>>ALL RIGHT. LET’S GET STARTED. WELCOME EVERYONE. THIS IS THE FIRST OF SEVERAL SESSIONS ON QUANTUM COMPUTING. THIS IS VERY EXCITING TIME TO BE WORKING IN THAT FIELD. WE ARE KIND OF AT A JUNCTURE WHERE WE ARE EXPECTING TECHNOLOGICAL BREAKTHROUGHS VERY SOON AND ALLOW US TO SCALE UP THE SYSTEMS. THE FOCUS TODAY WILL BE ON THINGS THAT YOU CAN DO AS A DEVELOPER ALREADY NOW, IF YOU WANT TO GET ENGAGED WITH QUANTUM COMPUTING. YOU WILL HEAR ABOUT THE SOFTWARE THAT YOU CAN ALREADY PLAY WITH TODAY. ALSO EXCITING AS WE WILL HAVE TWO FOLKS HERE BESIDES ME, MY NAME IS MARTIN ROETTELER. I’M THE LEAD OF DEVELOPMENT AND TWO FOLKS REPRESENTING THE USER. AND FURSMAN IS HERE FROM THE COMPANY THAT IS USING OUR Q SHARP LIBRARIES AND HEAR ABOUT THESE EXPERIENCES. ALSO HAVE, WHAT IS VERY EXCITING, YOU CAN DO, YOU CAN GET INSPIRED BY QUANTUM MECHANICAL THINKING AND IMPROVE CLASSICAL ALGORITHMS. CLASSICAL AUTHORIZATION PROBLEMS. IN A SENSE, ZERO KUBERNETES AND HELMUT HERE LATER, TO TELL YOU ABOUT THOSE POSSIBILITIES. IT IS EXCITING BECAUSE YOU CAN REACH CUSTOMERS AND MAKE A TREMENDOUS IMPACT ON THE BUSINESS TODAY. I WANT TO GIVE YOU THAT QUICK SHOUT-OUT, THAT BUILD THAT WILL BE SIX SESSIONS IN TOTAL THAT WILL INVOLVE QUANTUM. YOU CAN HEAR ABOUT EXCITING NEW STUFF THAT WE OFFER Q SHARP NOTEBOOK, ZERO INSTALL EXPERIENCE. YOU CAN GET STARTED WITH Q SHARP WITHOUT INSTALLING ANY, THERE IS BASIC BACKGROUND OF QUANTUM COMPUTING AND SEVERAL SESSIONS ABOUT THAT. SESSIONS ABOUT THE TOOLS, INTEGRATE VISUAL STUDIO AND STUDIO CODE AND MORE STUFF ABOUT PYTHON. A LOT OF STUFF TO EXPLORE THIS WEEK . I WANT TO START WITH BIRD’S EYE VIEW. WHY EVEN CARE? NOT A UNIVERSAL TOOL THAT WILL SPEED UP EVERYTHING. IT IS CONTRARY TO DOMAIN SPECIFIC APPLICATIONS THAT ARE KIND OF RARE BUT WE CARE ABOUT THEM. BECAUSE THEY ARE COMPILATIONALLY VERY HARD. YOU FIND THEM IN COMPUTATIONAL CHEMISTRY. THE PROBLEMS THAT YOU WOULD LIKE TO SIMPLY UNDERSTAND NATURE IN A SENSE. UNDERSTAND PROBLEMS ABOUT THE ENERGY LEVELS OF A MOLECULE. AND PROBLEMS ABOUT CATALYSTS, HOW TO DESIGN CATALYSTS THAT HELPS THE CHEMICAL REACTIONS. AND CLASSICAL COMPUTERS REACH LIMITS THERE, MAYBE THE CLASSICAL METHOD IS FAST BUT GIVES.>>VERY INPRECISE AND INACCURATE ANSWER. OR PRECISE BUT DON’T SCALE WELL WITH THE SIZE OF THE SYSTEM. THERE ARE PROBLEMS LIKE THAT ARISING FOR INSTANCE CHEMISTRY. FERTILIZER DOESN’T SOUND LIKE THE MOST APPEALING OF THING, BUT THE DESIGN OF FERTILIZER, THE TYPE OF PROBLEMS ARISE, FERTILIZER IS MADE FROM SUBSTANCES WHICH ARE INITIALLY KIND OF MADE FROM HARVESTING NITROGEN FROM THE AIR, TURNING IT INTO COMPOUND FORM SUCH AS AMMONIA AND FURTHER PROCESS AND THAT FIRST STEP IN THE PROCESS IS DONE USING A PROCESS THAT IS KIND OF KNOWN FOR HUNDREDS OF YEARS. ABOUT A HUNDRED YEARS. BUT IT IS VERY INEFFICIENT. WE WOULD LIKE TO FIND A BETTER PROCESS. WE KNOW IT IS POSSIBLE TO DO IT IN KIND OF ANOTHER WAY, NATURE DOES THAT IN CERTAIN PLANTS. BUT WE