New Photonic Chip Explained: The Next Era of Computing
ฝัง
- เผยแพร่เมื่อ 5 พ.ค. 2024
- Download Opera for free using opr.as/Opera-browser-anastasi... Thanks Opera for sponsoring this video!
Timestamps:
00:00 - Intro
00:52 - Computing with Light
04:33 - Taichi Chip
06:05 - Photonic Logic Gates
09:21 - Computing with Diffraction
10:40 - How Taichi Chip Works
13:05 - Results
B-rolls sources: Diffraction Limited, IBM
The paper: www.science.org/doi/10.1126/s...
LinkedIn ➜ / anastasiintech
Support me at Patreon ➜ / anastasiintech
Sign up for my Deep In Tech Newsletter for free! ➜ anastasiintech.substack.com
MY GEAR (affiliate links):
➞ Camera Sony Alpha 7 IV: amzn.to/3P53z7i
➞ Lens Sony 35mm: amzn.to/3P62v3f
➞ My Mic: amzn.to/3Ir7pUN
➞ MacBook Pro: amzn.to/3VbdgF4
➞ Lenovo Legion Gaming Laptop: amzn.to/3TmynD6
Download Opera for free using opr.as/Opera-browser-anastasiintech Thanks Opera for sponsoring this video!
Opera was bought by China a few years ago. I'm guessing the recent ad blitz is a response to the tiktok ban.
that is not a true optical chip. you must develop a true optical transistor that can deal with a actual laser beam. this design fails the test.
Never thought I would pay so much attention to photonic computing, (I am going to use Opera for washing the car as well).
YES! I would love a video on that
The possibility, "is still far away", of a mental-internet with a synaptic interface to human neurons with photonic-quantum processing connected in artificial neural networks, in which information is processed and decoded into qubits and subsequently converted into jpg. and or MP3/4 for human understanding, could be a way to advance technological development???
Minor correction: Light *in a vacuum* travels at 299,792,458 metres per second, but light in a fibre optic cable travels 30% slower at around 200,000,000 metres per second. We could, in theory, increase that speed by using different materials for the fibres, but we will probably never get close to the vacuum speed of light.
Thank you, although I'm not sure how minor 30% is.
@@Lost-In-Blank
High-frequency traders are paying fortunes to reduce cable length by an inch.
such traders use air radio links instead of fibre optics.
what's 30% amongst friends. It goes from INSANELY fast to just blazing fast...
@@Nilmoyprobably because radio really travels at the speed of light ?
Well, definitely a separate video on how the photonic computing would work.
Isn't this video that video?
@@ryanmcgowan3061he's referring to her comment at 4:30 about photon quantum computing at room temp :)
@@chrisfirgaira He must have forgot the word "quantum" then, because this whole video was basically how "photonic computing" works.
any university physics class will present a "light table" where lasers and lenses/prisms perform calculus operations at the speed of light. Very old an open tech. Many fighter jets from 30 years ago use "photonic processors" to achieve flight stabilization for example
Scientists from Tsinghua University China have developed Taichi photonic chip, if want to know more how it works then learn mandarin chinese.
Anastasi: "would you like me to do a video on..."
Me: "yes!"
😂😂😂
A yes man..
You're a gentleman and a scholar. You'd watch anything she publishes, even behind a fans' only paywall!
I understand and agree completely with the proposal!😍
weird world, eventually gaming PC RGB lights will actually increase performance 😅
the term go fast stripe could end up being true. I love that more than I reasonably should.
I love how animated and invested you're in things that interest you. I like seeing how excited you are each time you publish a new video.
I always think exactly the same whenever I watch one of Anastasi’s videos. It’s scientists and engineers with this sort of passion for their subjects that drives all these innovations that we see reported on here. (As well as being passionate about a subject some serious brain power is also required to push forward the frontiers of one’s field of course.)
Coldfusion had an episode about the progress with graphene transisitors. Things are heating up. I love it. Thanks for a wonderfull reaserch news.
The video on analog computing by Undecided with Matt Ferrell is also worth mentioning.
bandgap too small.
actually she had an episode on graphene transistors as well, like two month ago: th-cam.com/video/wGzBuspS9JI/w-d-xo.html
photonic computing still relies on clock cycles & is limited to the number of gates (fanout) do to attenations. most photonic compute is limited to matrix compute\comparision since its basically one operation per clock cycle. All of the other logic\compute is handled by the electronic (silicon) systems.
