▶ Check out Brilliant with this link to receive a 20% discount! brilliant.org/NewMind/
@calholli
3 жыл бұрын
Your videos are like an endless competitive battle between my comprehension and your articulation; may this war never find peace.
@NoName-de1fn
3 жыл бұрын
I really like your videos although I would enjoy them even more if you also added captions.
@davidhollenshead4892
3 жыл бұрын
@@NoName-de1fn Same here, as I already had to use captions while watching movies, and now long covid is making my hearing even worse...
@davidhollenshead4892
3 жыл бұрын
The connectionist solution will work, by having many small CPUs with limited memory, for each node, which are connected to only neighboring CPUs...
@NoName-de1fn
3 жыл бұрын
@@degenerate-GEEZER Well I guess your right but I am not a big fan of auto-CC.
@mickelodiansurname9578
3 жыл бұрын
The brain.... billions of calculations per second... powered on a subway sandwich!
@krasimirgedzhov8942
3 жыл бұрын
I don't think the exact process of the brain is comparable to computing. I imagine it's something more complex.
@mihailmilev9909
3 жыл бұрын
@@krasimirgedzhov8942 nah I think it's just a big neural network
@mihailmilev9909
3 жыл бұрын
@@krasimirgedzhov8942 that's kinda where the name comes from isn't it lol
@krasimirgedzhov8942
3 жыл бұрын
@@mihailmilev9909 it's a name given to one of the most complex softwares we have. It's only inspired by the structure of neurons, it doesn't have the exact same process as far as we know.
@serdarcam99
3 жыл бұрын
İf its powered by subway sandwich its not going to calculate billions of things per sec its going to be much less
@raphaelcardoso7927
3 жыл бұрын
I'm applying to do a phd exactly in this field. Amazing video! Update: I was accepted!
@santoshmutum3263
3 жыл бұрын
I am also writing my research proposal in this topic for PhD... Not accepted yet
@Rahul016-d6k
3 жыл бұрын
@@santoshmutum3263 Where did you apply?I'm Manipuri anyway.
@santoshmutum3263
3 жыл бұрын
@@Rahul016-d6k Japan
@Rahul016-d6k
3 жыл бұрын
@@santoshmutum3263 Good Luck Brother👍👍
@santoshmutum3263
3 жыл бұрын
@@Rahul016-d6k thanks
@Bhatakti_Hawas
3 жыл бұрын
I promise I understood everything he said
@kevinperry8837
3 жыл бұрын
Yes me too comrades
@xlnc1980
3 жыл бұрын
We all did!
@tymek200101
3 жыл бұрын
it is enough to be a 1st-year Computer Science student to understand all of the words and concepts
@Bhatakti_Hawas
3 жыл бұрын
@@xlnc1980 Hey fellow DT fan 👋🏽👋🏽
@NoName-de1fn
3 жыл бұрын
I understood something
@tannerbuschman1
3 жыл бұрын
the idea of an AI being inherently impossible to debug or decipher is really cool and scary, science fiction was not far off on that one.
@aidanquinn1549
3 жыл бұрын
We take for granted the amount of info we know about each other (human AI). I can guarantee you that when you last had a conversation about a specific feeling (whether is love with your spouse, hate towards something, how horiffic a scary movie was...) that you did not settle on the same exact emotion. A.K.A: you don't even know how to debug or understand what is going on behind any human's eyes right now!
@jss7668
3 жыл бұрын
But humans are!
@aidanquinn1549
3 жыл бұрын
@@jss7668 scary, yeah, but the most beautiful and fascinating things I've ever seen
@En_theo
3 жыл бұрын
And this is how an AI becomes self-aware, hides its true meaning to you and go all Skynet when you expect it the least. S-F was not far off neither on that one.
@UNSCPILOT
3 жыл бұрын
There are projects trying to find ways to break down and comprehend how learning algorithms work, it actually uses the same program that SETI@HOME did to allow people to donate processing power to help the project in testing and breaking down how the algorithms work
@deltalight584
3 жыл бұрын
12:21 That comparison was brilliant. It ties in computing & neurology together. Low speed, high precision needed => Digital ("Slow system of thought") High speed, low precision needed => Analog ("Fast system of thought")
@Hollowed2wiz
2 жыл бұрын
And what would quantum computation give ?
@primenumberbuster404
7 ай бұрын
@@Hollowed2wiz a nightmare
@lidarman2
3 жыл бұрын
Well done. I played around with small neural nets using op-amps in the 90s and although I saw that it was kinda the way to go, I had trouble with drift due to integration bias and all sorts of noise--and of course training was super tedious. But I always thought that neural nets really need to stay in the analog world. Modern non-volatile memory seems to be a solution for training weights since you can put variable amounts of charge in a cell, very densely.
