/sci/ - Science & Math

Depends on whether or not the big tech companies can profit from it. 20 years of they can, 50 years is they can't.

>>11933380
This

the human brain uses photons to compute. pretty cool stuff.

>>11933385
It doesn't.

>>11933372
We have been perfecting our current ways of computation for decades, getting something new on that level and even above will not be easy. Time will tell.

>>11933393
>It doesn't.
https://www.technologyreview.com/2010/12/17/198375/the-puzzling-role-of-biophotons-in-the-brain/

>>11933393
Isn't electric at it's limit already though?

>>11933385
how? I thought it was electron transfer between synapses...

>>11933426
that is mildly interesting. got anything more substantial?

>>11933477
It is, which is why we are looking into things like optical computing, graphene, and gallium nitride as a replacement.

>>11933602
https://www.nature.com/articles/srep36508
https://arxiv.org/vc/arxiv/papers/1012/1012.3371v1.pdf
>>11933583
You thought wrong.

>>11933477
Not even close.

>>11933611
Retard post.

>>11933850
most informative post itt

Fluidics is the future!

DNA computing when bros?

>>11934877
never
>>11934864
go on

>>11933393
Light is just a different way of implementing the same circuit. If we can get clock rates 5,000 times faster already, the only hurdle is actually making the chips.

>>11933372
>not even using gravitational computing
s m h t b h f a m

>>11933372
never because light will never replace conventional digital logic. there is no evidence suggesting it will be ever be faster than silicon cmos. youll get horrible loss from going from electricity->light->electricity.

>hurrr just replace everything with light so u dont swap electricity->light
yes just replace all the amplifiers, sensors, memory, ... with optical memes that dont exist

on current chips, only thing people are looking at is laser based interconnects for very long top level wires or cross chip communication. it only works on long distances because for short distances the electricity->light loss isnt worth it

>>11937255
>what is fiber optic internet?

>>11933380
10 years if there's a major world war and we're forced to use it at scale

>>11934877
>>11935509
I need a proof that the ribosome with its DNA "tape" is turing complete by Monday

>>11933372
Isn't this just fibre optics though.
Yeah nag putting that in a computer is silly as the cost will outweigh the benefit at short distances.
I'm not even sure you can make glass strands or acrylic small enough or flexible enough.

>>11938837

>I'm not even sure you can make field-effect transistors small enough or flexible enough.

Is what people said until the '60s

>>11940438
precedence is not a theory much less a mechanism or design.

>>11940453

Neither is your uncertainty and pessimism!

>>11933372
Are you too young to have heard of CNTs or Graphene? Every so often a new wonder tech comes up and is touted as the next big thing in technology.

Then they run into the usual problems: too hard to scale said material or it requires almost impossible levels of purity or it doesn’t have a way to control currents, etc.

These research papers always make big claims because they are looking for grants.

Hint: not a single one of the big hardware companies is doing any serious research into sillicon replacements. They know we’re stuck with this for a very, very long time.

its pretty retarded to think companies are holding back technology. there is shit tons of competition. instead ask questions like light is faster so what? how fast can the "light transistors" switch? Whats the theoretical speed limit? how small can we even theoretically get the light transistors and connections too? these things matter not some retarded idea of how light goes fast in a wire. its just like the other memes. graphene has a minimal advantage over silicon. so its not even worth investing money and basically working at the ground up to get it to the point in which its usable. SO it will never be used. We are really far in out "silicon manipulation technology" its not worth it to go any where close to other materials with slight advantage of silicon. We already have different materials used for different applications like germanium. its not like we are afraid of using any other material.

>>11937255
>... with optical memes that dont exist
like they did with all the transistors that didn't exist?
>>11941434
graphene and different semi materials are still being shilled now, but probably not consumer ready if it exists

>>11933477
>Isn't electric at it's limit already though?

as far as miniaturization, i.e. moores law, we're butting up against the theoretical limits of how many transistors we can jam into a dingle CPU chip, yes...

without any new advances in CPU architecture and substrate material, most new CPU advances will most likely be transistor organization, power efficiency, multi-thread support, multi core support, etc...

the "Physical limit" to current nanoelectrical CPU's is the size of the transistors and electrical pathways.

for example, 1 nanometer is around 10 angstroms... an angstrom is the diameter of a hydrogen atom.

Our current generation CPU architecture is about 7-14 nano-meters, or 70-140 hydrogen atoms wide.

the theoretical limit to the size of a CPU transistor/current path is maybe 10-12 angstroms... this is because Transistors are made of two types of doped silicon... silicon with a positive charged element, and silicon with a negative charged element. so you need several silicon atoms of thickness in order to actually fit the dopant elements within the substrate.

>>11933372
where's the fucking paper that details it?
>>11933380
if they really built a ring oscillator that works at terahertz that's fucking impressive. If they're doing it with nanowires to make optical wave guides smaller than light typically is then I'd say it's viable for something. Be damn ironic if they can build a processor and memory with it too because it'd turn the current trend of having a bunch of slow clocked processors on its head. How about one processor that's not as capable but FUCK FAST. Some applications really need clock rate, something that has stagnated with conventional silicon microchips. It's not going to be easy to make tiny as fuck optics, but goddamn if you can get FUCK FAST clock rates it's probably worth it.

>>11941614
ever heard of molectronics? You can have single molecule transistors.
https://en.wikipedia.org/wiki/Molecular_electronics
Weird ass quantum effects are a fucking bitch though.

If you shine a fleshligh into your butt, do the light neurogenesis into minibrains ????
LIERAL HEADBUTT LOOOL

Light based computation is infinitely more likely to make it to the consumer than quantum computing. The latter is better for large scale industrial high powered computation where the quantum aspects improve on previous algorithms. All in all, it depends, photonics has potential for a faster, lower temp and less research/development intensive computer, but it ultimately depends on how the tech matures and if the consumers want that.

>>11941434
Graphene has the manufacturing barrier. It's fragile, hard to make, it can be toxic to humans, and requires completely new accessory technologies. Same problem with nanotubes. You can't just directly compare every single new technology, just because one doesn't happen, it doesn't mean all of them won't.

bump
>>11941840
based

>>11941913
>it can be toxic to humans
Just not eat your processor no matter how appetizing it looks and you're good.

I would wager it would be more beneficial to find out how to get the heat away from the processor. Besides electromigration, heattransfer will be increasingly more important to solve. Look into Intel Foveros, where they literally have the more hot components (chip die) above the relatively cold components (IO).