I don't think we'll see much on an advance for photonic computing for a long time. I think the current function of photonic matrix will be replaced with analog since its far easier to design an analog matrix than photonics & the speed is about the same. The issue with electric digital comparisons is that takes lot of logic gates to perform comparisons or matrix math functions (multiple clock cycles). Its pretty simple & fast to do it in analog, All you need to do is pay the compute cost to convert a digital value into an analog & back into digital. The analog work can be done less than one clock cycle.Its easy to do comparison, additions, subtractions, mulitiplications & division using analog circuits.
2:44 My father worked with Charles Kao at STL in Harlow, UK, but in a different team. My father developed the first plasma etcher while there. I can't say "invented" as the idea had been around for a while, but no one had been able to achieve useful etch rates before.
Im sure you are proud of your father, man. Kudos
I work in the field of photonic integrated circuits, and this is the most complex circuit ive ever heard of. Great video and analysis !
As you mentioned, true wall plug energy efficiency of photonic circuits do make it a less-attractive solution for computing, which is often overlooked in these papers. It often comes down to material science to come up with new ways to decrease the energy bill.
This tech barely scratches the surface. It's why my interest shifted to photonics years ago. The writing is on the wall for silicon. Btw, nice diffraction grating background. 😵💫
sorry to break it to you, nothing is replacing transistors on silicon. quantum computing will have a huge impact on computing and opens up a new set of solvable problems, but quantum chips are only faster for those limited set of algorithms.
@@loo_9Sorry to break it to you, quantum computing is 99.99% hype, 0.01% real and useful.
@@tantzer6113isn’t that what @loo is saying?
luddites
I do miss photonics, unfortunately I'm not able to pursue career in it....
This is very insightful and eloquently explained. Thank you Ana for posting it and please consider recording a video on quantum computers with photonics chip.
I'm bestowing to you my "Mr. Wizard" honor for your wonderful science explanations.
Mr Wizard was a man named Don Herbert whose kids' tv show taught basic "tabletop" physics on a kitchen set with random kids. He was one of my heroes as a kid and as an adult for his kind respect towards the kids. In a similar fashion your pleasant concise delivery makes learning a joy.
Thanks
"Watch Mr. Wizard". Never missed it.
@@mgeldern Mr Herbert died on my birthday, 12 June. I loved that guy.
Quantum photonic chip video would be awesome.
You better eat a quantum bread
Quantum photonic AI being
kinda sounds like something from star trek
I swear I'm commenting from a Quantum phone on Quantum TH-cam.
@@beowulf2772 I seem to remember "positronic network" or something similar in Data's brain. So they were using antimatter in their fiction :D
I just love listening to your commentary on these scientific articles and explanation.
Just a little ditty from the past... At Bell Labs (where the Laser was invented for the eventual Maser application to fiber optics which was also invented there), in around 1991 there was a small group of Physicists that were developing a "quanta gate" that they hoped could eventually evolve to replace the transistor (also invented there) based 4 and 5 ESS (electronic super switches) in their Central offices. The Bell System was broken up shortly after, and the labs were disbanded into what is now a Nokia office complex. Love your videos and your enthusiasm. I wish you were my daughter. 🙂
I mean moving photons around aught to be more efficient than moving electrons around right? I feel like this all really comes down to the discovery that high purity silica fibers can transmit signals orders of magnitude further than the shoddy stuff originally used to assess the value of the technology. That, and all the other material property discoveries made over the past 40 years in the field of optics finally coming into practice. Silicon photonic computing being a bit of an academic/business community effort in Europe right now and all.
As far as I know Bell labs also headed the movement to continue analog telephony over digital. Which would have been unreasonably expensive, and overcomplex, but arguably could have lead to a ground up analog internet.
We could have had live video conferencing over fiber optics in the 60s. Worth noting that there's still a similar planning, and funding problem holding us back now, we even "over-invested" in fiber infrastructure at one infamous point telecoms history(although there were bigger problems at the time, and the public was hungry for scapegoats).
The complexity scaling of continued analog development would put the timeline in a comparative stall-out for a while, but at some point the high exponential growth on continua data computing would blow digital out of the water.
I like to think we met the problem somewhere in the middle and used digital as a well timed stop-gap. 🙂
@@Leadvest Currently working at Nokia myself. Nothing related to the Bell Labs department but still in the area of large-scale telecom. I think you're spot on with your conclusion. There is an aspect of telecommunications that we often forget : it's not only phone calls, but also the Internet.
There is an high degree of complexity at the application level in order for us to enjoy the benefits of watching TH-cam in 4K from any device anywhere in the world.
Intuitively I would say the bottleneck is the pace at which we can route traffic, which is more of a decision making process rather than purely scaling up.