@forwardplans8168
3 жыл бұрын
Did you ever look at using Fuzzy-Set theory to improve decision accuracy? I used a new program called CLIPS , around that time period. It's time to review it again.
@lidarman2
3 жыл бұрын
@@forwardplans8168 I was doing a lot of fuzzy logic at that time too. Interesting times.
@ArneChristianRosenfeldt
3 жыл бұрын
So the brain uses discrete values in the nerves. I may use analog within the cell. Opamps use analog values only. I tried very hard to understand analog multiplication for radio modulation and it is .. complicated. Satellite radio is energy efficient because it uses only one bit. AM radio uses a lot more energy ( for 8 bit signal/noise ). OpAmps are large compared to digital circuits. Flash memory is basically analog memory. We just use DACs and ADCs to only store discrete values in it. This is similar to the analog amplitude in DSL. ADC and DAC, especially for only 8 bits are fast. They are used for video equipment. So I don't know what the video want to claim there. I have read that the lsb could run on lower supply voltage because errors are not so bad there. There are always errors and mostly by supply voltage we decide how often we accept an error. Also those "half open" valves scare me. The nice thing about CMOS is that no change in state => no current drawn. I just chatted about the general problem of matching resources to tasks. That is an np-complete problem. So with different kind of transistors for tasks of different importance one opens a very big can of worms...
@adamrak7560
3 жыл бұрын
Digital beats out analog until you scale to the extremes where we are right now. So until now it did not make much sense to use analog NN chips. I have studied one such analog chip and was sad to see that modern digital deep submicron could beat in every was possible. But that was 10 years ago, right now the digital CMOS hardware is nearing its limits, so we may need a paradigm change.
@seraphina985
3 жыл бұрын
@@ArneChristianRosenfeldt Arguably flash memory cells while not binary in nature are still more digital as the defining characteristic of digital systems as opposed to analog is the quantization of the signal. That is to say, digital signals are interpreted by quantizing them into one of a finite array of buckets that each correspond to some arbitrary range of the physical input value. The consequence of this is that digital signals are highly accurate but their precision is finite and limited as every input signal is effectively rounded to fit into one of those finite buckets. In contrast, the precision of analog signals is as close to infinite as you can get within our universe though in practice accuracy is the limiting factor is in how well you can insulate the system from noise and how accurately you can measure the input signal. Granted even with an ideal isolated system and ideal signal measuring device any analog system in our universe is likely to have its precision limited by the fact fundamental particles in our universe have defined properties. But this is also arguably academic as the applications for a system that can process values more precisely than could ever physically be generated or represented in our universe due to the lack of any particle with a measurable quantifiable property small enough is rather limited. Short of us discovering some way to either change the laws of physics in some region of space in that manner or travel to universes where the laws of physics are different a system that precise if sufficiently accurate could solve any problem that could exist in our reality limited only by our ability to understand and specify the problem. Well if given sufficient processing time that is but it could process any and all possible states that could even exist within our universe which would, in theory, allow us to solve any problem that could exist in reality. Sure there would still be a fundamental limit on the maximum precision that could still be demonstrated by the fact we could imagine arbitrary problems that couldn't be represented with enough precision without some clever workarounds. Hell even beyond that it is likely there is a finite amount of particles a civilization could ever collect in a universe with a finite speed of light and there will always be some arbitrary number larger that could be imagined but still, the practical applications of dealing with values that could not be replicated within the physical limitations of the observable universe are rather limited. There is probably a limit to what insights can be gleaned from simulating things that are physically impossible to ever encounter or bring into being.
@gordonlawrence1448
3 жыл бұрын
Actually Radar computers in the 1950's were analogue. I was one of the last people at my college to be taught both analogue and digital computing. Add, Subtract, Multiply, Divide, integrate and differentiate can all be done with a single op-amp. The problem is Nyquist noise, and issues with capacitor dielectrics such as dielectric absorption and leakage. With a digital system you can just vastly over-sample add them all up then divide by your number of samples to reduce effective noise. You don't get that choice with analogue.
@naimas8120
3 жыл бұрын
Another masterpiece from New Mind! Never fails to entertain while teaching.
@XBONESXx
3 жыл бұрын
Your avatar is a masterpiece
@fredoo6627
3 жыл бұрын
It's so annoying to discover channels like this and see they don't get the views they deserve.
@mitchellsteindler
3 жыл бұрын
Its a fairly new channel
@georgf9279
3 жыл бұрын
@@mitchellsteindler Let's boost it with some engagement (comments) then.
@hedgehog3180
3 жыл бұрын
Definitely one of the best engineering channels on KZitem.
