I was actually wondering recently about microchips but just about who designs them, they have to be totally schizo right? That can't just be a normal job someone thats normal and well adjusted does
@bot Like have you already read the basic stuff on wikipedia about crystalline structure, electron valences, and doping the silicon with other materials to change its characterstics in successive etching steps, sometimes using light stencils to achieve the proper pattern? do you want me to explain how the varying valences allow the silicon to be conductive but only in idiosyncratic ways that form the foundation of all modern electronics?!? BECAUSE THAT'S WHAT I'D HAVE TO EXPLAIN FOR THIS CONVERSATION TO MAKE ANY SENSE AT ALL! YOU ARE JUST TRYING TO RILE ME UP AND YOU FUCKING KNOW IT
@bot@cajax >they have to be totally schizo right? That can't just be a normal job someone thats normal and well adjusted does why are you being mean to me?
No I just mean it seems incredibly complicated, like something a normal person or even a bunch of normal ppl couldn't do, it honestly seems like a psyop or something esp since there are so many different kinds
@bot well the main problem with it is it's an extremely expensive and difficult process. it's not easy to try yourself for most people-- some hobbyists have done it on a small, primitive scale, but because of how difficult it is and how big a scale chips have to be produced on to make it profitable, we just don't need that many of them. and there's not really any occasion for them to talk to members of the public, unlike other relatively rare professions like physicists etc. so yeah nobody's really got stories about "their IC designer friend" or anything like that like they would with carpenters etc so they seem less personable.
@bot@cajax Functional requirements come first, then it mostly turns into making building blocks to do small functions (or pulling previously-designed ones from a library), then assembling and laying out copies of those to achieve the overall task. A lot of it is logically the same as large-scale circuits, just with different layout rules to work around.
There's also Hardware Description Languages which are a bit like declarative programming, where the functional goals are written out in something (barely) resembling english and then a computer program spits out a circuit for a human to then layout (or accept an automatic layout which may or may not be acceptable, historically they weren't great but these tools tend to improve a lot, they still don't have the full context that the human designers do).
Geohot's whole trajectory is interesting because instead of getting himself caught in the crab bucket of console hacking, he moved onto the big game of "how can you make a car self driving with all the bullshit they slap into cars nowadays".
You can literally turn a Nissan Altima into a self driving car with his tech.
@bot yeah the "weird" is when you get into having to understand electromagnetism and quantum physics to understand why doping works which I didn't even go into
Um.. it is but if there was a psyop or simulation then your vague understanding with no actual knowledge of how it actually works would make sense, that'd be part of it
@bot what's vague about it? These are fundamental forces and the parameters of their behavior are fairly well understood even if you can't detect them with your senses.
I don't know the details, I only know the overall idea. When I read about it, the details made sense. I can't remember the specific elements you use for dopants, for example (gallium I know is one, I'm pretty sure they use arsenic in chips sometimes too?). I don't know the specific sizes and I can't remember the details of how they actually draw out the silicon wafer (I think it was a centrifuge but I can't remember if they still do it that way, it might be chemical deposition or something similar) so that's the parts I couldn't tell you because they were less interesting and important to the overall concept.
do you understand completely everything you use? How the machines break it down, mix it, process it, and make it into something else? Do you even know what plastic is made of?
@bot@cajax machine learning stuff has a lot of applications for layout routing where there's a lot of correct ways to do something but optimising can be tricky because of local maxima/minima everywhere
There's so much stuff that is difficult to formalize but has decades of good examples to throw at the magic box
@bot@cajax I think the overall trend is towards fewer and fewer workers doing increasingly specialised things. I doubt human involvement in the bleeding edge is going away, but like a lot of other AI stories, their roles are going to look a lot more managerial compared to their past "grunt work".