Perhaps now is the time for another step forward, or should I say "a step backwards". Exploring the past and the technological discoveries we discarded might allow us to make further progress than we think.
Awesome report Anastasia. The world is headed for big changes, this is a big leap. Thank You for your channel.
This definetly going to power the next age of computing devices….I have been betting on this for a long time
Yes me too , Had the idea over 20 years ago . Also had an idea for a laser powered lawn mower two years later it was for sale for about a million dollars .
@@matthewcalifana488sure buddy. Sure.
What company should one invest? Graphine computing breakthrough major just announced too
@@matthewcalifana488sure buddy, sure
What company is she talking about in this video??
Great coverage of photonic processing. Not new though
many jet fighters from 30 years ago use optical processing to achieve flight stabilization (same calculus being run continuously from sensors through near-instantaneous output)
It interesting to see the "new chip technologies" to be commercialization of very expensive and proven military techniques from decades ago.
RADAR to visual film was a great application of laser processors too, that goes back to the 1960's.....
"Not new though
many jet fighters from 30 years ago use optical processing to achieve flight stabilization"
For example patent #5093802 publicly available from the US Patent office from 1989.
Just the public patents show tech that is decades ahead of what the generally clueless population believes is new.
AND the patent law has a classified section that is NOT publicly published. When an inventor files a patent the "classified section" decides IF it is to be classified "secret, top secret etc"
The inventor is then made an offer "they can't refuse" and if those inventors have a problem with it......Malaysia Airlines Flight 370.
The hard part is that little red circle. Interference is relatively straightforward but if you don't do anything else you run into problems since light is linear so the smallest mismatch in the interfering signals can reveal a strong field far down the line.
I presume they are either leaving the pure optical realm or using some unusual non-linear effect in the red circle but what is it?
The speed limit is the round-trip time within that circle. If it is a hundred wavelengths long, the 1/1,000,000,000,000,000 s in the video's icon would correspond to a wavelength of the order of 3 nm. Off by two orders of magnitude.
The nonlinearity is probably a change in the index of refraction depending on light intensity.
@@rainaldkoch9093 Sorry, what is off by 2 orders of magnitude?
@@petergerdes1094 The switching time is not 1 fs = 1/1,000,000,000,000,000 s but of the order of 100 fs, at best.
@@rainaldkoch9093 Ok, but who said it was? Did she say it in the video and I missed it? I was just a bit confused bc it sounded like you thought I said that.
@@petergerdes1094 1:11
I appreciate the research needed to communicate this in a way that I can understand. Thank you. Your channel is one of my favourites!
I've been hearing about photonic chips for 40 years now, so it's about time.
I really appreciate all the effort you put into understanding the topics in order to distil a compact summary. Many thanks.
I'm glad I found this channel. A friend of mine told me > 25 years ago that chips will be using light at some point. It made sense. Now here we are
At some point in 2070-2080 when we will be dead. This technology matures too slow.
This is insane, you explain it so well too! this is by far one of my favorite channels now, you rock!
Inside a fibre optics, light travels at 2/3 the speed of light in the vacuum. While this is indeed quite fast, it is not close to the fastest way to transfer information on earth. One way that it is faster is using mmW.
Note: This later method has the drawback that there is a need for line of sight between the two ends, so it would not be a good replacement of the existing fibre optics network (and I think that not that many users care about the difference in latency).
I'm in the semiconductor industry (over 20 years) and this is fascinating! Thank you!
Classic computing has kind of been stuck on a certain plateau for a while now and we need a big breakthrough to move on to the next thing. Photonic computing is probably 10 years away for practical home applications, but it's exciting to see the first steps.
I'm sure that I don't understand like I should, like being able to explain what you've said to an interested party but you are so BEAUTIFUL that i cant wait for the next one to drop.
Excited for light based computers :)
Anastasi, Thankyou for your insights into computer chips.
It's wonderful to be kept on top of leading edge development, You're doing an outstanding effort with this, Anastasi..Thank you so much and I'm always looking forward for your next delivery. 👌
Appreciate your hard work Anastasi, I learn so much from your videos. Thank you.
Anastasi, you have a good mind and you're very talented at making unimaginable complexity understandable for the average person. It also doesn't hurt that you so obviously love this field.
I always enjoy hearing and seeing your intelligent deconstruction of advanced technologies.
You're like the audience whistle-blower who reveals the how-I-did-it of magician's tricks!
Not everyone has the skillset to pull this off but you do it - routinely!
Every one of your presentations leaves me feeling a little smarter than I was before watching you.