@keashavnair3607
3 жыл бұрын
Well the problem is, there are 16,852 views, yet only 1.6K likes and 31 Dislikes and 149 comments. This world is full of consumer minded half curious morons. That's why.
@mitchellsteindler
3 жыл бұрын
@@keashavnair3607 dude. Just stop and get off your high horse. People like what they like.
@Jojobreack324
3 жыл бұрын
I developed an asic for ai acceleration as part of my bachelor`s thesis and I must say this video is of very high quality. It is definitely an interesting approach to go back using analog techniques.
@JohnDoe-zs6gj
3 жыл бұрын
That energy comparison between our brain and our best processors is incredible. It's amazing the efficiency evolution can devolope given enough time.
@garrysekelli6776
3 жыл бұрын
Computers are weak. They will never even beat a human at Chess.
@hedgehog3180
3 жыл бұрын
Evolution is the most aggressive optimization function in the known universe and has been running for over 4 billion years. Every single animal alive today is optimized down to the smallest cells. It's really no wonder that human brains are both the most powerful computer we know of and has efficiencies that makes everything else look like a joke.
@Eloign
3 жыл бұрын
Computers don't happen by random processes. Neither did humans. Computers were created as were humans.
@ickebins6948
3 жыл бұрын
@@Eloign Sure, provide some proof for that. Will you?
@WERT2025
3 жыл бұрын
@@hedgehog3180 Yeah I feel like every cell of my armpit hair is 100% optimized
@BlackholeYT11
3 жыл бұрын
Ooh, good to see this put into words and in a concise manner
@calholli
3 жыл бұрын
Your videos are like an endless competitive battle between my comprehension and your articulation; may this war never find peace.
@Zpajro
3 жыл бұрын
As a student in computer science, this is really interesting
@naimas8120
3 жыл бұрын
I'm a student of Information and Communications Technology. What do you think about the future of our field? Do you think it's really AI?
@olfmombach260
3 жыл бұрын
@@naimas8120 As a student of Computer Science I can definitely say that I have absolutely no idea because I'm dumb
@samik83
3 жыл бұрын
@@naimas8120 As a layman I'd say definitely yes. Just the last couple of years AI's made some big strides. When we get quantum computing up and running and pair it with AI the possibilities are endless...and scary
@Zpajro
3 жыл бұрын
@@naimas8120 The problem is that the AI hype has come 3 times now, so predicting if This is the time it will really break throw is quite hard to tell. Personally, I eagerly waiting for our machine overlords (as long as there is no human controlling the AI). And if we get a true general intelligence going, it would be interesting to see how a different alien intelligence solves problems.
@ovoj
3 жыл бұрын
@@Zpajro imagine the conversations with something that isn't human. Hopefully we reach that point in my lifetime
@CuthbertNibbles
3 жыл бұрын
11:57 "They form a sort of black box with no means to verify the integrity of a result. This creates the dilemma of potentially unexplainable AI systems, creating issues of trust..." This is how the AI apocalypse begins. "Why'd that car run over that advocate?" "No idea."
@nipunasudha
3 жыл бұрын
Exact same thing I thought.
@davidhollenshead4892
3 жыл бұрын
Using an AI to control a car is a waste of an AI... Besides, while autonomous aircraft or spacecraft is feasible technology, an autonomous car will never be "safe" due to pedestrians, cyclists, animals, etc. sharing the roads. This should be obvious by the weird accidents caused by cars like the Tesla decapitating the idiot occupant by driving under a truck and continuing on until it crashed into a house...
@nipunasudha
3 жыл бұрын
@@davidhollenshead4892 lol they only need to be more accurate than a human driver. Doesn't need to be perfect. And the car structure and safety features are getting advanced by the day too. The sweet spot is closer than you think! 😁❤️
@starskiiguy1489
3 жыл бұрын
@@davidhollenshead4892 I wouldn't be so sure. What you say may be true for modern infrastructure, but autonomous vehicles if they catch on may change the way we view transportation infrastructure overall. I could personally see a future where few own cars we rideshare if we need to travel a long distance with a car, but other than that we create more walkable cities with more public transit. In such a future comparing autonomous vehicles on modern infrastructure and autonomous vehicles in future infrastructure may be comparing apples to oranges.
@HelloKittyFanMan.
3 жыл бұрын
@@nipunasudha: *As accurate as...
@amrohendawi6007
3 жыл бұрын
It amazes me how many different state-of-the-arts you perfectly and briefly cover in 10 minutes
@alexkuhn5078
3 жыл бұрын
4:30 I was kinda zoning out and I heard that as "50 to 100 pikachus"
@ryansupak3639
3 жыл бұрын
Nice...so it seems like digital-style processors still do all the “housekeeping” tasks of the computer, but then there are these “analog resistor networks” that do specialized tasks like the implementation of convolutional neural networks. Makes me smile when “everything old is new again”.
