How Quantum Computing Opens Up A New Frontier For Software | Michael Brett | TEDxBermuda

something truly remarkable has happened on the Internet's in the past 10 years and you might not be aware of it yet and that's that the cost of making a prediction has radically decreased so the unit cost of trying to make an assessment of what might happen next has become so cheap that it's happening all the time all around us on the internet and we use it all the time so when Siri when you ask Siri a question and she responds what she's doing is making a prediction of what it was that you intended to say and what kind of answer you need that when you ask what the best route to get home is on through the traffic that's a prediction about what the the optimal routing and shortest time would be to get home and so we're seeing all these new applications come to fruition based on predictive analytics and there's a whole new technology suite out there that's helping us do that but there's also a challenge even with all the computational power in the world we're running up against limits of what's possible in the number of predictions that we can make and how fast that we can go so my question for you today is what would you do if you had access to all the computing power in the world to make a prediction what's the most important prediction to you and your life maybe it's what might happen to your sports team this year or something to do with your health or even what's the best route to get home from the conference today and what to do next and so over the next couple of minutes I'll take you on a little bit of a journey about what predictive analytics means and how we might do this together so prediction is all about patterns recognizing patterns and then describing them and understand what might happen next and this is the simplest pattern that I could come up with that we could all do together so it's a straight line we can describe it using an algorithm that we learned in high school y equals MX plus C and we could probably all figure it out relatively quickly and if we expanded the canvas make a very accurate prediction about what might happen next what happens if the pattern gets a little bit more complex so this pattern is now starting to work in multiple dimensions there's obviously some underlying rules and rhythm that governs what happens here but over time as the patterns become less intuitive to our human brains we need a little bit of help from computers to help us figure out what's going on behind the scenes then probabilities get involved this is a pattern it's becoming more complex again and as the tree expands we can see that there are some underlying rules that are governed by mathematics but there's some probabilities that come into play every time the branch splits into two and as humans it's very difficult for us to figure out what might happen next with this but if we use a computer and we design an algorithm use some machine learning to help train that we can get a fairly accurate prediction about which direction it might go at which time maybe it won't be a perfect prediction but over time we'll get relatively accurate expectation about what happens next but then chaos sets in these kind of patterns are almost impossible for us as humans and almost impossible for machines to figure out as well we could spend a lot of computational time trying to understand what the underlying rules were behind this particular chaotic pattern we might be able to get there with a lot of computational effort and some really smart work going into it but it's really challenging and natural systems are more chaotic again there's some obviously some patterns underlying this behavior but what if we add an intelligent system to that as well so how might this shoal of fish behave over time how do they react to each other which direction might they go it's extremely challenging to work this out and then we look at one of the most complex things that human beings do almost every day when we drive a car we're making predictions about what might happen next all of the time almost intuitively whether that dog is going to run out from behind the bus whether the taxis going to turn in front of us what color the light is going to change we're doing this woman very naturally and intuitively as we we learn how to drive but when we build self-driving cars and put these into the same environment with us as human beings the amount of predictive capability that those machines need to interact with us in that environment very reliably is extremely challenging and so the amount of computational power that's needed to make those predictions in to train those machines is immense and we're starting to run up against the limits of the computational power that supports that one of the underlying rules that's helped us create computational power over the last 52 years is Moore's Law the doubling of computational power and over we've been riding that wave as we've created new algorithms and new computational power since then but we're starting to reach one of the fundamental limits of how Moore's how we keep up with Moore's law and that's the size of the transistor that we can fit onto a single chip this is one of Intel's most recent chips the gap between those little branches on there is about 14 nanometers now if you squeeze your fingers together as tight as you possibly can the gap between your fingerprints is tens of thousands of times bigger than what's on this screen this is getting down to the atomic limit of how small we can make something we're almost literally counting the number of atoms to make sure that there's enough electrons to create electricity to flow across that gap and to create an electronic circuit and so there's a real challenge right now about how do we keep up with Moore's law how do we keep creating more and more powerful computers when we're reaching the limit of how small we can make something and therefore how many transistors we can cram onto a single chip and so there's a number of different techniques to today and we're seeing like multi-core computers and new technologies like field programmable gate arrays and all sorts of wonderful things but one of the technologies that is extremely promising and potentially giving us a major breakthrough and computational power is quantum computing so what do you see on the screen here is something truly remarkable this is developed by a team at the University of New South Wales in Sydney Australia the blue stuff is silicon and the red stuff is phosphorus and that is a single phosphorus atom that's been precisely engineered right where the team wanted to be and then some control lines next to it that can write to and read from the information that's embedded within that atom so this is the building blocks of a quantum computer a quantum computer says well what if we stop what if we ignore the limit of building a transistor at the atomic level and we go inside the atom what if we go down to the next layer of physics into quantum physics and start to make use of the strange behavior the weird and wonderful world of quantum physics and build computers that rely on that physical process as opposed to classical computers in the work around there so this is the building blocks of some of the first quantum