Thank you for your hard work and effort - it shows!
I always look forward to your videos and I'm always glad that I watched them!
Keep up the wonderful work!! 😊
Good presentation for someone who knows bit n bit of the working of photonic computing. Thanks to Tsinghua lab for the breakthrough. Hopefully this will break the US n Weat stranglehold on EUV machines in the future
You are awesome! I like everything about this video. Your humor too :) Keep it up
This sounds very promising, thank you for explaining it so well 😊
Yes, please, make more videos on photoinc computing.
your approach to content is so inspiring, keep up the great work!
Hello, its been a long time that this topic has been on my mind? Thank you for your updated information 👏👏👍👍
This gives "parallel computing" a whole other dimension! Maybe "simultaneous computing" might be a better description.
One minor thing in the video -- the "a" matrices is pronounced like the "a" in "ate" rather than the "a" in "cat". In IPA terms, it's the /eɪ/ diphthong.
The jokes were really funny and the tongue in cheek commentary on inferring interference was an excellent follow up on the last episode. Additionally, the information was inspiring and I can’t wait to see more.
Awesome video.
Very high potential technology.
Thank you for your time and effort producing this content on photonic computing technology.
Thank you for the explanation and for the layman’s view of the tech.
It will be interesting to see how this will combine with layered chips.
Reminds me of what Optalysis once offered now upgraded to next level. Great vid, thx for sharing😉
I love these videos. Thank you for all the hard work you put into them
Fascinating delve and update into a subject long in the mostly speculative world. Thank you!
Glad you enjoyed it!
this looks like the first useful optical computing chip. thanks a lot for this episode.
The energy reduction aspect is huge! That will save data centres millions of dollars. This is the most important development by far. It really doesn’t matter if these chips never make it into the pc or phone market.
TY for this explanation & creating this video.❤
Absolutely yes, make another video about this technology!❤
4:31 Yes, please!
Yes please make a video about that topic I want to learn and I love learning from you
Excellent explanation of aria in Opera, too, looks interesting.
Thank you for the very competently presented discussion on photonic computing, Anastasi.
My pleasure!
Wonderful. Important details have been presented very well.
Thanks for pointing out the overall energy requirements comparison with semiconductor chips. From an SoC design methodologist.
@anastasiintech Not only do you have a stunningly beautiful mind, but you compliment it with great, and much appreciated humor - I love the inclusion of your cat in showing the REAL value of lasers! :D I believe capturing and training photons to participate in our processing needs is the holy grail of computing... As you say, "1000 times" speed improvement. LOve the vids!
excellent way explaining, i am so glad
Brilliance and beauty combined.
Anastasi too, of course 😄
Unfortunately this will take quite some time till it reaches the consumer market.
How are you Italian and based in Singapore 😮 your accent is amazing and your content is breathtaking, thank you ❤
Laser is a coherent light because only 1 waves go out not necessary focused ;) , the best way to use photonics is in using matrice you can calulate all matrice in 1 times ;)
THANK YOU - your channel is one of the best of TH-cam verry interessting content of high quality even visually verry nice - and your ever lasting smile a real pleasure again THANK YOU
please do a separate episode on it I beg you please.
Very interesting. So much has changed in the industry since I retired 12 years ago, it's hard to keep up. Also I've used Opera for years, solves all the little irritants of Chrome and Firefox, but this is the first time I've seen it advertised.
Splendid: in every possible which way. Very enjoyable instruction, learned a lot, thanks :)
Yes. I want to see a video about quantum computing at room temperature using photonics
One has to start somewhere. Miniaturization takes time. Doing reconfigurable complex calculations at relevantly higher speeds than traditional silicon components has long been a hurdle left for early photonics chips to achieve before we could consider them for general-computing tasks. With this set of new developments, we see a real path forward on these issues - with actual hardware to show for the modeling and preliminary research. That is already a huge leap forward for light-based computing. Working through the problems of scaling gets a huge boost because of the parallelism and speed you rightly highlight. If progress happens on that as quickly as this development did, with this new chip/set, the miniaturization issues will have plenty of length of runway with which to work.
Analogue and photonics are looking increasingly important going forward. Thanks for sharing your impressions of the Tai Chi and the field.
Thanks! I instal Opera from your link.
Fabulously clear, interesting & exciting! So useful to have all these new developments explained & described as the world moves relentlessly towards the remarkable power of AI that has the potential to be an extraordinarily great blessing for all of humanity. Thank you for sharing!
Wooow... well done! So 60 years ago we didn't even have lasers. Now, we compute on photonics. The mind boggles!