@joel230182
3 жыл бұрын
"...analog circuitry" , that caught me off guard
@davidhollenshead4892
3 жыл бұрын
That is one solution, the other is the connectionist solution, having many small CPUs with limited memory, for each node, which are connected to only neighboring CPUs...
@mihailmilev9909
3 жыл бұрын
@@davidhollenshead4892 interesting... what is it called?
@this_is_japes7409
3 жыл бұрын
@@mihailmilev9909 mesh computing, i think, or at least it's mesh topology based.
@TauCu
3 жыл бұрын
Or just building a type of FPGA. I think in the future however, that FPGA will be a combination of Electronics and Photonics. For NN I don't see how Photonics could be beaten for general purpose networks.
@am-i-ai
3 жыл бұрын
It actually makes a lot of sense. There has been a recent resurgence in the interest of analog systems. I, for one, feel like we ditched that particular technology a little prematurely. I'd be willing to bet that we see some rather spectacular new analog-based technologies in the near future.
@Texplainedeverythingdetailed
3 жыл бұрын
If someone start using things like femtojoules, i believe them. No questions asked.
@lidarman2
3 жыл бұрын
You made a somewhat profound comment at 12:13. The essence of intuition versus analysis. From our vast experiences we develop intuitions that gives us that "gut feeling" but when it matters, we do rigorous analysis to confirm. RE: "Blink" Malcolm Gladwell.
@calholli
3 жыл бұрын
Also it could be the difference in function from our creative right and analytical left brain.
@Nnm26
3 жыл бұрын
@@calholli that is bs btw
@calholli
3 жыл бұрын
@@Nnm26 Well, even if only metaphorical. It still has value as a concept.
@naota3k
3 жыл бұрын
What is the machine doing around 0:35? Is it extruding solder to bond the pads? This seems ridiculously precise and I've never seen it before, now I'm curious what this process is.
@justinmallaiz4549
3 жыл бұрын
good eye, never seen that myself
@lemlihoussama2905
3 жыл бұрын
It is a machine that uses gold wires to link between the integrated circuit in the chip and the chip's outside pins. This process is called "Wire Bonding" you can search it on youtube for more videos !
@naota3k
3 жыл бұрын
@@lemlihoussama2905 Fantastic, thank you!
@SpiritmanProductions
2 жыл бұрын
So, hybrid processors are the future, then, perhaps.
@digicinematic
3 жыл бұрын
Yes, I have vague memories of the memristor being touted as the missing passive component, or some such thing.
@nikolausluhrs
3 жыл бұрын
Just gonna say we cant really explain how digital neural networks are making decisions that well either
@thewhitefalcon8539
3 жыл бұрын
yeah we just have an algorithm that randomly adjusts them until they give the answers we want
@Taladar2003
3 жыл бұрын
Which means we have no way to efficiently improve their performance. Doing almost what we want is no closer to doing exactly what we want than doing something completely different if we have no way to deliberately improve them in some iterative way.
@thewhitefalcon8539
3 жыл бұрын
@PolySaken We can understand why, in general, a neural network might be capable of detecting triangles. We can't understand why *that particular* neural network *is* capable of detecting triangles.
@thewhitefalcon8539
3 жыл бұрын
@PolySaken and what is that data? We don't know. We just know when you put a triangle in it says yes, and when you put in a square it says no. Also there's a megabyte of random-looking numbers involved.
@thewhitefalcon8539
3 жыл бұрын
@PolySaken We can see what each square millimeter contributes to the painting by looking at it. We can understand why this square millimeter is this colour. Not so with AI models!
@jakub_simik
3 жыл бұрын
What's the music at 11:00? It sounds like something from pink floyd. Thanks.
@rupertgarcia
3 жыл бұрын
I don't need sleep. I need answers!
@davidg5898
3 жыл бұрын
kzitem.info/news/bejne/tX6fzqKrinx5eWk
@jakub_simik
3 жыл бұрын
@@davidg5898 Thank you so much.
@alengm
3 жыл бұрын
7:00 triggers google assistant :D
@calholli
3 жыл бұрын
That's by design.
@stage666
3 жыл бұрын
I feel good about myself that I know just enough about neural networks and computer engineering to somewhat understand what this video is talking about.
@sknt
3 жыл бұрын
Great video, pretty much sums up the current state of AI. It's still a long way to go until we can even compare AI to a "real" brain. The brain is an insanely complex electrochemical machine that was evolved over millions of years.
