Тёмный

IBM’s Incredible TrueNorth Chip || Neuromorphic Computing 

NeoScribe
Подписаться 127 тыс.
Просмотров 32 тыс.
50% 1

Наука

Опубликовано:

 

4 окт 2024

Поделиться:

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист
Посмотреть позже
Комментарии : 81   
@DataSmithy
@DataSmithy 4 года назад
Thank you. this is the first time I've heard that new terminology. I will have to keep my eye on "neuromorphic" technology
@pauloneill9880
@pauloneill9880 4 года назад
Spellbinding! Hey Neo who would have thought you could produce something that creates awe from IT. New level reached my friend , bravo!
@neoscribe2295
@neoscribe2295 4 года назад
Thanks so much!
@lowcielva6285
@lowcielva6285 4 года назад
Your video always great. It gave me a bigger picture to see what's happening in the world of technology right now and the future. Keep up the good work 👍🏻
@neoscribe2295
@neoscribe2295 4 года назад
Thanks!
@commieTerminator
@commieTerminator 4 года назад
Since the memory unit isn't separate, this type of technology could be prone to 'forget' data just like a biological brain.
@viggipedia
@viggipedia 4 года назад
I'm thinking more a hybrid of long term CMOS based storage with slower recall speeds. (Kinda analogous to Cache and RAM respectively)
@pwaaoolhtims4344
@pwaaoolhtims4344 3 года назад
Finally got a chance to look at the house and see how it goes and if you want to be a part of the address is it possible to get a better idea of what you want to do that 😉😉😉
@covoeus
@covoeus 3 года назад
The chip could be designed in a way to enable capturing and restoring 'snapshots' of the individual neuron states. It could also be hooked up the traditional storage mediums and 'taught' to use it like you would record data on a piece of paper.
@mickelodiansurname9578
@mickelodiansurname9578 3 года назад
Well yes and no... in a complete system it can of course simply store weights and biases off chip and access them later... a kind of neuromorphic cache stored on a standard digital retrieval system. And storing a machine learning model is just that, it really is just text in the form of weights and biases. But internally? Well yes the processor will immediately 'forget' its previous state on every single iteration ... its forgetting things millions of times a second faster than you can... lol. But ultimately the 'memory' is a system that results from the current state of voltages between axons on the chip... which do change over time yes... so yes, it can forget things, that's a thing. Unlike your brain however, which also has an issue with recall? Well you use paper and pen or a smartphone to remind yourself... the machine can as I said store such things and is even pretty much psychic since it can also rely on the machines around it to help it out. You can't access someone else's head to get a clue about something you never learned... these machine s can of course.
@mickelodiansurname9578
@mickelodiansurname9578 3 года назад
@@covoeus its is actually referred to in machine learning as a 'save state' ... well done.
@tarsierontherun
@tarsierontherun 4 года назад
Dude I missed you!
@neoscribe2295
@neoscribe2295 4 года назад
Thanks! Good to be back, much more to come!
@BitcoinSLU
@BitcoinSLU 4 года назад
facts
@antoniushe
@antoniushe 4 года назад
Nice progress in the editing, interesting topic ~
@neoscribe2295
@neoscribe2295 4 года назад
Thanks!
@kenneth1032
@kenneth1032 4 года назад
Great content as always. Keep safe during this period.
@neoscribe2295
@neoscribe2295 4 года назад
Thanks so much Kenneth. You keep safe as well.
@photovideooz4084
@photovideooz4084 4 года назад
Brainchip Inc has already got the most powerful self-learning Neuromorphic Chip out called Akida, it's available for order via Socionext soon.
@robbyleonard1092
@robbyleonard1092 4 года назад
Happy you're back man
@neoscribe2295
@neoscribe2295 4 года назад
Thanks Robby!
@TheChangeYT
@TheChangeYT 4 года назад
Solving neuromorphic architecture is one key to the future
@AMR-bf8nx
@AMR-bf8nx 4 года назад
Why don't create a supercomputer with thousands of these chips for mimicking a whole human brain? We can call it Heuristically Programmed Algorithmic Computer .
@yashaswikulshreshtha1588
@yashaswikulshreshtha1588 4 года назад
u deserve more likes and views than anyone else's video on this topic. Respect for U keep it up
@oscarheggelarssen3394
@oscarheggelarssen3394 Год назад
liked the video, but the background music was a bit intense
