FA+
Shop
10269
Views
Views
466
Favorites
Favorites
Category
Artwork (Digital) / All
Species Dinosaur
Size 1280 x 727
File Size 166.2 kB
Report this content
★
More from Thesociallyawkwardpinguin
Listed in Folders
![[c] AMT Bulwark](http://d.furaffinity.net/art/thesociallyawkwardpinguin/1548268806/1548268806.thesociallyawkwardpinguin_12.png "Click to change the View")
[c] AMT Bulwark
"At its introduction in 2194, the AMT Bulwark was near-universally praised for its unprecedented mobility, modularity, ruggedness, and most of all the revolutionary self-learning neural network that made it the first self-aware walker designed by the Askhari. Together with the Starling-class exploration vessel and the Mark VI exosuit, it quickly became one of the most iconic technologies of the interstellar exploration age, found all over newly founded colonies - and eventually all over newly opened frontlines.
Seventy years later, the droid's novelty has pretty much entirely worn off. But the walker's near-legendary ruggedness and defense manufacturers' unwillingness to switch up their streamlined manufacturing lines means that hundreds of thousands of Bulwarks are still produced with only the slightest modernization tweaks. Over its lifetime, countless variants of the chassis were used on the ground, in space and underwater; by field engineers, combat medics and war correspondents. Even a slightly eyebrow-raising road-legal variant, piloted rather than automated, is commonly found on former fortress worlds.
***
Full view to see details!
Sometimes I get really bored of macro and try drawing other things :p As such I often give priority of sfw commissions that doesn't involve a lot of fluids and unwilling participants.
And here's the next commission of last last month for
of one of the unit of his empire in stellaris :D Blurb of text by himself.
Next drawing will be uploaded in a week (hopefully the macro boarding party for those of you who follow me on twatter).
Seventy years later, the droid's novelty has pretty much entirely worn off. But the walker's near-legendary ruggedness and defense manufacturers' unwillingness to switch up their streamlined manufacturing lines means that hundreds of thousands of Bulwarks are still produced with only the slightest modernization tweaks. Over its lifetime, countless variants of the chassis were used on the ground, in space and underwater; by field engineers, combat medics and war correspondents. Even a slightly eyebrow-raising road-legal variant, piloted rather than automated, is commonly found on former fortress worlds.
***
Full view to see details!
Sometimes I get really bored of macro and try drawing other things :p As such I often give priority of sfw commissions that doesn't involve a lot of fluids and unwilling participants.
And here's the next commission of last last month for
of one of the unit of his empire in stellaris :D Blurb of text by himself.Next drawing will be uploaded in a week (hopefully the macro boarding party for those of you who follow me on twatter).
Category Artwork (Digital) / All
Species Dinosaur
Size 1280 x 727px
File Size 166.2 kB
Not exactly, I'm delaying pic release to maintain an illusion of regular uploads :p This pic took around two weeks while working on other commissions too.
"Weeks" is a big word, more like one hour here and here.
And no, not flight capable, just rule of kool stuff x)
"Weeks" is a big word, more like one hour here and here.
And no, not flight capable, just rule of kool stuff x)
what types of neural networks did it use?
like seriously, neural networks don't learn when they have got a lack of error, and so that isn't a neural network, do you mean that it is a generative network that tries to predict frames ahead of it, because i could see that being used. and the way you describe it is not how neural networks are implemented, rather networks are used to create function of which are static once the desired outcome is achieved, no need to learn once it reached peek precision, i have made neural networks in the past and your use of them to somehow humanize a robotic ware-machine, is confusing.
like seriously, neural networks don't learn when they have got a lack of error, and so that isn't a neural network, do you mean that it is a generative network that tries to predict frames ahead of it, because i could see that being used. and the way you describe it is not how neural networks are implemented, rather networks are used to create function of which are static once the desired outcome is achieved, no need to learn once it reached peek precision, i have made neural networks in the past and your use of them to somehow humanize a robotic ware-machine, is confusing.