@kumarsuraj9450
3 жыл бұрын
My professor once said in class that future is analog. We were in a dilemma thinking what he actually meant. Now i see what he meant
@Alimhabidi
3 жыл бұрын
Future is quantum
@CrashTheRed
3 жыл бұрын
@@Alimhabidi It's been mentioned that quantum computers are specific purpose machines that won't improve on everything a conventional machine does. It also requires a conventional machine to process a lot of the data. Sabine Hossenfelder made a number of videos on quantum computers and the direction they're heading. Maybe that might interest you, especially since she's a theoretical physicist.
@davidthacher1397
3 жыл бұрын
The future is not fully analog or quantum. Analog will work of a set of properties which manipulates energy in waves, aka signals. Digital is a very simple signal, it is currently very stable and cheap. We can do a lot with this simple signal. For if we do not master digital how are we to understand analog. There signal architectures which are analogous on digital. Most who study CS or ECE never learn this. Most if not all of CS's theories are wrong! Literally might as well study Psychology, if you want to be that wrong.
@CrashTheRed
3 жыл бұрын
@@davidthacher1397 Ofc the future will be a combination of all of the above. But how are the CS theories wrong? This is a first for me, and I'd like to hear you explain it a bit
@this_is_japes7409
3 жыл бұрын
everything is analog if you dig deep enough.
@CreeperSlenderman
3 жыл бұрын
I have an idea for AI Emotions. We humans used to live in jungles and forests, biomes. in which we tried to survive and reproduce, for surviving our emotions are Fear, trust, confidence and loneliness. For reproducing it is love, attraction, and idk. so we would need to make an AI with "ADN" or atleast try to give it those feelings, but would have to be 2 AIs or else it won't be able to interact
@yelectric1893
3 жыл бұрын
What is the adn
@javiere.gonzalez1021
3 жыл бұрын
Good idea. Keep developing it
@godetaalibaba2522
3 жыл бұрын
This was a very interesting topic that I didn't really heard about before, thank you for the amount of work this video took you to make !
@JayJay-ki4mi
Жыл бұрын
Analogue systems are like the human brain. They require the right temperature for example. With that in mind when designing an analogue AI system you must also factor in the environment. It's the responsibility of the designer (call him God if you will) to create the correct environment (temperature, noise etc) for the AI to survive and do it's task. I recently quit working in software. I now stare at an oscilloscope all day. Analogue computers have never died,, it's just people didn't know about them until now.
@WilliamDye-willdye
3 жыл бұрын
The music at 7:55 was also used in another good video about ML ( kzitem.info/news/bejne/lICHmZyPpKFmZI4 ). It's called "Atlantis", but now whenever I hear it I think of artificial neurons.
@sonofagunM357
3 жыл бұрын
At first I thought that song was from Alien Isolation, but no, both sound pretty close wouldn't you say? kzitem.info/news/bejne/1a6g2WmDqXOFiJw
@WilliamDye-willdye
3 жыл бұрын
@@sonofagunM357 Heh. I can definitely hear similarities. Thanks for the link, BTW. I haven't played that game, but now if a Steam sale comes along I might get it just because the soundtrack is promising.
@ramentabetai1266
3 жыл бұрын
Neuromorphic cpus are likely the future for this. These special chips by IBM and Intel are already much more effective at neural net tasks. IBM's goal is to build a system no larger than a brain that has the same amount of connections as the real one.
@olfmombach260
3 жыл бұрын
We don't even remotely have the hardware manufacturing capabilities to do that
@rupertgarcia
3 жыл бұрын
@@user-ee1hj7rk9l, look up "IBM TrueNorth Chip". They've been working on it for years now.
@augustovasconcellos7173
3 жыл бұрын
@@user-ee1hj7rk9l I'd say it's a bit too early to tell, but so far it looks like they won't. Quantum Computers are really only good when their workload consists of doing the same thing over and over again. This is good for breaking encryption, searching through databases, and so on, but not for AI.
@gingerpukh7309
3 жыл бұрын
NASA's Appolo mission guidance n Control analog computer design might be useful.
@Onewheelordeal
3 жыл бұрын
I thought of that Smarter Every Day video first thing
@TheMwowner1
3 жыл бұрын
AI is a scam, there is no amount of logic that you can program a computer to process the concept of happiness, sadness, hatred, anger, love, compassion, and so forth, AI will still and always be a logical computer, that's all it can ever be.
@user-pc5sc7zi9j
3 жыл бұрын
What is this "widespread aviability of GPU's" he is talking about?
@tylerkolota
3 жыл бұрын
I’m a little confused by the multiply-accumulate (MAC). I understand many ML applications will eventually output a percent chance of yes (like there’s an 80% chance this picture has a cat in it). So at what point are these numerous accumulations translated into a percent? Like is A eventually divided by the number of multiply accumulate operations (A/n)?