@ErnieRMC
@ErnieRMC 4 года назад
Wow!! Terrific video of videos. The bad call you Dad! And the graphics... can you point me to a path of creating them.
@neoscribe2295
@neoscribe2295 4 года назад
Thanks so much for the comment! My visual style is heavily influenced by Nerdwriter1, kaptainkristian and the vox playlist: earworm. The best advice I can give is to just start creating content and you will build your skills over time!
@ErnieRMC
@ErnieRMC 4 года назад
Great advice and help. Thank you. I did a catalog for 20 years and web ms front page web. Now learning wp. Practice site GeeWhiz.store testing fun site. Old site RMConnection.com
@ErnieRMC
@ErnieRMC 4 года назад
Again great video!
@PedramNG
@PedramNG 4 года назад
Amazing!
@defaultHandle1110
@defaultHandle1110 Год назад
IBM has now developed a chemical / magnetic based, memristor type chip for AI applications. It is not transistor based.
@teamsalvation
@teamsalvation 4 года назад
This is awesome!! So, is this "as simple" as replacing a GPU with an MSC160 Board - putting aside the need to re-write the CUDA Kernels into whatever language or waiting for google to update TensorFlow? How does data get into the TrueNorth chips? Can the system be setup to directly stream raw video data, from a video camera, into the MSC160 and then some HW interrupt would trigger "the kernel" to run?
@RynaxAlien
@RynaxAlien 2 года назад
Gaming GPU?
@ijknm2531
@ijknm2531 3 года назад
thare is a companey called brainchip that reacently signed a contract with NASA to send a neuromorphic chip into space. the chip is called akida and its main function is to be implanted into sensors on rovers and the like to make those sensors inherently smart and not rely on a central computer.
@daniel_960_
@daniel_960_ 4 года назад
Great video! Happy to discover a new topic I haven’t heard of yet, especially already in product stage. Just realized I haven’t watched your past 2 videos, will do this now.
@neoscribe2295
@neoscribe2295 4 года назад
Thanks!
@Kaizoku_Official
@Kaizoku_Official 4 года назад
Awesome soundtrack dude 👍
@neoscribe2295
@neoscribe2295 4 года назад
Thanks!
@nekrofilzombi
@nekrofilzombi 3 года назад
I love C+
@dqxva9ep7a-d1h
@dqxva9ep7a-d1h 18 дней назад
"My CPU is a NEURAL NET processor, a learning computer..." When the art/movie mimicks life.
@PassportGaming
@PassportGaming 4 года назад
Can’t wait to see how Tesla’s chip holds up against this beast
@papr3ka295
@papr3ka295 4 года назад
the Tesla V100 SMX2 can do 15.7 FP32 Tflops
@PassportGaming
@PassportGaming 4 года назад
Papr3ka Wow, that’s 5.4 more than the ps5
@frozenstarlight3132
@frozenstarlight3132 3 года назад
Actually, not really. Folding@Home reached above 2 exaflops a few days after this video was posted
@Galomortalbr
@Galomortalbr 4 года назад
and here is born skynet
@abhitechtalk369
@abhitechtalk369 4 года назад
What is the speed of nuron??
@jonathanozik5442
@jonathanozik5442 4 года назад
"Biological neurons can fire around 200 times a second on average."
@amandhingra4947
@amandhingra4947 4 года назад
Sources would make it even better
@anaghyati8110
@anaghyati8110 3 года назад
fascinating
@fred101j7
@fred101j7 4 года назад
what's the flops of a Truenorth chip anyone know?
@sanghaforce2951
@sanghaforce2951 3 года назад
flops calculation method is different for every kind of chip also real world and theoretical values aren't same
@DavidVonR
@DavidVonR 4 года назад
Terminator is coming.
@The_Conspiracy_Analyst
@The_Conspiracy_Analyst 8 месяцев назад
YAWN! This has been done TO DEATH already. It's just asynchronous dataflow architecture. Check out GreenArrays, etc. It's also a basic structure of the GOOD FPGAs and CPLDs. IBM has been known to do this -- make a project to just hype and troll to the extreme for marketing purposes. Anybody seen Watson these days?
@rmt3589
@rmt3589 3 года назад
Can't we just use a C++ Converter of some kind?
@64devil
@64devil 3 года назад
So God should've just kept up simple or woah maybe it was when we ate the fruit
@petersuvara
@petersuvara 4 года назад
Music Credit Please
@plafar7887
@plafar7887 3 года назад
It's pronounced "Von Noiman".
@Vikcreed
@Vikcreed 3 года назад
HOLY SHIT!~
@davidcowan6433
@davidcowan6433 4 года назад
Oh man, that gave me a nerdgasm. 😝
@neoscribe2295
@neoscribe2295 4 года назад
Lol
@michaelandrews4783
@michaelandrews4783 3 года назад
Sure makes Apple computers look like little kids toys
@HardKore5250
