Daddy! Daddy! – Yes DADDY! What do you want? Will you buy me a Quantum Computer? Not before 2032, little one! Hi, everybody. I am Marco, Cuso for my friends and welcome to… the last episode of Quantum Computing series or, “How Quantum Computers will change the way we calculate 1 + 1” This is also the 100th video on the channel If we include a promotional one that now is not listed Therefore I am very happy that this event happens on this video that I find very important, and I hope you’ll find it very interesting and will share it a lot! This video is born also thanks to an external co-operation and I mean there will be nobody actually here with me. Manuel Occorsi, from the page “Close-up Engineering” published some articles related to how the IBM computer works that you can find and read in the link I leave below in the info box. Manuel is studying Computer Engineering and writes for Close-up Engineering a web page that, together with Scientificast, Astronauti News and, of course, La Fisica che non ti aspetti manages the Telegram Channel “Telegram Scienza”, that you can find in the info box. Today, we’ll see together how the IBM Computer works: how it is built, and what are the issues that, as per today, make it impossible to build a more powerful computer IBM Quantum Computer has 5 qubits. Several different techniques have been developed to build qubits by exploiting different quanto-mechanic effects. In this case, IBM computer uses qubits based on Josephson Effect. This kind of qubit has the big advantage of being built directly on single silicon chip using well established technologies. These structures are deposited on silicon (and maybe we’ll talk about JJ in another video) but they require very low temperatures, because the deposited metals must be put in a superconducting state. IBM computer is constantly kept at around 15 milli Kelvin. using a so-called dilution cryostat. It is worth mentioning that 15 mK is a temperature even lower than outer space. That means that building the Quantum Computer out there wouldn’t be enough. The system must be kept at an even lower temperature. These qubits can be put in states equivalent to those obtainable with the logic ports which we talked about in the second video using microwave pulses that are directed to the qubit by using wave guides that are deposited directly on the chip. So, the whole system is obtained directly on the chip. Once they are put in the desired state (0, 1, a specific coherent superposition) The qubits remain in that state until we tamper with them again by modifying the state or measuring them. Or better… They should remain in that state. Unfortunately, the presence of residual photons can lead to interactions that are equivalent to measurements of the qubits and this can change their status in an unpredictable way. For instance: If I prepare a quibt in an “fair” superposition of states that is, where the probability of getting 0 or 1 is the same when we measure it, so it’s 1/2 I would expect, when I perform the measurement, to get 0 half of the times and 1 half of the times.. IBM Quantum Computer has a command called “Idle” Idle does nothing. It waits. It waits for the same amount of time needed by any other command to be executed. When I measure the qubit, I would expect to get half of the times 0, half of the times 1. BUT! If I repeat the experiment FOR REAL more than once (and IBM QC allows you to repeat it 8192 times in a row) I will get the 0 state more often. Why? Because it is possible, although unlikely, that an interaction with a photon that shouldn’t be there makes the qubit to lose its superposition of states and collapse into either the 0 or 1 state. This process is known as de-coherence In principle, it should’t be a big deal, ’cause if the statistics is the same even if the measurement is performed before my command by the machine I still would expect the qubit to collapse in either the 0 or 1 state with an equal probability and then remain in that state. But things don’t go this way. State 1 is an excited state, so, even if this process is very unlikely at those temperatures, a qubit can spontaneously collapse from state 1 to ground state 0. And this changes all my results. We can clearly see that if we prepare a qubit in state 1 and wait for 20 idle cycles before measuring it. In theory, I should always get the qubit in state 1. In practice, there”s a small chance for it to collapse to 0. This is obviously a first issue, as my results may be affected by de-coherence. So, it’s necessary, although not simple, to develop some means to either know that de-coherence occured, so my data must be post-treated, or to avoid it in the first place If you’re interested, I will link in the info box two rather technical articles about the subject. Another serious issue is information transport from one QUBIT to another one. Contolled- (C-) NOT gate, a fundamental gate for any quantum computer applies NOT operation to a QUBIT if another QUBIT’s value is 1 So it is very important to be able to make such a comparison. But, how? In a classic computer I can, oversimplifying, draw a wire between my two bits, my two transistors. I can’t be so trivial here. How do I solve the problem? IBM doesn’t tell everything about this and how they actually built their QC, but they managed to solve the issue, at least partially. because if you look at the structure, only one QUBIT (Q2) is actually connected to the other four, which are not connected among them. So, gates like C-NOT or Toffoli, which involve more than one QUBIT have to involve by design the central QUBIT (Q2). Another important problem, which is not related to IBM QC, but is inherent to Quantum Algorithms, is the randomness of results. Let me explain. Grover’s Algorithm, which we talked about several times in the previous videos, doesn’t always give out the correct result. There is only a finite probability that GA calculates the correct answer for instance it finds the correct book, page, line and position where the word I am looking for is among all the books of the Library of Congress. But when the number of elements to search through is large, this probability increases but it tops at 50% So, when N becomes huge, and GA can work at its best its best is that it will give the right answer every other time. What the ???? How can this be acceptable considering the potential of this machine? We have to remember a couple fo things: first of all, a QC takes much less time than a CC to calculate the result, orders of magnitude of time less. so, even if I had to repeat the algorithm 10, 100, 1000 times, it would still take much less time than a CC completing the calculation only once. Ok, the CC will give the right result, maybe, but the odds that GA picks the wrong results 10 times in a row is less than 1 in 1000. The second thing is that checking hte result is extremely simple and fast. you just have to take the selected book, look up the selected page, and line, and position and see if the word is correct. if so, I’m done. Otherwise I will start again, but still taking less time than a CC. In the end, will we ever have a QC in our pockets? Will the hardware dimensions ever shrink down to those of a smartphone? IBM staff doesn’t think so, as the hardware is extremely complicated. and even if CC followed a similar path, starting from huge machines as large as a room and ended up in tiny objects we can keep in our pockets, and that still have more computational power than those ancient computers, with QC things seem to be much more complicated. But, there’s another thing we can do. It would be possible to use QC’s just like we can do today with IBM’s QC: on the Cloud! The concept is that we’ll have a regular smartphone where we’ll type our problem which will be sent to a QC at some research centre THAT will calculate the results and send them back to my phone. So my phone will just be a terminal of the QC. This is a much more reasonable scenario, and also one that is closer in time. But how much? An estimation is very hard to make. Some say that in 20 years we’ll have QC’s available as products. QC’s that can actually be used for real purposes. but that would require an increase in the number of QUBITs IBM’s staff makes another forecast: when QC’s will exceed 50 QUBITs their computational power will surpass any supercomputer we’ve ever built and will ever build. A CLASSIC supercomputer, of course. Well, my friends, this video and this series end here. I really liked making this series and I hope you liked it as well. I had a great feedback with respect to sharings and comments and talking about comments, I’d like to thank everyone who, thanks to their expertise, know about computers and who made comments much more interesting answering precisely other users’ questions. Thank you very much, and thanks to all who just watched the video, liked it or shared it. I think that’s the way the community on Youtube should work. See you in the next video, thanks for watching and remember: Be curious: Physics is were you least expect it! Daddy! Daddy! DADDY!!!!