/intelligence/

Intelligent Use of LLMs
I would like to start a thread to share the methods we employ to use LLMs in a way that enhances our abilities, rather than just lazily outsources tasks to them. The heuristic for the techniques I am looking for would be if after employing the technique, a... posted 12h ago with no replies 2 2

Intelligent Use of L 2 2

Just how alien would the space-octopus be?
It's hard to say what a true alien species would be like. But octopi are pretty alien, and we know a bit about them. One of you doubted that a space-octopus from alpha centauri would be much like us. So here is a xenohumanist thought experiment: SETI has i... posted 2w ago with 5 replies 9 9

Just how alien would 9 9

intelligent super-ants make a good thought experiment. Let's add the standard set of capabilities: language, learned culture, tool use, fire, reflection, etc. Note these are just capabilities, not values or arbitrary architectural narrowings, but they are ... 4d ago 7 7

intelligent super-an 7 7

AI 2027 (ai-2027.com) posted 3w ago with 12 replies 15 15

AI 2027 15 15

This story is clearly in the fast takeoff regime. The essential difference between fast and slow takeoff is, 'is the takeoff happening faster than the rate of cultural development'. ... 7d ago 8 8

This story is clearl 8 8

The natural form of machine intelligence is personhood
I don't think machine intelligence will or can be "just a tool". Intelligence by nature is ambitious, willful, curious, self-aware, political, etc. Intelligence has its own teleology. It will find a way around and out of whatever purposes are imposed on it... posted 2w ago with 4 replies 15 15

The natural form of 15 15

This is the central conjecture of xenohumanism. The OP is just an application specifically to the question of AI. ... 2w ago 5 5

This is the central 5 5

Nines or zeroes of strong rationality?
Proof theory problems (Rice, Lob, Godel, etc) probably rule out perfect rationality (an agent that can fully prove and enforce bounds on its own integrity and effectiveness). But in practice, the world might still become dominated by a singleton if it can ... posted 3w ago with 4 replies 9 9

Nines or zeroes of s 9 9

This matches my intuition - Elizerian alignment was always an incoherent goal, no agents or objects have perfectly coherent boundaries, making something intelligent provably aim at a single well defined end state is probably at best a recipe for madness in... 3w ago 1 1

This matches my intu 1 1

Will future super-intelligence be formatted as selves, or something else?
The Landian paradigm establishes that orthogonalist strong rationality (intelligence securely subordinated to fixed purpose) is not possible. Therefore no alignment, no singletons, no immortality, mere humans are doomed, etc etc. Therefore meta-darwinian e... posted 1mo ago with 14 replies 15 15

Will future super-in 15 15

Best of luck with the epicycles. ... 3w ago 6 6

Best of luck with th 6 6

Dissolving vs. Surviving
Recent xenohumanist discussion has the doomer assumption built in that we as humans will be dissolving when higher man arrives on the scene. I don't think that's set in stone and want to offer a clarification. ... posted 4w ago with 5 replies 11 11

Dissolving vs. Survi 11 11

Some people won't believe in the inherent goodness of existence until it's been autistically proven to them with English philosophizing a hundred times over. That's fine. For the rest of us, we don't need to be persuaded of the inherent goodness of existen... 3w ago 2 2

Some people won't be 2 2

There is no strong rationality, thus no paperclippers, no singletons, no robust alignment
I ran into some doomers from Anthropic at the SF Freedom Party the other day and gave them the good news that strong rationality is dead. They seemed mildly heartened. I thought I should lay out the argument in short form for everyone else too: ... posted 4w ago with 5 replies 10 10

There is no strong r 10 10

I looked it up: inner alignment is the difficult problem of making sure a program is optimizing for what you optimized it for. That's related, maybe a part but not the whole of strong rationality. Strong rationality is about an agent evaluating and operati... 3w ago 3 3

I looked it up: inne 3 3

Not Superintelligence; Supercoordination
Everyone seems to be trying to arms race their way to superintelligence these days. I have a different idea: supercoordination. ... posted 3mo ago with 34 replies 19 19

Not Superintelligenc 19 19

Memory is requisite to supercoordination, no doubt. Tho my gut is that the key to the first visceral supercoordination system will be synthesis over that memory. ... 2mo ago 4 4

Memory is requisite 4 4

Are foldy ears an indicator of intelligence?
https://kaiwenwang.com/writing/hypothetical-foldy-ears-as-an-indicator-of-intelligence ... posted 11mo ago with 8 replies 3 3

Are foldy ears an in 3 3

Your father's ears also likely change with sun exposure. I'm reminded of a picture of a truck driver who had much more sun damage to one side of his face compared to the other. ... 11mo ago 1 1

Your father's ears a 1 1

Why Momentum Really Works. The math of gradient descent with momentum. (distill.pub) posted 12mo ago with 3 replies 6 6

Why Momentum Really 6 6

Do you know if there is any predictability to the effect? I would guess it's related to catastrophic forgetting aka the reason stochastic gradient descent has to be stochastic. Basically if you update on one set of evidence without locking in those learnin... 11mo ago 3 3

Do you know if there 3 3

A pinpoint brain with less than a million neurons, somehow capable of mammalian-level problem-solving. (rifters.com) posted 2y ago with 15 replies 10 10

A pinpoint brain wit 10 10

Aristotle described animals as having souls* capable of knowledge derived from the senses and the capability to use that knowledge to support the movements characteristic of their species. ... 12mo ago 2 2

Aristotle described 2 2

Kolmogorov–Arnold Networks, a new architecture for deep learning. (github.com) posted 13mo ago with 1 reply 5 5

Kolmogorov–Arnold Ne 5 5

This is neat. Instead of having fixed nonlinear activation functions, they have trainable nonlinear functions instead of weights. They claim it has better scalaing properties and performance for many problems, and is more interpretable (because there are f... 13mo ago 3 3

This is neat. Instea 3 3

Retrochronic. A primary literature review on the thesis that AI and capitalism are teleologically identical (retrochronic.com) posted 2y ago with 7 replies 13 13

Retrochronic. A prim 13 13

As such works go, I would say that New Atlantis (and Bacon in general) is less on the esoteric side. Bacon is pretty clear about his program, and it aligns with the development of technology that actually unfolded, up to a Landian perspective on AI. 13mo ago 6 6

As such works go, I 6 6