@HardKore5250 4 года назад
Pokhi chip is the latest.
@cybercapri
@cybercapri 4 года назад
Yep, Skynet is on the Horizon, Enjoy life while you can... LOL...
@bassplayer807
@bassplayer807 4 года назад
I’ve always wanted to get into A.I/ Robotics as of late, I’m 27 now but don’t wanna work in food manufacturing because it’s a dead end job, in the near future, A.I could replace my job by 2030, and that’s not a doubt, I think by 2030 the Mckinsey report said that 800 Million jobs would be lost at the start of that decade by advancements in automation/ robotics/ A.I, but by 2022 100 Million new jobs will emerge from A.I/ data science, I’ve started to learn C programming for robotics and python for A.I this month
@cybercapri
@cybercapri 4 года назад
@@bassplayer807 I started working with PCs back in 1986 and shortly thereafter A.I. was a thing, I believe it was a thing before that. ELSA was the A.I. at the time for PC while Apple introduced Lisa. A.I and SciFi go way back but given the newer designs of CPUs and GPUs A.I. are more practical these days then decades ago. C seems to have a serious future given its overall age so you are wise to pick it up now; python also has a solid future. Unfortunately for me Basic, Fortran and a few others were quickly phased out while C just kept chugging along and I never got into C as deep as I should have. Automation and/or A.I. will be the new slave labor of the future and will cut costs dramatically and also require less manual labor but there will always be some form of human labor to maintain the robots until they become fully aware and all bets are off then... LOL... Too bad Corporations turn excess profit into GREEDY PROFITS rather than reducing the product price. Greed is the worst drug the Earth has and those controlled by it could not spend their wealth if they lived 5 lifetimes yet GREED causes them to obtain MORE... Take Big Oil for example, did you know their profit margin is 85 Trillion Dollars per year? Big Oil could pay off the US Debt and buy every American a 50,000 USD Vehicle and still profit over 20 Trillion Dollars each year. Imagine what it would be like to get a new 50k vehicle for FREE every year, how nice would life be... And by every American I mean every Man, Woman, and Child. NOPE Big Oil HORDS their Profits rather than give back to those that make them rich but I digress... Cheers...
@__-tz6xx
@__-tz6xx 4 года назад
Congratulations! You are incredibly basic. You said the obvious comment, the comment which plagues all of the comment sections of almost every video about Artificial Intelligence.
@Machiavelli2pc
@Machiavelli2pc 4 года назад
Exactly why Neuralink exists. I direct hedge against Skynet.
@wolf359loki
@wolf359loki 4 года назад
Too bad our Memory is RAM and as soon as power is lost so is the data.....
@dokerb3d60
@dokerb3d60 3 года назад
ok humans do this 0.93625*0.123553 that is 1 flop, supercomputer can do it 200000000000000000 times per second. yes brain are smart lol
@_jw_harper
@_jw_harper 3 года назад
#milesbennetdyson
@pauloneill9880
@pauloneill9880 4 года назад
4 thumbs down? LUDDITES!
@SerPapus
@SerPapus 4 года назад
The reason why humans brain don’t use a lot energy is because we probably game something up while evolution was going on.
@runneypo
@runneypo 4 года назад
more likely because if the brain used any more energy it would be extremely difficult to gather enough fuel to survive on a day to day basis
@SerPapus
@SerPapus 4 года назад
runneypo it probably learn to use it efficiently but fuck how did we become self aware is crazy to me
@emailphone4728
@emailphone4728 4 года назад
Like the video, weird music choice though...
@AndrewMSconfig
@AndrewMSconfig 4 года назад
Music choice is awesome :D
@emailphone4728
@emailphone4728 4 года назад
AndrewMSconfig I like your positivity :P
Далее
End of the silicon era. Processors of the future
19:26
Просмотров 322 тыс.
Neuromorphic: BRAINLIKE Computers
35:52
Просмотров 91 тыс.
skibidi toilet multiverse 042 Trailer
01:57
Просмотров 2,8 млн
Why Brain-like Computers Are Hard
17:44
Просмотров 239 тыс.
The AI Hardware Problem
13:26
Просмотров 529 тыс.
comparing GPUs to CPUs isn't fair
6:30
Просмотров 290 тыс.
Brain-Like (Neuromorphic) Computing - Computerphile
13:58
Analog Computing is GENIUS - Here's Why!
15:28
Просмотров 483 тыс.
When you Accidentally Compromise every CPU on Earth
15:59
The TrueNorth Journey: 2008 - 2018
15:35
Просмотров 3,7 тыс.