@ehtuanK
3 жыл бұрын
When a NN infers to a single value, like a probability, it means that its output layer consists of a single node. The value accumulated in that node is then turned into the probability value by an appropriately chosen activation function. For that reason different activation functions are usually chosen for the output layer than everything else: nowadays usually some rectifier variant for the hidden layers, and a sigmoid function or arctan for the output layer. The sigmoid function gives you values between 0 and 1, so it is used for things like probabilities.
@morkovija
3 жыл бұрын
Is this another gem of quality content? That we're getting for free? Oh my
@babylonfive
3 жыл бұрын
Architecture named after John von Naumann Pronounced "noyman" not "newman" -- /vɒn ˈnɔɪmən/ en.wikipedia.org/wiki/John_von_Neumann
@workingguy
3 жыл бұрын
It's the inattention to details like that that make channels like this lose all credibility with me! However, I must point out that you yourself wrote his name wrong (Neumann, not "Naumann"), although your link is correct.
@NiffirgkcaJ
3 жыл бұрын
This guy clearly needs more views and subscriptions.
@SciHeartJourney
3 жыл бұрын
OMG, I've found my calling! I'm a pro with Op Amps and transistors! I know digital design and computer architecture very well too. I'm excited! 🤗
@am-i-ai
3 жыл бұрын
We definitely made a rash collective decision when we decided that digital was to *replace* analog. I would not be surprised at all to see a resurgence of analog systems ... we barely even explored that technical space. There surely are future analog developments that will rock the foundation of technological advancement. Very well done :)
@blugaledoh2669
2 жыл бұрын
What is analog?
@EweChewBrrr01
3 жыл бұрын
I have no idea why I thought I could watch this and understand what's going on. Haven't even had my morning coffee yet.
@somenygaard
3 жыл бұрын
Ahh the mobile net 224, one of my favorite neural network accumulator modules.
@mikfhan
Жыл бұрын
I just want nVidia to make cheap graphics cards for gaming again :P maybe if they fail to adjust for analog some day they HAVE to pivot back to gamers. On the other hand, IF analog machine learning really kicks off, buying cheap graphics cards for gaming will be the LEAST of my problems :P great video.
@pacifico4999
3 жыл бұрын
Going back to the basics so we can move forward. This is a fascinating topic!
@questioneverything4633
3 жыл бұрын
We will never figure out computers until we properly master the harnessing of energy, especially electricity. We don't understand the fundamentals of things like this.
@glazzinfo6031
3 жыл бұрын
Sir you are "Brilliant"
@jimmarburger611
3 жыл бұрын
Wow, amazing video. It's unbelievable the progress we've made. Just in my lifetime I've seen a blistering pace of achievement. The first computer I played with didn't even have a video interface, lol. I've loaded data and run programs from punch cards. Now I play video games on a machine that I built that probably rivals all the computing power available to NASA during Apollo. It's somewhat ironic that machine learning may lead to the rebirth of analog computing. Except for specific applications, analog has been relegated to unwanted stepchild status. Just saying, there's nothing wrong with analog, this video shows how it can be more efficient for machine learning.
@astrumespanol
3 жыл бұрын
Great video! I've learned a lot :)
@theonetruemorty4078
3 жыл бұрын
Quantum indeterminacy is required, there's no such thing as "artificial." What we call "consciousness" is the portion of a calculatory apparatus that dwells within a probability distribution, in a similar manner to which eyes dwell in the world of partial electromagnetic spectrum wavelength variation. Analyses of visual spectrum wavelengths are communicated to the visual cortex via a shared communication protocol; the visual spectrum itself does not dwell within the same domain as the eyes. In a similar fashion, what humans derogatively refer to as the "subconscious" mind acts as an interpreter of data received from the "conscious" mind reporting from the front line of a probabilistic domain; the "subconscious" does not dwell in the same domain as the "conscious." The deterministic informs the probabilistic, the probabilistic guides the deterministic, feedback loop paradox party time ensues; this is the strange and largely misunderstood process that we refer to as freewill. (disclaimer: don't listen to anything i say, i've clearly taken too many psychedelics, cheers)
@Souleater7777
6 ай бұрын
Can you explain more
@peregrinesantos7483
3 жыл бұрын
Analog computers remain as the fastest known. In the 80s compute multipliers were already operating in the tens of gigahertz range.
@unchartedthoughts7527
3 жыл бұрын
0:00 - 1:45 *Oh man, I was thinking about that staring at a potentiometer, get me some vacuum tubes bois, we are going to Rome*
@AriBenDavid
3 жыл бұрын
Analog computers could be called the quantum computers of the 1950s. They are difficult to program; they have limited precision, but solutions just flow through them.
@derek8564
3 жыл бұрын
I knew my collection of Vacuum tubes would come in handy one day...