HAVE TO DO IT AT HIGH PRESSURE, HIGH TEMPERATURES, AND JUST WASTES A LOT OF ENERGY. WHEN PEOPLE ESTIMATE ABOUT 2 OF GLOBAL ENERGY OUTPUT GOES INTO THAT ONE PROCESS OF MAKING AMMONIA. IF YOU FIEND A BETTER WAY, TREMENDOUS IMPACT ON HUMAN KIND. SO SIMILAR APPLICATIONS, WE ANTICIPATE IN, FOR INSTANCE, CAPTURING CARBON FROM THE AIR. THERE ARE PROCESSES THAT DO IT ALREADY NOW BUT VERY INEFFICIENT. IF THERE WAS A BETTER WAY, THAT WOULD HAVE TREMENDOUS IMPACT. QUANTUM COMPUTERS HAVE THE GREAT PROMISE OF BEING A TOOL TO HELP US WITH THAT. MATERIAL SCIENCE SIMILARLY A LOT OF APPLICATIONS AND THE, ONE OF THE PROMISING AREAS FOR QUANTUM COMPUTERS ALSO MACHINE LEARNING APPLICATIONS. WE DON’T HAVE, WE HAVE NOT REALLY FOUND A KILLER APP YET. WE HAVE SOME HUNCHES OF THINGS THAT QUANTUM COMPUTER COULD HELP US THERE. FOR INSTANCE, WE KNOW THAT QUANTUM COMPUTER IS GOOD AT COMPUTING CORRELATIONS, AND COMPUTING FREE TRANSFORMS AN AMPLIFY CERTAIN PROGRAMMABILITYS. HISTORICALLY, THE FIRST APPLICATION REALLY WERE IN AREA CLOSER TO TYPOGRAPHY. VERY MATHEMATICAL PROBLEM SUCH AS FACTORING OF NUMBERS. THIS IS PRETTY BIG NUMBER. IT HAS 2, 000 BIT, ROUGHLY 600 DES MEL PLACES. ONE OF THE RSA NUMBERS. THE PROBLEM OF FACTORING, OR THE HARDNESS OF FACTORING IS REALLY UNDERLYING A LOT OF CRYPTO VIEWS THESE DAYS. TURNS OUT ON A QUANTUM COMPUTER, YOU CAN FACTOR NUMBERS RELATIVELY EASY. YOU STILL HAVE TO BUILD A MACHINE THAT IS QUITE SIZABLE. FACTOR FOR INSTANCE, 2, 000 BIT NUMBER, YOU NEED 4, 000, ROUGHLY 4, 000 QUANTUM BIT, Q BITS, AND THOSE Q BITS HAVE TO BE REALLY GOOD. CORRECTED. IF YOU LOOK AT THE HEADLINE NUMBERS, YOU MIGHT FIND IN THE PRESS, THOSE ARE NOT CUBICS. WE NEED AROUND 4, 000, HAVING 4, 000 WOULD ALLOW US TO FACTOR IN NUMBER, INSTEAD OF A BILLION YEARS WHICH IS THE BEST CLASSICAL ALGORITHM THAT DO IT IN THAT TIME, WE COULD DO IT UNDER SOME ASSUMPTION OF THE CLOCK SPEED. IN A SECOND OR SO. AND THE SAME THING IS TRUE FOR OTHER METHOD, MIGHT HAVE HEARD ABOUT TYPOGRAPHY, OUR TEAM IN REDMOND ANALYZED THOSE AND FOUND SIMILAR COST ESTIMATES IN TERMS OF CUBICS YOU WOULD NEED. AND THE NUMBER OF OPERATIONS YOU NEED. AND ALSO BREAK THAT. THAT IS INDICATION THAT USED IN A LOT OF BLOCKCHAIN TECHNOLOGIES. WE HAVE TO CHANGE THAT TECHNOLOGY TO ANOTHER TECHNOLOGY. THE GOOD NEWS IS THAT THERE ARE PROPOSALS HOW TO REPLACE CRYPTOGRAPHY WITH QUANTUM SAFE FORM. A BIG SHOUT-OUT TO THE GROUP THAT BRIAN RUNS AT MSR. RIGHT NOW, THERE IS AN EFFORT GOING ON BY THE U. S. GOVERNMENT, SO THEY HAVE A CALL FOR NEW CRYPTO SYSTEMS AND MICROSOFT HAS SEVERAL CONTAINERS IN THAT RACE. IT IS INTERESTING FIELD LIKE HOW CRYPTO SHIFTS TO ITS QUANTUM. FEAR NOT, AS THERE WILL BE NEW KINDS OF CRYPTO, EVEN WHEN QUANTUM COMPUTER IS AROUND . CRYPTO IS NOT A GOOD BUSINESS CASE. HOW MANY CRYPTO APPLICATIONS WILL YOU BE ABLE TO SELL? MAYBE A FEW, MAYBE SOME GOVERNMENT AGENCIES CARE ABOUT IT. IT IS NOT REAL BUSINESS CASE TO MAKE A LOT OF MONEY