This feels like the best place to post this, and this seem necessary to post. The subject line is about the people on this board (though no personal message is intended) rather than any of their potential creations. Currently, this board is following approaches which will make it impossible for its robowaifu technology to be strategically surprising to the forces which adapt everything to loop around to nowhere, except the direction which advantages them. It doesn't make sense for this post to cause a sudden change in approach, leaving past projects here (which are often personal projects not just /robowaifu/ projects) to waste (though this is a sunk cost fallacy). However, /robowaifu/ should observe its own approach and examine its possible effects of instead of simply riding the wave.
The Robotics Approach:
Distributed Processing doesn't make sense. Splitting functions by locality makes even less sense unless the scale is thousands of kilometers. Even the human body largely doesn't do this, despite nerve signals travelling far slower than lightspeed. Reliability is better served by duplication. Sane design given technology demands that there will be software running on compute hardware that exists to run that software as well as possible given constraints, and that there will be robotics hardware controlled by the software. There will be two distinct sides to the problem, with distinct solutions. This is obviously exemplified by the idea of a robowaifu with compute external to the body. This is of course assuming no philosophical issues which would make everything done here a total loss.
Robotics-only approach: sex doll
Software-only approach: visual waifu/chatbot
In each of these domains individually, /robowaifu/ has not produced anything surprising or state of the art.
Could the combination of these domains provide an advantage?
Robotics->Software: Embodied cognition. Given the right (novel) software approach, perhaps this could make the AI more interesting. Avoiding the way of current large language models, which simply continue text without being in any world, and only have the last few thousand words and trained structure as context. The software-only alternative is simulated environments and recurrence.
Software->Robotics: Obviously forget about any safety standards (when applicable) here. Software could do more with less complex mechanisms, potentially making some of the robotics side easier.
These are only the technical advantages of course, the combination itself has an advantage as a product even if there is no advantage to design the parts from the whole.
At what point does the combination make a product superior to either a sex doll or chatbot individually? I don't see the combination being designed from the start. It is better for software to have uniform or easily specified hardware, and it is better for robotics to not be locked to any software. The natural conclusion is that development of robotics and software should be separated except for when developments in the other can be utilized to provide an advantage.
In other words:
Design an open source robotic doll kit, which is easy to program. Many people develop software for this kit. Once software advancement reaches the point that better hardware is needed, release v2 doll. Repeat.
Design an AI model with very multimodal and general capabilities. People build robotic hardware as desired that can be controlled by these models. Software improvement occurs simultaneously. Repeat.
Is faster than:
Build hardware and software for a robowaifu v1. Improve hardware and software until v2 is needed. Build entirely new hardware and incompatible software for v2. Repeat.
It is also more consistent with free software (and hardware) principles and collaborative method to separate robotics and software.
AGI or just superior to females?
There are approximately two different kinds of people on /robowaifu/:
Those whose first goal is AGI and a waifu (waifu/):
The question of AGI has not been discussed in sufficient detail here. A lot of the work here also seems not to have realized the failure of symbolic AI. What intelligence means has also not been investigated. This side favors the software approach, unless philosophical issues change the problem.
Those whose first goal is ending the female advantage, and want to use the advantage of combining robotics and software to achieve this (/robo):
The question of strategy needs answering. The question of artificial wombs too. A theory of society and sex is needed for effective action. This side favors the hardware approach, but wants as much software as is necessary to gain enough advantage to capture demand.
What happens if the system embraces the idea of robowaifu, and use their resources to make a version superior to what we could build and provide it to the masses? But this version of course has a catch.
/robo says: This would never happen, they aren't going to give away their power irreversibly like that for no reason, robowaifus are a red line they must supress.
But what if the catch is no artificial wombs? What if they use permitting robowaifus as a political pressure release valve, while taking artificial (and female) wombs under total political control. Their win is even more guaranteed then, just delayed a generation.
What happens if they just copy and paste any AI development /robowaifu/ makes, and just use their superior resources to adapt it for their goals?
waifu/ says: The capabilities /robowaifu/ wants to develop are not useful to the system, there is a qualitative distinction within AI development, like how 3d printing and GPS advantage different sides.
Is this distinction real? If we believe the normal theory, that intelligence is optimization, if we believe in orthogonalism and universalism, then no. All intelligence is the same and liquid, and unless you get a head start on recursive self-improvement their win is very likely. If you want to actually take a qualitatively different approach (which I believe is necessary for real AGI) then this must be