mastodon.ie is one of the many independent Mastodon servers you can use to participate in the fediverse.
Irish Mastodon - run from Ireland, we welcome all who respect the community rules and members.

Administered by:

Server stats:

1.5K
active users

#systems

3 posts3 participants0 posts today

I very much doubt the AI singularity thing is gonna happen, because complexity doesn't scale linearly.

Example, if I make a game prototype, I can usually do it in a day. But to make a complete game out of it takes years.

Why? Because making 1 system is easy. But making a bunch of systems that all interact is exponentially more complex.

Neural nets aren't that different. To go from LLMs to AGI, you need to add a variety of neural modules that all interact, and whoops there goes complexity, up on the exponential scale.

Intelligence gives species an edge - but if that's the case, then why didn't we yet evolve to be hyper-intelligent over the past million years? We still kill ourselves regularly doing the dumbest shit. And fall for super obvious psychological traps.

Well, it's pretty much like with computer processors - at some point, you can't really go faster physically, so in that field, core speed improvements stopped around the ~4ghz mark - and instead, designers went multi-core to go parallel.

Works great for stuff that doesn't interact much, but as programmers know, if these cores run threads that have to interact with other threads (or their data, really) processed by other cores, you get slowed down a lot by having to synchronize things. Also, you need a good design to optimize this synchronization, but it will always be a hurdle.

Brains are similar. Evolution kinda got stuck with the brain design it started out with, so it could be more optimal if it were allowed to start from scratch with the knowledge we now have - and basically, that's what we can do now with AI - but in the end, there's limits to how good a system you can design.

It's the exponential complexity thing. You can't get out of this! It's a hard boundary, like a law of our nature, that can't be broken.

Systems will always consist of subsystems that interact with a variety of other systems - on all scales! On a tiny scale, you could consider each neuron a system, that interacts with multiple other systems (neurons). On a massive scale, you can consider each human a system that interacts with other systems.

Add to this another bonus "law of nature" that, the more complex things become, the more prone to collapse they are, and you have a hard limit. You can only have that much intelligence going on in this universe/reality. It all doesn't scale well.

(fun case in point: we have over 8 billion human beings running in parallel, and we're still dumb as fuck as a global society)

This toot went a bit all over the place, but I think you get my point. I should write a paper on this some time 🤔 I'm not in the academic loop, maybe some people already did; but then it's odd that I never see someone mention it, and many people seem to think something like a technological singularity is a possibility.

So maybe I'm on to some insights here that are relatively novel? That, or I'm just wrong 😁

So, I'm happy to hear counter-arguments to my, ehh, thesis? tootsis? 🙏🏼

who wants to pay me to reproduce this illustration with some cleaning up and in colour?

Drawn on FF Kronprins Haakon in 2021 for a talk on the ship (which, unfortunately, I've lost :/ ). The left side was published in the 2022 Fram Forum, the right side was a bit too spicy of a hot take in a field which still insists we can do it all from spaceborne platforms 😬❄️

“There’s a common misconception that state-of-the-art technology has to be expensive, energy consumptive and hard to engineer. That’s because we have been persuaded to believe that #InnovativeTechnology is whatever bombastic billionaires claim it is, whether that’s commercial #spacecraft or the endless iterations of generative #AItools.

As the #Canadian #technologist and #engineer #UrsulaFranklin once said, fantasies of technology would have it that innovation is always “investment-driven, shiny, lab-born, experimental, exciting”. But more often than not, in the real world, it is “#NeedsDriven, #scrappy, on #location, #iterative, #practical, #mundane”.

The real pioneering #technologies of today are genuinely useful #systems I like to call “frugal tech”, and they are brought to life not by eccentric billionaires but by people doing more with less. They don’t impose #TopDown#solutions” that seem to complicate our lives while making a few people very rich. It turns out that genuinely #innovative #technology really can set people free.” — Eleanor Drage

#OpEd / #EleanorDrage / #tech / #FrugalTech / #Cambridge #LeverhulmeCentre / #FutureOfIntelligence <theguardian.com/commentisfree/>

The Guardian · Enough of the billionaires and their big tech. ‘Frugal tech’ will build us all a better worldBy Eleanor Drage

#systems #architecture #management #dailyreport
Revolutions in system evolution. Part 1: Definition.
Rapid transition in structure, behavior, or interactions.
Shifts are tied to changes in interactions, reshaping
structure and producing new behaviors.

What leads to the tipping point? 1) Resource
exhaustion, 2) accumulation for transformation, 3)
external factors breaking hierarchy.

During transition, new properties break hierarchy,
changing main interaction forces. After revolution, the
system stabilizes in a new state with new dynamics and
static hierarchy.

Why do systems evolve with revolutions? I guess,
hierarchy or structure creates barriers and tension due
to new, unhandled interaction forms.

Hierarchy acting as a dimension. Maybe space or time is a
hierarchy form in a greater system.

#systems #architecture #management #dailyreport
Revolutions in system evolution. Part 2: Loops & complexity.

Feedback loops amplify or weaken changes. Positive loops
(e.g., core heating in stars) push systems toward
revolutionary transitions, while negative loops
stabilize them.

Complexity emerges as systems evolve, enabling new
behaviors (e.g., multicellularity,
industrialization). This drive is a hallmark of
evolution, exploiting interactions for higher-order
states.

Examples:
- Stars transition from main-sequence to red
giants, then to white dwarfs or neutron stars, adapting
to new conditions.
- Biology's transition to multicellularity enabled new
structures (tissues, organs) and further revolutions
(vertebrate development).
- Societies' Agricultural Revolution led to settled
communities, enabling the Industrial Revolution's urban,
mechanized structures.

My freelance quote for your software project.

50 - if you tell me what you want, I confirm the requirements, and I build it for you
100 - if I have to pry a description of what you want out of you like pulling a tooth
250 - if I have to build it on a "company laptop" you supply
350 - if I have to regularly video call with you for no particular reason
400 - if I have to use JIRA
450 - if I have to use Github
500 - if I have to use Zoom
750 - if I have to use Teams
1,250 - if you ever want to "watch me work" by sharing my desktop
1,500 - if the project is fixing something your employees built for you
1,800 - if it's fixing something a previous contractor built
2,500 - if it's fixing / completing a project a previous, cheaper freelancer failed to complete
19,500 - if it's fixing something created with "AI"

#AI#LLM#quote