Hey, I'm the one who wrote that bit! I was actually thinking it's a model of an biological neural network (AKA a synth brain) rather than artificial neural network, which is what you're describing. Honestly though either option is just "it's strong AI ok" translated into casual technobabble C:
That's also an artificial neural network. What I was actually thinking is a model brain of some saurian critter with dog-level intelligence - a biological neural network, just... computerized. Yes, it's incredibly overkill for murderbot purposes, but this is the future, the idea of giant doggo mechs is fun, and it adds some fun implications that the raptors will eventually figure out consciousness uploads (woaaaaaaaaah).
that just shows how little you understand, ANN (attifical neural networks) are not the same as a brain, first let me expain how the (nerural) part works, its called a perceptron, it works by taking 2 or more inputs, mulitplying them by a weight, lets call it W, and then adding it all together, then giving of either a 1 or a o depending on if the answer is less or more then a pre determined number, so the perceptrons equasion would be sum == ((a*w)+(b*w)). simple. the inputs could be anything from the amount of people at a concert to the angle that a object is rotated, and the output is fed into another perceptron. one step above now, now lets learn about ANNs, those perceptrons can be linked to more perceptrons and a somewhere in the network can be a output, that output, or outputs, is the answer, the many inputs is the question. now how it learns, is that it looks at the question and gives its guess at an answer, and then is compared to the right answer and then looks at what it did wrong, it looks at what weights where needed for it too give the correct answer, it changes it and looks at another question, when it gives correct answeres all the time it is given new questions it hasnt trained on, and no longer learns once it is deemed correct, and then is compiled into code and is no longer a neural network, but rather a set of instructions.
for one, a design based on a brain is inefficient, perpetuating signals (neurons) is very in efficient at transport and transformation of data, and the amount needed inside a single network to fully simulate even a worms brain, let alone a raptors brain, would need super computers way outside mores law, not just that, but a raptors brain is a stupid idea, they where non heriacal preditors that even hunted their own species. and probably had no concepts of logic moral and instruction based learning, why not just copy a dogs prefrontal cortex and do everything else with predesigned circuts rather then a constantly changing network so that the property of a dogs ability to make decisions, miked with its artificially created (as in no receptors or anything) emotions, it would be what you wanted without in heriting the normal physical traits of a dog, as many of the aspects of dog behavior are based around their inabilities, such as panting is the only way they have to cool down as the fur traps sweat. if your looking for a perfect system of death machine then why not go full programmed intelligence rather then adapted intelligence from animals
>and the amount needed inside a single network to fully simulate even a worms brain, let alone a raptors brain, would need super computers way outside mores law
It's 2194.
>a raptors brain is a stupid idea, they where non heriacal preditors that even hunted their own species. and probably had no concepts of logic moral and instruction based learning
There's a raptor, in a spacesuit, doing a blep, standing right next to the mech. There's a graffiti of the raptor doing a blep on the mech. That should imply these raptors are in fact sentient and capable of learning.
>why not just copy a dogs prefrontal cortex and do everything else with predesigned circuts rather then a constantly changing network
Because that doesn't sound as cool as giant doggo mech friends that can bond with their owners and have distinct personalities.
Bottom line is: this isn't hard sci-fi, so I reserve the right to do whatever I want.
It's 2194.
>a raptors brain is a stupid idea, they where non heriacal preditors that even hunted their own species. and probably had no concepts of logic moral and instruction based learning
There's a raptor, in a spacesuit, doing a blep, standing right next to the mech. There's a graffiti of the raptor doing a blep on the mech. That should imply these raptors are in fact sentient and capable of learning.
>why not just copy a dogs prefrontal cortex and do everything else with predesigned circuts rather then a constantly changing network
Because that doesn't sound as cool as giant doggo mech friends that can bond with their owners and have distinct personalities.
Bottom line is: this isn't hard sci-fi, so I reserve the right to do whatever I want.
Comments