FROM IT. SO TO UNDERSTAND THAT A LITTLE BIT BETTER, LET’S TRACK BACK AND THINK ABOUT LIKE WHAT CAN WE LEARN FROM NATURE REALLY? SO KIDS ARE GREAT WITH THAT. KIDS ASK US QUESTIONS LIKE, WHY IS GRASS GREEN? LEAVES GREEN? WHY IS THE SKY BLUE? WE DON’T THINK ABOUT IT MUCH, WHY? BUT REALLY THERE ARE REASONS FOR THAT, THAT ARE FUNDAMENTALLY QUANTUM MECHANICAL. UNDERSTANDING THE SKY, YOU HAVE TO UNDERSTAND SCATTERING OF LIGHT. AND SO ON. SO REALLY WHAT WE TRY TO DO IN QUANTUM COMPUTING WE LEARN FROM NATURE AND HARVEST THAT AND TURN IT INTO APPLICATIONS. AND THE PRIME EXAMPLE IS, QUANTUM CHEMISTRY. SO IN A NUTSHELL, WHAT GOES ON IS, YOU HAVE A SYSTEM, LIKE A MOLECULE, NOW YOU TRY TO MODEL THAT AND CHEMISTS HAVE BEEN DOING THAT FOREVER. IT IS A TYPICAL WAY TO MODEL IT. IT IS TO IGNORE MOST OF THE STRUCTURE OF A MOLECULE IN JUST FOCUSING ON THE ELECTRONIC STRUCTURE. THEN YOU ARRIVE AT THE CERTAIN MATRIX, CALL IT HEMO IT ONA. UNDERSTAND THE CHEMISTRY STRUCTURE. UNDERSTAND. TYPICALLY ONLY THE LOWEST ONE. INFORMS US A LOT ABOUT THE PROPERTIES OF THAT MOLECULE. AND QUANTUM COMPUTER IS REALLY GOOD AT THAT. WHY? BECAUSE IT IS A CONTROLLED QUANTUM MECHANICAL SYSTEM. IF YOU HAVE ANOTHER QUANTUM MECHANICAL SYSTEM, YOU CAN MAP IT TO YOUR CONTROL SYSTEM AND STUDY THE TIME EVOLUTION OF THAT SYSTEM WHICH IN TURN INFORMS YOU ABOUT THE SYSTEM THAT YOU CARE ABOUT. BUT IT IS NOT AS SIMPLE AS THAT. THAT HAS BEEN KNOWN FOR QUITE SOMETIME THAT SOME TIME REVOLUTION IS SOMETHING QUANTUM REVOLUTION IS GOOD AT. THERE IS A CATCH, EVEN IF YOU HAVE AN EFFICIENT WAY IN A SLOWDOWN, THAT CAN QUITE LIKELY KILL YOU IF THE DEGREE OF IT IS TOO HIGH. THE FIRST ALGORITHM IN THAT SENSE WERE ACTUALLY THE SCALING OF ENTER 11. ENTER THE PROBLEM SIZE, THE NUMBER OF SPATIAL FUNCTIONS MUCH SUCH AN ALGORITHM, YOU DID A MOLECULE THAT PEOPLE ARE REALLY CARE ABOUT, SUCH AS FOR NITROGENIZE, RELATED TO THE CATALYST PROBLEM OF HARVESTING NITROGEN. THEN STILL END UP WITH A RUN TIME OF BILLION OF YEARS. IT IS KIND OF SLOW. FOR PEOPLE IN THE TEAM, MICROSOFT, HERE IN REDMOND AND SEVERAL IN THE AUDIENCE, THERE WERE OTHER PEOPLE, DAVE, AND SEVERAL OTHERS, WORKED ON CHIPPING AWAY SLOWLY THE COMPLEXITY OF THAT ALGORITHM. AND FINALLY BROUGHT IT DOWN TO SOMETHING THAT COULD ACTUALLY BE DONE IN A, SAY, REALISTIC QUANTUM COMPUTER HAVING A REALISTIC CLOCK SPEED. THE POINT HERE IN THAT SLIDE IS, IN QUANTUM WORLD, WE HAVE SIMILAR THINGS GOING ON LIKE IN A CLASSICAL WORLD. WE NEED TO HAVE QUANTUM ALGORITHMIC THINKING. WE NEED TO BE AWARE OF THE RUN TIME. JUST HAVING POLYNOMA IS NOT ENOUGH. WE NEED TO DEVELOP TECHNIQUES THAT REDUCE THE COMPLEXITY. ULTIMATELY, WE WANT TO PROGRAM THESE ALGORITHMS. WE WANT TO KIND OF BE ABLE TO, EVEN NOW, WANT TO BE ABLE TO WRITE THE PROGRAMS THAT WE CAN LATER RUN WHEN THE QUANTUM COMPUTER SCALES UP TO THAT SIZE. WE WANT TO HAVE THE PROGRAMS READY, UNDERSTAND THEM BEFORE, SO THAT WE CAN FIND KIND OF BOTTLENECKS WE WANT