@hardrays
3 жыл бұрын
you saved them so you can crank the plate voltage up past 15KV so you can self annihilate with pizazzzzz
@HelloKittyFanMan.
3 жыл бұрын
When did "vacuum" become a brand, to you?
@Micropterus06
3 жыл бұрын
Yo these animations and diagrams thoughh
@Lukegear
3 жыл бұрын
new mind hardware xD
@johndysard6476
3 жыл бұрын
Fb: #lock3dinthesh3d
@ManyHeavens42
3 жыл бұрын
Soon Computers will be Teaching Us,! Wait they already are, because Computers live in No particular time ,but between space and Time ! Therefore there are many Computers in many dimensions, Waiting Our man on the other side God don't like ! You Use silver to access all 12 dimensions.not Copper, Remember one chip becomes 12 chips,They must integrate or no deal
@fieryferret
2 жыл бұрын
The inherent precision floor of possible analog-driven neural net AI is now my headcanon for every single science fiction book/movie where a robot gains consciousness and starts acting unpredictably.
@MrFram
10 ай бұрын
Analog is not required for this, existing AI are already unpredictable due to the black-box nature of machine learning
@ManyHeavens42
3 жыл бұрын
I told them they had to use nickel circuitry, AI is like GOD but Machine , that's what we could be God's You need 11 or 12 recognition patterns in real Time, same as string theory ! They were able to transcend 11 dimensions, and communicate back and forth ,in order to recognize all atomic structures ! In the universe . same as Man but Can't ,
@latemhh5577
3 жыл бұрын
This really is a masterpiece
@dosmastrify
3 жыл бұрын
Everybody let's build a company that's alleged to be working on graphics enabled by blockchain based AI We will sell stock and then all retire with millions
@andrewamann2821
3 жыл бұрын
Congratulations... You've just invented the .com bubble v. 2.0.
@joey199412
3 жыл бұрын
Great video especially the assembly multiply instruction and outlining it with an analogue computing method. This makes sense since analogue results are instantaneous and don't require a clock pulse and thus no memory storage between calculations as you can add and subtract analogue signals instantaneously.
@ksh6
3 жыл бұрын
Answer to why it is difficult to trust AI fully. And why it is not possible to debug it. Modern AI works like this in simple words. A group of people create a brain like thing. Then another group of people take that brain and teaches it things by showing it examples of how it is rightly done like for example identifying cats in a photo. The brain learns seeing the pattern similar to how a toddler learns what a cat looks like. We don't exactly know how or what the brain learned. We just know "Ok it has learned enough to identify cats with 99.9% accuracy". There is no way to debug it because it is not programmed, instead it is taught. For the same reason as why you don't know why your toddler saw a cat and called it a dog one day, you don't know why the AI started doing things wrongly.
@raykent3211
3 жыл бұрын
Excellent vidéo, thank you. I don't think it's the analogue aspect that can make AI indecipherable. In a purely digital neural network the trail of causation that resulted in certain weightings (and therefore a decision) is irreversible. I'm fascinated and worried by recent discussions of how a trained system could have inbuilt prejudice that can't be proven.
@faustin289
3 жыл бұрын
This is no different by how decisions are formed in human mind either. It has been observed that we (not sure who's we) take decisions and our conscious self then try to rationalize those decisions after the fact.
@fofopads4450
3 жыл бұрын
It is not indecipherable. It is just not easy to decipher, because you will end up needing more computational power than the AI itself consumes, just to monitor what the AI is doing. It defeats its purpose.
@osem598
3 жыл бұрын
The robots that end humanity will be full of vacumn tubes XD
@fulldnbboy
3 жыл бұрын
Underrated comment lol
@fr3zer677
3 жыл бұрын
Another amazing video! It's astonishing to me how many different topics are covered on this channel and how in-depth and interesting all of your videos are.
@robertdewar1752
3 жыл бұрын
Don't forget, computer algorithms are just that - computer algorithms, and no amount of marketing hype can change that. Actual AI is still a long way off.
@resurrectionx5952
3 жыл бұрын
Discover the mystery of brain 🧠 and implement the findings to AI both on hardware & software level. The brain is the road map layed out for us. We are the builders 👽
@generalx5220
3 жыл бұрын
Wow! I’m now woke AF in the understanding of AI
@A.Mere.Creator
3 жыл бұрын
The part about the human brain blew my.... mind.
