

Why do we view using the level of abstraction immediately below as cheating while the levels below that are taken for granted?

Indeed. It didn't write it's own TCP implementation either. Nor operating system, nor compiler, nor build its own wafer etcher...

My comment probably sounds facetious but I'm actually intrigued by why we seem to view using the level immediately underneath as cheating whereas the levels below that are taken for granted?

Smarmy and weird was probably just the vibe i was giving off. Incidentally, I've always found Claude easier to talk to than most people lol

Hey, when i was in my twenties i told every girl that i was "pacing the frontier"... The right pace makes the denouement just that much sweeter.

What?

Thanks for pointing this out. I'm only reading these comments because i liked the interaction combinator Bend language.

Pivots are cool but why reuse the name and cause confusion? What is the old Bend called now?

DSPy seems like the right comparison and this is the first comment I've seen mentioning it.

Thanks for putting this together. I'm surprised the cost saving is so little though. I expected much more based on the post.

tl;dr templating language/tool for markdown from Obsidian

Looks useful!

I was hoping this would be about a new way to augur the events of the day from my morning toast to ensemble with the tea leaves at the bottom of my morning cuppa, alas ...

Thank you for this. Knowing both those words, i could never work out the original/history of the terms.

ICQ was my first messaging experience. Good old times!

Me too!

I agree with you on the markets but not on institutions and organisations.

Organisations pursue goals with agency and there is even Darwinian competition between them in many cases.

I kinda agree that the financial system is intelligent in the problem solving sense but disagree that it is pursuing a particular goal in the sense that institutions and organisations are.

It's more akin to an ecosystem to me. In nature we also have ecosystems organising and reusing resources but I would only ascribe intelligence to living organisms pursuing individual goals with agency.

Haven't thought too hard about the distinction though so happy to be shown a better principle.

Thanks for the reply. I was hoping to get some debate going and have my thoughts challenged in order to see how they hold up under scrutiny. You're right that I believe in Functionalism and Strong AI and in previous comment I only really stated my belief so please allow me to try and make a bit of an argument for it. My position probably falls under the "System reply" to the Chinese Room argument although I feel it's more restrictive in some aspects so perhaps it corresponds to a particular subtype (I don't know as I haven't read enough in the area).

My basic argument is that human *intelligence* (and understanding - I consider understanding as part of intelligence and consciousness separately but I can see how many might group understanding with consciousness and leave intelligence on its own) is a form of collective intelligence. Individual neurons acting based on local chemical and electrical signals collectively produce behaviour that we call intelligent, i.e. it can solve problems, respond to questions, etc ... . A particular configuration of neurons acting together I will call a particular software or algorithm (I understand that separating the software from the neural substrate in that way is part of what's in dispute, particularly by Searle, but that's my belief and what I feel is required to make the mapping to the Chinese Room argument). The Chinese Room thought experiment then is trying to refute Strong AI by arguing that the person in the Chinese Room has no understanding of the exchange with the outside world because they are just manipulating symbols incomprehensible to them. I agree that the person has no understanding. That person on their own would not be able to string together a sensible reply without the instruction cards. However another person with the same instruction cards, i.e. algorithm, would be able to reply sensibly, pass a Chinese Turing test, etc ..., thus demonstrating that the intelligence/understanding resides in the algorithm. The person is just an execution substrate. The mistake in the Chinese Room argument is identifying the locus of understanding in the person, which imho is failure of imagination. It's akin to saying that humans can have no understanding because each individual neuron has no understanding of our actions, something which the existence proof of human intelligence refutes.

What then of consciousness? Here I also have some beliefs which I'm happy to write down but in my assessment they're less substantiated so make of them what you will. I saw some things by Joscha Bach recently and I liked what he said about agents and consciousness as a form self-recognition of the agent as an entity in the environment and predictive modeling of future states of that entity in order to achieve the agents goals and realise some form of agency. I don't know where neuroscience and cognitive science stand on this but made sense to me as a sensible model of consciousness. Therefore I would even go so far as to say that I believe that entities such as the United States of America (and any country or institution really) are probably conscious in some form, because they are aware of themselves, model themselves, and try to project their agency in the environment that they find themselves in. We are not directly aware of this because we are like single neurons in our brain and this happens at a scale much larger than us. Moreover I suspect that as you go up each rung on the emergence ladder, the temporal scale of the phenomena gets an order slower as that level is composed of the interactions at the level below it. So the speed of interactions of our neurons is an order faster than our conscious processes and similarly the emergent bodies and institutions that we form part of act an order slower so their processes are more difficult for us to perceive.