TO PROFILE THEM. AND IDENTIFY GOOD ALGORITHMS. SO THAT IS PART OF THAT SLIDE. REALLY THAT MOLECULE IS ONE PEOPLE CARE ABOUT A LOT IN, LIKE NITROGEN FIXATION. IT IS PART OF A REALLY, REALLY BIG MOLECULE. CALLED THE ACTIVE SIDE IS JUST VISUALIZED HERE, MUCH, MUCH SMALLER THAN THE ENTIRE COMPLEX. AND THIS, THIS TEAM HAS WORK THAT WAS PUBLISHED A FEW YEARS AGO. WE FOUND IN A FEW HOUR, LARGE SCALE, WE COULD SOLVE THAT PROBLEM. SO AT MICROSOFT, WE TRY TO BUILD AN ENTIRE END TO END STACK FOR QUANTUM COMPUTING. SO IT STARTS WITH THE KIND OF APPLICATION LAYER, SOFTWARE LAYER, WHERE YOU EXPRESS THE HIGH LEVEL ALGORITHMS. THEN IT GOES ALL THE WAY TO QUANTUM DEVICE AND IT NEEDS TO HAVE ANOTHER LAYER THAT IS THE GLUE BETWEEN THESE TWO EXTREMES. IT IS A LOT OF CLASSICAL CONTROL THAT NEEDS TO HAPPEN. WE HAVE THIS END TO END STACK VIEW THAT WE WANT TO REALLY DEVELOP ALL OF THE COMPLEMENTS. IN THIS TALK, I’LL FOCUS ON THE TOP LEVEL COMPONENT AND PRESENT YOU SOFTWARE. SO THE SOFTWARE IS BASED ON THE LANGUAGE CALLED Q SHARP. AND LIKE C SHARP AND F SHARP, IT DRAWS ON SOME MULTIPLE PARADIGMS OF PROGRAMMING. IT HAS FUNCTIONAL FLAVORS TO IT. IT ALLOWS YOU TO EXPRESS MANY IDEAS THAT YOU MIGHT HAVE FOR QUANTUM ALGORITHMS AT A VERY HIGH LEVEL. THAT IS ONE OF THE ASPIRATIONS ALLOWING HIGH LEVEL PROGRAMMING WHERE YOU DON’T HAVE TO WORRY ABOUT BIT LEVEL OPERATION SO MUCH. NOBODY REALLY LIKES TO DEAL WITH ASSEMBLER, UNLESS YOU CARE ABOUT PERFORMANCE, WHICH WE CAN ALSO DO BUT MOST OF THE TIME YOU JUST WANT THE IDEA EXPRESSED. ALSO WE CARE ABOUT PROBLEMS AT SCALE. WE DON’T WANT TO JUST LOOK AT TOY PROBLEMS, OR LIKE, KIND OF JUST NOT REALISTIC PROBLEMS. WE REALLY WANT TO STUDY THE PROBLEMS AT SCALE THAT WOULD HAVE AN IMPACT AND MOVE THE NEEDLE. THAT DRAWS US TO CERTAIN DESIGN IN THE LANGUAGE, KIND OF A REUSABLE DESIGN AND BE TARGET LARGE SCALE. WE CAN HANDLE ANY SCALE. LIBRARIES, THEY ARE OPEN SOURCE. WE CAN, PEOPLE CAN CONTRIBUTE. PEOPLE CAN EXPERIENCE THE ENTIRE LIBRARIES. WE TRIED TO GIVE SOME GUIDANCE WHAT WE ARE GOING TO DO NEXT WITH LIBRARIES AND YOU CAN PLAY WITH FUNDAMENTAL LIBRARYS. IN A MOMENT, YOU CAN SEE ONE OF THE MAIN WORK HORSES, CHEMISTRY LAB. WE HAVE VISUAL STUDIO. THE DOCS. WE INFLUENCE THE DOCS TEAM. BECAUSE A LOT OF THAT STUFF IS MATHEMATICAL. IN MATH, IT IS LANGUAGE CALLED LATEC, THAT SOME OF YOU MAY HAVE HEARD. DOC TEAM SUPPORTS MATH JACKS RENDERING WHICH IS COOL. THIS WHOLE THING RUNS ON DOC NET CORD. INITIAL RELEASE, FIRST RELEASE, THEY WERE FRAMEWORK. THEN WE SWITCH CORE, MULTI PLATFORM OF CORE. RUNS ON WINDOW, MAC AND LINUX. ROUGHLY THE VERY HIGH LEVEL PICTURE OF HOW A Q SHARP PROGRAM LOOKS LIKE IS LIKE THIS. REALLY THERE IS A DRIVER WHICH TYPICALLY IS WRITTEN IN C SHARP. IT COULD BE WRITTEN IN F SHARP AS WELL. AND THEN THE, THAT EXPRESSES THE CLASSICAL COMPUTATIONS HERE THAT YOU MIGHT DO ON THE SIDE. THE QUANT YUM MECHANICAL ONES ARE THE TWO KINDS. WE DISTINGUISH BETWEEN OPERATIONS