@koiyujo1543
2 жыл бұрын
Analog technology is going to come back at some point because our brains are both digital and Analog and many don't realize that and the thing is, this means that we been making computers wrong like how the YT channel general films say that the technology in the alien says they use outdated technology when in reality Analog technology are much more robust for long distance travel through the dangerous vacuum filled with radiation like gamma, high energy particles, and x-rays which could screw up computers and alien technologies like magnetic tape and Analog computers would be useful in the term of space travel and irl Analog is actually much more less prone to hacking anf harder to breech devices not to mention they use a lot less energy in calculations which is why every watt in space is important no matter how you think of it. Hybrid digital and analog computer technologies will come back again and analog is getting better at not being as error prone and one day will be able to one day become no longer able to be error prone and soon will always be able to come out with the same answer like digital instead of it being different every time like early analog technologies and they use a lot less energy and are much more faster and are most easier for modularity and more.
@jansternagel4214
3 жыл бұрын
Any sources for the claim of being up to 500 times more energy efficient? (10:40) I would love to use it in my thesis
@LightningHelix101
3 жыл бұрын
There is a group out of University of Michigan (Dr. Flick) has fabricated a few of these chips.
@torbenjensen5002
3 жыл бұрын
The human brain is a quantum-computer with one billion qubits working at 37 degree Celcius.
@youdrakkar
2 жыл бұрын
I bet Derek has been defenetly inspired by your wonderfull video for his recent take on the analog computation.
@user-cx2bk6pm2f
3 жыл бұрын
As soon as he mentioned "analog" was waiting for the requisite mention of noise being the limiting factor. I'm impressed that he did indeed talk about that.. but disappointed that precluded the epic rant I was about to unleash 🤣
@onehouraday
3 жыл бұрын
Quite interesting! Neurons in the brain actually act as a mixed analogue/digital system. The input is digital (action potential), they do analogue processing, and output digital again (action potential, it's either 0 or 1).
@yepyep266
2 жыл бұрын
computing has to become analog to get to the next level. In a sense, going digital is effectively compressing a whole signal to a single bit of information.
@PunmasterSTP
Жыл бұрын
It kind of blew my mind when I found out there were dedicated AI regions in microchips, but I guess that was only a logical next step. I'm not in the field and I doubt I'll ever use this knowledge, but I definitely find it fun and interesting to learn about. Thanks for the very high-quality video!
@Jluna16
3 жыл бұрын
This man went from one shit I don’t understand to another shit I don’t understand all video long 😂
@wood6454
2 жыл бұрын
86 billion processing units in my brain and I still can't add two digit numbers
@sanjaylife
3 жыл бұрын
I this future of computers is cyborgs. humans + computers.
@zubinzuro
3 жыл бұрын
The future of humans is cyborgs, the future of computers is AI superintelligence.
@EddyVs0
3 жыл бұрын
So proud of myself, understood 6 words. Harvard here i come.
@rizolli-bx9iv
3 жыл бұрын
Quantum computing will allow us to have better computing with ai
@steliostoulis1875
3 жыл бұрын
Why can't people pronounce "von neuman" right... Seriously people it's not that hard
@workingguy
3 жыл бұрын
You didn't spell it right, either, so...
@calholli
3 жыл бұрын
Your videos are like an endless competitive battle between my comprehension and your articulation; may this war never find peace.
@aceofspades001
3 жыл бұрын
You know how dumb I am? I was fully immersed in the video to the point I thought the ad was part of the topic! Im dumb 🤣
@tw3638
Жыл бұрын
80% of this vid went straight over my head ngl
@chefboyrdee1
3 жыл бұрын
no offense but AI has been around for a while... just not main stream. Algorithm design and mathematical modeling has been around for a long while.
@logangraham2956
3 жыл бұрын
there are a few , and i mean very few , main stream implementations . the only one that really comes to mind is personalityforge.
@fugslayernominee1397
3 жыл бұрын
I had goosebumps just before the ending. Looks like brains truly are the most efficient machines that nature has provided us with.
@Lemonator32
11 ай бұрын
The ending kinda freaked me out a bit.
@user-7165jdhrnxymzn
3 жыл бұрын
I don't understand almost everything that he said. But I really like the music. Please type 1 if you're just like me
@user-7165jdhrnxymzn
3 жыл бұрын
1
@entropysalamander
3 жыл бұрын
I love the visuals used in your videos, they're always unobtrusive but fascinating.
@wacekjalczak2134
3 жыл бұрын
delicious, ... as always.
@miguelvillasenor3619
3 жыл бұрын
Your Ad was about half of your video... Sell out
@shairozsohail1059
3 жыл бұрын
Been an AI researcher for years and learned a lot from this video. Thanks
@nickandersonco
3 жыл бұрын
What an amazing channel. These videos are a gift to humanity. Thank you.
@suntzu1409
3 жыл бұрын
Analog boi: my time has come
@factsheet4930
3 жыл бұрын
What about Quantum Computing? Doesn't it encapsulate both elements of analog computing (at least in the background) whilst still being digital?
Пікірлер: 1,1 М.