computers that we're going to be able to use to keep up with our predictive analytics capability and it seems around the world that are racing to build these right now this is a chip that's built by Google it was released earlier this year this is their quantum computer it's a slightly different approach to the the one out of Australia but this is a real quantum computer it looks like a regular chip it's about the size of your thumbnail but inside there is some truly remarkable engineering to control information at the at the quantum level at IBM they've also built a quantum computer you can use it today you can go to ibm.com slash quantum and sign up and start using this quantum computer and start programming on it these are accessible real machines the shielding that you see around the the computer there that's part of a refrigeration system that cools these chips down to almost absolute zero so that we can start to play with the quantum mechanics that's going on inside there's startup companies as well the team at a at a firm called Righetti have recently built a chip as well and starting to make these more and more accessible to software engineers like me that are really interested in the algorithms that sit on top of them and so we've seen an explosion in the amount of application work that's going into quantum computing over the last couple of years this has been a field that's been highly theoretical for the last 20 years or so but as these chips have become accessible software engineers and computer scientists have started to look at the applications that can sit on top of that and things like machine learning and simulation have become some of the most promising application areas of quantum computing we're really excited so let's take a look at how this is different to a classical computer as you know the classical computer with a gate it's either a 1 or 0 on or off there's no other possibility for that that transistor to be in a state but with a quantum computer we can create a qubit a quantum bit and that's got some probability of being a 1 or a 0 and when we read that qubit it flips into place so we can store information within the probabilities embedded in the quantum physics and this is extremely powerful because it allows us to encode kind of more information and more possibilities within the same number of keep qubits as we get in the classical computer so let's imagine a classical computer that's got two transistors that means we've got four possibilities to work with it's either 0 0 0 1 1 0 or 1 1 and if we wanted to find the right answer to what might happen next with the prediction we would have to step through each possible one one at a time each possible state and and see if that was the the correct answer for us but with a quantum computer we can reorganize this into two qubits and store four possible states within there with some probability that we can measure simultaneously so it's got all of these states existing at the same time with the probability that we'll pull something out of it so rather than stepping through them one at a time we can keep all of this information stored there and measure once and get the right answer from that that we can then use to make our prediction and this is where I'm going to start to blow your mind we can start to play with the probabilities others and entangle multiple qubits together and while it's computing play with the probabilities so that we can get the right answer for us out of that so let's say for example that we always wanted the two qubits to work in opposite directions so then we take the probabilities from being 25% for each possible state and change that to being 50% for two states there so while our computational work is going on we're playing with the probabilities inside the system and helping us find the right answer out of them what's truly remarkable about this is that with the we only need a small number of qubits to store a lot more potential information than in a classical computer and so with just 30 cubits we could potentially run faster than the fastest supercomputer that exists in the world today with 300 cubits that is more computational power than all of the computers that exist on the planet today in one computer and we're at the really early stages of building these machines we've got a long way to go but it's incredibly exciting to do this and so what can we do with the quantum computer it will transform the way that we do predictive analytics and I don't think you'll access one in your pocket there won't be one in your in your cell phone but within a cloud platform there'll be multiple quantum computers multiple different types of computers running alongside each other so you'll be accessing CPUs graphical processing units all sorts of different machines and a quantum computer is a special case that runs alongside those and allows us to work with probabilities to solve some hard problems the ones that we're particularly interested in in the very early days of quantum computing are machine learning helping us do predictive work large-scale optimization problems like traffic routing and quantum chemistry understanding how molecules work together how cells behave down at that level and it's all about prediction and so we can't do this alone there's a whole new workforce that we need a whole new skill set that's required to build these applications we need to combine things like software engineering and quantum algorithm research data analytics into into quantum software engineering and then create a skilled workforce around quantum data analytics to make use of these remarkable new machines and that's where I invite you all to come on this journey with us we're at the beginning of a new revolution in the way that computers are built it's opening up and the remarkable new set of applications that we that were previously intractable to us that we couldn't solve even with all the computing power in the world today and so what we're doing at my company is hoping to train people to do that work we're running hackathons around the world we're running training sessions to take software engineers and computer scientists and data scientists that haven't worked with quantum computers before and to teach them the fundamentals of our these new computers work the kind of algorithms that we can run on top of them and open their mind to the potential of the applications that we can run with these new computers what we're seeing is that despite all the theory that we've had over the past 20 years or so about what's possible with the quantum computer that there's as we put tools into the hands of people and they start to explore this for the first time their imagination starts to run wild about what they could potentially do with an ability to play with probabilities and the ability to play with the potential of quantum computing and so we're incredibly excited about where these computers might take us or at the very beginning of this journey but we've got the potential to open up a whole new level of computational power and do some remarkable things with it [Applause]

Channel: TEDx Talks

Views: 3,483

Rating: 4.884058 out of 5

Keywords: TEDxTalks, English, Technology, AI, Computers

Id: gpF-4JY88BI

Length: 14min 12sec (852 seconds)

Published: Tue Dec 04 2018