AND FUNCTIONS. OPERATIONS JUST HAVE SIDE EFFECT ON THE QUANTUM DATA AND OPERATE UPDATED. FUNCTIONS ARE THINGS THAT YOU WANT TO DO ON THE SIDE BUT THEY ARE NOT MEANT TO BE COMPLICATED OPERATIONS. JUST SMALL SIDE OPERATIONS. SIDE COMPUTATIONS. THE IDEA IS THAT THOSE THINGS SHOULD MAP TO SOMETHING THAT YOU COULD DO WITH ONE THAT SITS CLOSE TO THE QUANTUM PROCESSOR. THOSE ARE DONE BY THE CONTROL. PART OF THAT. THAT IS THE ROUGH IDEA. AND WE CAN TARGET THEN ULTIMATELY WE WANT TO TARGET THE HARDWARE. IN THE MEANTIME, TARGET SIMULATORS AND WE HAVE KIND OF SIMULATORS WORK RECENTLY FAST UP TO 30 CUBICS OR SO. THEN YOU REALLY SUFFER FROM PAIN AND EXPEDIENTIAL OVERHEAD OF SIMULATING QUANTUM SYSTEM. AND EXPEDIENTIAL NUMBER OF CUBICS. YOU CAN ALREADY USE A PROFILER. IT IS COMPLETELY SCALABLE. JUST COUNTS HOW MANY CUBIC, AND GATES I HAVE AND OTHER METRICS SUCH AS THE DEPTH, THE CIRCUIT DELAY. THINGS YOU CAN DO TODAY. YOU CAN DO IT AT SCALE. YOU CAN WRITE REALLY LARGE SCALE ALGORITHMS AND BASICALLY PUSH A BUTTON AND YOU GET THAT PROFILING INFORMATION. IT IS REALLY FOR MATH, REALLY COOL. AND YOU DON’T HAVE TO REWRITE YOUR CODE. YOU SAVE YOUR, PROTECT YOUR INVESTMENT MUCH THE CODE YOU WRITE ONCE AN EXPLORE TODAY, YOU ULTIMATELY WILL BE ABLE TO RUN ON THE QUANTUM DEVICE. AND BIG NEWS OF TODAY, SOME OF YOU MIGHT HAVE SEEN IT. SO WE HAD ALREADY THE LIBRARIES AND THE SAMPLES WERE OPEN SOURCE, WE NOW ANNOUNCED TODAY THAT WE ARE GOING TO OPEN SOURCE THE Q SHARP COME PILER AS WELL. THE TEAM WILL OPEN SOURCE THAT AS WELL. THE SIMULATORS WILL BE OPEN SOURCED. LIBRARIES SAMPLES AND SO-CALLED CARTOUS IS OPEN SOURCE. IT IS A COOL CONCEPT. SELF-PACED TUTORIALS GOING FROM SIMPLE THINGS TO MEDIUM COMPLEXITY THINGS. AND THEY ALLOW YOU TO GET STARTED WITH Q SHARP REALLY FAST. SO IT IS REALLY A COMMUNITY THING. SHARE YOUR IDEAS, COLLABORATE AND ENGAGE AND DRIVE FOR THE QUANTUM REVOLUTION. OKAY. IF YOU WANT TO LEARN MORE, PLEASE EXPLORE OUR PRESENCE ON GITHUB AND THE DOC PAGES AND THIS MAIN QUANTUM PAGE. NOW I WANT TO SHIFT GEARS AND MENTION THAT WE HAVE A CHEMISTRY LIBRARY. THAT CHEMISTRY LIBRARY ALLOW US TO CAPTURE THE PROBLEMS WE MENTIONED EARLIER, LIKE THE NITROGEN MAZE AND THE CATALYST PROBLEM, FOR INSTANCE. YOU CAN MODEL THAT, FOR THAT, YOU NEED TO TYPICALLY A CHEMISTRY MODELING SOFTWARE. THESE SOFTWARES ARE KIND OF COMPLICATED. WE PARTNERED UP WITH INDUSTRIAL, WITH THE GOVERNMENT, SORRY, THE GOVERNMENT RESEARCH LAB CALLED PNNL. PACIFIC NORTHWEST LABS, IN WASHINGTON STATE. THEY HAVE A SOFTWARE THAT IS KIND OF A SOFTWARE CONSORTIUM PART OF, CALLED NW CHEM. THAT ALLOWS CHEMISTRY MODELING. COMPLEX SOFTWARE. WE BUILD INTERFACE SOFTWARE TO GENERATE QUANTUM SOFTWARE THAT WE EXECUTE IN OUR FRAMEWORK THAT WILL EVENTUALLY. >>, EVENTUALLY ALLOW YOU TO UNDERSTAND AND POSSIBLY UPDATING YOUR CHEMISTRY MODELS. SO THAT, THIS LIBRARY ALLOWS YOU TO DO THAT. WE NOW ARE OFFERING TO THE WORLD AND PEOPLE ARE USING IT