Powerfull thank you for the explaine ❤
This is one the best chip videos on TH-cam
Absolutely Fascinating! Thank you for sharing your insights.👍🙂
Very cool topic… great job explaining it!
Thanks for your hard work.
In the future AI will be able to make all the videos on your channel in half a second. Each Unit of Effort (UE) is a valuable commodity, like gold. I sadistically await your next video, knowing that you spent hours and even days making it, with numerous frustrations. In the days of AGI we will pay a premium for that.
This technology is very promising. It is still relatively fringe.
I love this channel so much.
All your are like "the chip that will change everything" and it's true everytime 😮
This is why TwoMinutepaper says "what a time to be aliiiiive"
Thank you for info on the latest advancements - Brilliant! - Cheers from Seattle 🍻 (very much appreciate your hard work)
Math in light, makes perfect sense, same way it works in millimeter wave.
We did some simple math this way with microwave power amps. We had division by a power of 2, followed by multiplication of amplitude, then combination by the same power of 2, and to move frequency we filtered different wavelenths, multiplied two sines to obtain a product of the two which was selectivly filtered to obtain just one of the two products.
The largest problem was controlling the waves so they do not radiate out of the circuit. With fiber optics or coax this is easy but with printed circuits it becomes a problem. But an optic printed circuit will have trouble keeping the light where it is supposed to go.
We were mainly after a larger signal. All the circuit traces are transmission lines of different impedances. Its just the way you look at it. Mostly I tested for best results in a lab, but worked with the chip designers to optimize results much like you are doing, your tech level is not to high for me to understand thanks for the videos! I will watch them all!
This looks very promising! And thanks for the asmr 😊 :)
Awesome, thank you.
Good video, thank you
Augmented reality. At minute 8:02 your background interacts with the animation.
Thanks for your english! I use it for learning.
WOW, awesome, thanks for this video!
Great explanation !!
ngl i love this topic, could listen to it for days :)
I much like the way you explain the principles and mechanics involved in chip construction. Had no need to correct your terminology. Obviously, you know how to explain what you are talking about which is far more important than the exact wording. Having said that, the way you clarified this here was fun! - Very interesting video, thank you.
The laser they used was 1550-nm fiber laser. It outputs few tens of milliwatts on average. Peak power per pulse has nothing to do with laser power consumption at all. You can have megawatts per pulse just because watt is J per second. The shorter the pulse, the greater the power. The laser they used is a tabletop laser with modest consumption. Also, one should not look at consumption numbers for the research-grade instrument. Diode lasers can have their efficiency reaching 70%, especially in the infrared. Then, also keep in mind that the whole power dissipation will occur outside of the chip.
And suddenly I remember everything, whoah. Good video!
your are so smarter .🥰🥰
Thank you so much.👍👍
Well explain
*Note that waves can cancel ONLY in situations that give the wave energy some other place it can go.* This is necessarily true because of the principle of Conservation of Energy. If one wave carried the normal positive energy and the other carried negative energy, then Conservation of Energy would permit cancellation. But *we never see waves carrying NEGATIVE ENERGY.*
Opera, yes: I’m using it right now!
As for miniaturization: the components can never be smaller than
the wavelength of the light they use, isn’t that so? Anyway, THANK you Anastasi!
Díky!
Thank you
Yes, light refractions can be observed from the lens of a flash light lens. Shine the light on a white wall, move the flash light near the wall and you will see black spots on the white wall from the lens reflection. Use phase reflections to cancel light. Got it! See we learn about technology from observing nature.
Light travels fast in optical fibers, but it is still a lot slower than radiowaves in air because the glass of the fibers is slowing the light quite a lot compared to vacuum speed. That is why some very low latency links are not donc with optimal fiber ljnks but radiowave stations.
in 2005 i got carried away with dwdm and spent two years trying to get up to speed on what i thought was coming, aaannnnddd it never happened. so much potential blew my mind and i didn't see the obvious, no one was ready to actualize that potential. oh well, other things happened.
It's like how Nokia's 50G PON was the first to utilize Digital Signal Processing, when everything moves to SOC (Photonics included), BANG!
Goddamn, that's a pretty serious improvement. I look forward to hearing more about those years down the road. Seems like quantum computing at room temp is pretty close now.
outstanding presentation! It feels like we're back in the 40's with the ENIAC computer that took up whole rooms. Except now we have other computers and AI to help miniaturize it. Give it another 10-20 years and it will be the size of a refrigerator and in another 30 years it will be a desktop. 40 and it will be the new iphone.