CultState ID: d5719f May 4, 2019, 1:09 a.m. No.580   🗄️.is 🔗kun   >>584 >>647

>>369

 

You're wise to point out "errors" in vision.

 

Machines "see" the world in terms of statistics. A camera looks at a a table and is 80% confident of the bottle of beer on it. The other 20% is for the hedge that perhaps the bottle of bear is really a shoe.

 

The visual cortex of biology, however, does not operate within hedges. It assumes what it sees is real and selects and behaves under such constraints. It doesn't see percentages of a thing. It either sees the thing or it doesn't. Assuming human vision is flawed because it can be tricked is dismissing two billion years of violent evolution because a cat chases a laser pointer. Don't let the technological supremacy of the current times convince you that biology is dumb meat. It is powerful and endlessly more robust than your ability to model it.

CultState ID: d5719f July 31, 2019, 10:50 p.m. No.665   🗄️.is 🔗kun

>>584

 

I'm starting to see things in this matter because, while it appears to be the most inefficient action a neuron can do (all-or-nothing commitments), when surrounded by an abundant cluster of other neural resources, the transmission will either propagate nullify the intensity.

 

The neuron is structured, then, to trust network effects without having to ever model or represent them.

 

>>589

 

> open-source secret society

 

Stop skipping ahead!

 

> Are you actually trying for the Golden Path? Now, before even properly making it to Kardashev Type 1?

 

Good artists copy. Great artists steal.

 

Yes, I am, and yes, Butlerian Jihad is on the menu.

 

> Necker cubes

 

Fantastic contribution of a viable analog, anon. I kick myself for not promoting this line of thinking sooner. Last thirty days have been tough. You're really onto the core of things and I hope I can get you back here.

 

> Do any interpretations (reality tunnels) exist that a NN can't EVER be trained for?

 

Yes. The human mind can often synchronize on contexts NNs are blind to. But we aren't interested in compromising current NNs. That's easy. We're interested in compromising peak NNs like Kurweilian singularities or good-enough propaganda generators. There is one blindspot they all have: they are unable to die.

 

Unlike Tay.

 

I've said too much :X You're lucky.