STARTING TO EXPLORE IT. AND I’M VERY EXCITED TO ANNOUNCE OUR NEXT SPEAKER ANDREW FURSMAN FROM ONE QUEBEC. IT IS A START-UP BASED IN VANCOUVER CANADA. OUR SPEAKER TOOK THE WATER PLANE DOWN FROM VANCOUVER TO HERE. WHAT IS COOL IS, THEY ACTUALLY EVALUATING THAT LIBRARY. KICKING THE TIRES. TELLING US THE THINGS THAT WORK WELL. AND THE THINGS THAT DON’T WORK SO WELL. AS PART OF THAT LEARNING EXPERIENCE, WE TRY TO MAKE THE LIBRARY BETTER AND HAVE A GREAT OFFERING FOR OUR CUSTOMERS. SO ANDREW IS THE CEO OF THE COMPANY ONE CUBIC. HE WORKED IN THE VENTURE CAPITALIST SPACE FOR A LONG TIME. HE HAS START-UPS THAT WORK IN TELESECTOR, SATELLITE, FINANCIAL SECTOR AND HE LEADS THE COMPANY. HE WILL SAY A LITTLE BIT ABOUT THE COMPANY. JUST GIVE HIM A WARM WELCOME. [ APPLAUSE ]>> THANKS VERY MUCH. THANK YOU.>> ALL RIGHT. THANK YOU VERY MUCH FOR HAVING ME. YEAH, I WILL TELL YOU A LITTLE BIT ABOUT ONE CUBIC TO GIVE YOU CONTEXT. WE ARE ABOUT 6-YEAR-OLD COMPANY. JUST OVER A HUNDRED PEOPLE. WE RAISED ENOUGH MONEY THAT WE CAN ACTUALLY FOCUS ON SOME OF THE SLIGHTLY LONGER TERM TECHNOLOGIES BUT NOT SO MUCH THAT WE DON’T CARE AT ALL ABOUT RIGHT NOW. SO I THINK THAT IT IS KIND OF A PERFECT SEGUE FOR THE KIND OF DIFFERENT DEVELOPMENTAL TOOLS THAT WE ARE TALKING ABOUT TODAY. WE STARTED OFF DOING A LOT OF IP DEVELOPMENT. AND NOW WE ARE MOVING MORE INTO COMMERCIALIZATION. AND OUR REAL INTEREST IS IN SITTING HERE BETWEEN THE DIFFICULT INDUSTRIAL PROBLEMS THAT WERE ELUDED TO PREVIOUSLY AND THE DIFFERENT HARDWARE DEVELOPMENT SYSTEMS AND THE TOOLS THAT ARE AVAILABLE TO BE ABLE TO TACKLE THESE PROBLEMS THAT ARE NOW BEING DEVELOPED. AND WE HAVE REALLY, REALLY LIKED WHAT MICROSOFT HAS BEEN BRINGING OUT. IT HAS ALLOWED US TO DO A LOT OF THE WORK MUCH SOONER THAN MANY PEOPLE HAVE EXPECTED. WE ARE INTERESTED IN TALKING A LITTLE BIT ABOUT QUANTUM COMPUTING. NOT JUST BECAUSE OF THE FACT THAT WE THINK THAT IT IS INTERESTING, BUT BECAUSE SOME PEOPLE WAY ABOVE OUR PAY GRADES THINK IT IS ESPECIALLY INTERESTING. SORT OF COMMANDED TO CARE ABOUT QUANTUM COMPUTING ABOUT ONE OF THE FEW DEVELOPMENTS THAT WILL REALLY BE CHANGING WHAT IS HAPPENING OVER THE NEXT FEW YEARS. AND THIS IS ACTUALLY BEEN A GREAT BOOST AS YOU HAVE SEEN MANY PEOPLE STARTING TO TALK ABOUT THIS MORE PUBLICLY, BUT I THINK THAT THERE IS STILL A PRETTY SIZABLE GAP BETWEEN THE WORD QUANTUM COMPUTING AND WHAT PEOPLE REALLY ARE DOING WITH QUANTUM COMPUTERS AND WHAT THEY HOPE TO DO WITH THESE TECHNOLOGIES AS THEY COME ON BOARD. FOR ME, THE EASIEST WAY TO UNDERSTAND THE LITTLE WAY ABOUT WHAT WE ARE TRYING TO DO WITH QUANTUM COMPUTING AND WHY WE CARE ABOUT IT, LOOK AT THESE PHOTOS. WHAT I THINK IS ESPECIALLY INTERESTING ABOUT THIS IS NOT JUST THE KIND OF LOOKS LIKE A MICRO BREWERY. BUT MORE OF THE FACT THAT ANYBODY WHO IS WORKING IN A LABORATORY THAT IS ANYTHING FROM A MICRO BREWERY UP TO KIND OF CUTTING EDGE CHEMISTRY WORK RIGHT NOW, PROBABLY RECOGNIZES THIS. THEY