Schematically then my beliefs organise roughly as follows:

- intelligence = algorithm: think neural network weights you can download of HuggingFace

- consciousness = self-aware world model + state: I speculate that the same NN world model weights running with separate state would perceive themselves as separate entities

Would love to hear some thoughts on this!

Cool, will check it out.

Yeah, wasn't claiming that it was particularly novel or divine insight, otherwise I might have tried to publish a paper about it. I was just recounting my experience on this matter.

Really? I'd love to read some write ups on this because I'm still doing it the old way. I was going to say that corporate experience runs counter to this but maybe not; you think the architect is more important than the developer, but actually there have been shifts over time to recognise ICs more and keep them in their roles while managers are more generic. Interesting food for thought!

Very true!

Also the accountability problem! Sometimes we try to hold the corporation accountable and sometimes the individuals. It never seems adequate though.

I found the documentary "The Corporation" (2003) very instructive in that regard, particularly the segment with the BP CEO where the point is made that a company made up ethical individuals can still act unethically, or unaligned to human values, just by virtue of the incentive and legal structure governing the corporation.

Agreed. Michael Levin's work and discoveries are mind blowing! I was actually referencing that with the "electrical signals" nod (as well as neuronal activity) and Alan Turing's reaction-diffusion theory of morphogenesis with the chemical gradients. I'm not a molecular biologist so i don't know anything about the details, but the emergence of global phenomena from these local interactions fascinates me.

A related point that I've been making privately for at least ten years to anyone that would listen, is that we've had Superhuman Intelligence for at least hundreds, if not thousands, of years in the form of corporations and institutions.

No single human could design a rocket to go to the moon or keep all Wallmart stores supplied on a daily basis. Moreover when something like the CEO or US president changes, the policies and actions change a bit but the majority actually stays the same. This is collective intelligence embodied in the organization and its institutional memory. This is also what i believe is wrong with Searle's Chinese Room argument - the algorithm is the intelligence.

Our cells and neurons act based on local chemical gradients and electrical signals and we wouldn't call any single one of those particularly intelligent. Yet collectively they produce human intelligence and even what we call consciousness!

Looks good , but I don't understand the licensing. The bottom of the read me says it's Apache 2.0, except for the /ee folder, but I don't see the /ee folder in the repo, at least not at the top-level. I also don't see anything about what the enterprise features are on the actual web page.

Oh this is awesome. Going to try this on my OCI Ampere instance.

2nd law of thermodynamics -> entropy go up

Yes, but now we have AI to summarize tweets.

The UK does seem more like a banana republic each year.

If you have kids for the status then you're going to have a bad time.

I love both of those and actually only started playing Mini Metro this month. Such synchronicity(!), although mediated by whatever is probably happening on the Play Store, this is less surprising then it would have been 50 years ago.

What is the best way to deploy these on CPUs, arm64 ones in particular?

I'm interested in the CPU inference application of these models with things like the FairyFuse kernels.

I've tried trillim previously but was disappointed that i got higher tok/s just with similar sized models through ollama using just Q4_K_M quants.

I see there is bitnet.cpp and litespark-inference. What else should i look at?

Thanks. This relates to some questions I've got. I was playing around with the previous generation smaller models and found that i wasn't getting any speedups from the T1 and T2 binary/ternary models compared to standard Q4 quants of straight qwen3.6 models. I was wondering whether unpacking of the ternary encoding was impacting inference speed?

If that's the case then why not just train at Q2? I guess the counterargument is that then you lose the nice properties of things like the FairyFuse kernels. I wish there were some good discussions of these trade off.