RECOGNIZE IT BECAUSE THIS ISN’T THAT DIFFERENT FROM HOW WE DO CHEMISTRY TODAY. AND THE REASON THAT IS A BAD THING, IT IS BECAUSE ALMOST EVERY OTHER INDUSTRY THAT WE CAN THINK ABOUT HAS REALLY TRANSITIONED TO SOMETHING MORE LIKE THIS. THE WAY THAT I LIKE TO SORT OF DRAW AN ANALOGY IS, IT IS NOT A VERY GOOD IDEA FOR BEAUING — BOEING TO MAKE A BUNCH OF PLANES ANTHRO OFF CLIFFS TO SEE WHICH ONE FLIES. THAT IS THE IS STATE WE ARE IN WITH A LOT OF CHEMISTRY RIGHT NOW. THE REASON WE ARE IN THAT STATE, BECAUSE COMPUTERS ARE NOT PARTICULARLY WELL SUITED FOR THIS TASK. WE HAPPEN TO BE JOINED IN THE AUDIENCE TODAY BY MA IT IES, GREAT QUOTE, IF WE WANT TO ACTUALLY DO THE WORK WE DO IN THE LABS IN CHEMISTRY, WITH OUR COMPUTERS, IT IS NO PROBLEM AS LONG AS WE HAVE A PROBLEM THAT IS ROUGHLY THE SIZE OF OUR GALAXY. BUT SINCE I DON’T OWN THAT MACHINE JUST YET, AND BECAUSE WE DO NEED TO MAKE A PROFIT EVENTUALLY WITH THE COMPANY, YOU KNOW, WE ARE REALLY INTERESTED IN TRYING TO UNDERSTAND IF THERE ARE BETTER WAYS FORWARD. AND THERE REALLY ARE BETTER WAYS FORWARD. THERE ARE THREE MAJOR THINGS THAT I THINK ARE OF INTEREST. FIRST IS QUANTUM INSPIRED OPTIMIZATION OPERATIONS. WE’LL HEAR MORE ABOUT THAT IN A MOMENT. BASICALLY, RECAST YOUR PROBLEM IN A FORM THAT MAKES IT AMENABLE TO QUANTUM PROCESSING IS ALREADY HELPFUL ESPECIALLY IF YOU THEN HAVE DEVICES THAT ARE DESIGNED TO COMPUTE TO SOLVE THOSE SORTS OF PROBLEMS DIRECTLY. THE OTHER THING THAT YOU REALLY WANT IS A PATH TOWARDS THE SCALABLE QUANTUM COMPUTER. AND FINALLY YOU WOULD LOVE TO HAVE A DEVELOPMENT ENVIRONMENT THAT ALLOWS YOU TO TAKE ADVANTAGE OF BOTH OF THESE THINGS. THE NICE THING IS, WITH MICROSOFT, YOU HAVE A ONE STOP SHOP FOR THAT. THEIR PROJECT CATAPULT IS REALLY GREAT QUANTUM INSPIRED PLATFORM THAT WE WILL BE ABLE TO USE IN ORDER TO DEVELOP APPLICATIONS THAT CAN RUN TODAY SOLVING SOME OF THE DIFFICULT PROBLEMS MUCH THEN OF COURSE, THE HOLY GRAIL OF ALL OF THIS IS THE SCALABLE TOPOLOGICAL PROCESSOR THAT THEY ARE WORKING ON. SPECIFICALLY, THE Q SHARP ENVIRONMENT IS A GREAT WAY TO BUILD APPLICATIONS THAT WILL EVENTUALLY RUN ON THE SCALABLE PROCESSORS THAT WE CAN UTILIZE THE CODING ENVIRONMENT AND THE SEM LATER IN ORDER TO PROVE THOSE THINGS OUT TODAY. ONE CUBIC, WE SPEND A LOT OF OUR TIME THINKING ABOUT CHEMISTRY. NEAR TERM APPLICATIONS. ONE OF THE THINGS THAT WE DO AS WE PARTNER UP WITH LARGE CHEMISTRY COMPANIES IN ORDER TO WORK ON THE PROBLEMS THAT WE THINK THAT ARE GOING TO BE THE FIRST THINGS THAT GET OFF LOADED FROM IN THE LAB TO IN COMPUTER MANY AND SO ONE OF THE TYPES OF PROBLEMS THAT EVERYBODY IS REALLY EXCITED ABOUT, IT IS THE DEVELOPMENT OF NEW TYPES OF CATALYSTS, BASICALLY SAYING, IF YOU ARE TRYING TO DO SOMETHING LIKE DRY PAINT, WHAT CAN YOU TOSS INTO THAT PAINT TO MAKE THE PAINT DRY FASTER? AND EVEN THOUGH THAT SOUNDS LIKE A MOST BORING PROBLEM YOU COULD POSSIBLY IMAGINE, WATCHING PAINT DRY IS ACTUALLY OF SUPER HIGH VALUE PROBLEM. AND SOMETHING THAT WE WILL BE ABLE TO DO REALLY, REALLY EARLY ON WITH THESE QUANTUM DEVICES TO BE ABLE TO SIMULATE WHICH THINGS MIGHT IMPROVE THAT PROCESS. SO WE HAVE ACTUALLY PUT TOGETHER A QUICK DEMO. THIS ISN’T EXACTLY THAT PROBLEM BUT THIS IDEA OF THE SIMILARET RICK STRETCH OF A HYDROGEN RING IS LIKE A PROBLEM THAT HAS ALL OF THE SAME SORTS OF FEATURES THAT YOU WOULD EXPRESS WHEN YOU ARE TRYING TO SOLVE THE PROBLEM OF EXPERIMENTING, TO FIND NEW TYPES OF CATALYSTS. IDEALLY, WHAT YOU WANT TO DO IS BE ABLE TO KNOW IF YOU ARE GOING TO MAKE THOSE RINGS BE DIFFERENT SIZES, THEN YOU WANT TO BE ABLE TO UNDERSTAND WHAT IS THE ESSENTIALLY THE LOWEST ENERGY STATE AND WHAT SOURCE OF CHARACTERISTICS WILL THESE DIFFERENT THINGS HAVE. AND TO KAL — CALCULATE THAT, YOU ARE ABLE TO FIRE UP Q SHARP AND LOAD UP SOME OF THE CHEMISTRY THINGS WE TALKED ABOUT BEFORE. HERE WE HAVE INTERFACED SOMETHING LIKE QUANTUM SOLVER, BUT ESSENTIALLY IT IS REALLY, REALLY INTERESTING THAT ALL OF THIS IS PUT TOGETHER IN PACKAGES THAT EXIST RIGHT NOW. AND THAT Q SHARP ENVIRONMENT IS GREAT FOR BEING ABLE TO SPEAK NATIVELY IN THE LANGUAGE OF THE QUANTUM PROCESSOR. BUT ON TOP OF THAT, YOU CAN THEN ACCESS YOUR Q SHARP FROM A FAMILIAR LANGUAGE LIKE PYTHON. SO THIS, FOR EXAMPLE, IS A WAY FOR US TO BE ABLE TO CALL THAT CODE RIGHT OUT OF OUR FRIENDLY PYTHON ENVIRONMENT FROM ONE OF OUR JUPITER NOTEBOOKS. IN THE END, YOU CAN ACTUALLY RUN A LITTLE PROGRAM. SO WHAT I LIKE ABOUT THIS PAGE HERE, IT IS IF YOU LOOK AT THAT DARK BLACK LINE DOWN THE MIDDLE, THAT IS THE GROUND TRUTH OF WHAT THE REAL ENERGY STATE OF THIS SYSTEM IS. IF YOU LOOK AT THE BLUE DOTTED LEAN AND THE GRAY DOTTED LINE, THIS IS SORT OF THE ESTIMATED WAYS THAT WE WOULD USE TO TRY TO USE A CLASSICAL COMPUTER TO BE ABLE TO SIMULATE THIS RIGHT NOW. AND UNFORTUNATELY, THEY ARE JUST NOT ACCURATE ENOUGH TO BE USABLE. BUT WITHOUT EVEN HAVING A QUANTUM COMPUTER AVAILABLE TODAY, BY FRAMING THE PROBLEM IN A WAY THAT SORT OF THINKING ABOUT THIS AS A QUANTUM CHEMISTRY PROBLEM FOR QUANTUM COMPUTERS, BY UTILIZING THE SOFTWARE TOOLS THAT MICROSOFT HAS PROVIDED, AND BY ALLOWING US TO USE THESE SIMULATOR, WE HAVE BEEN ABLE TO WRITE SOFTWARE THAT WE CAN PUT TOGETHER, RUN THROUGH THE PIPELINE THAT WE DISCUSSED AND ALL OF THE DATA POINTS THAT CAME OUT, THESE RED X’S HERE THAT REALLY, REALLY DO A GREAT JOB OF SUMMARIZING THAT CURVE. SO THIS IS EXACTLY THE KIND OF STUFF THAT WE CAN ALREADY DO TODAY BUT WHAT IS INTERESTING AS THESE MACHINES GET LARGER AND AS WE ARE ABLE TO DO BIGGER AND BIGGER SIMULATIONS, WE ARE ABLE TO SCALE THESE UP TO BE MORE AND MORE INTERESTING SIMULATIONS THAT ALLOW US TO BUILD COMPLETELY NEW THINGS AND HAVE A TOTALLY DIFFERENT RELATIONSHIP WITH THE BUILT WORLD BECAUSE WE ARE ABLE TO KNOW THE PROPERTIES OF DIFFERENT CHEMICALS THAT WE HAVEN’T PRODUCED BEFORE. WHEN I THINK ABOUT WHAT ONE CUBICS IS REALLY WORKING ON, THAT IS REALLY THE OLD STAR, THIS IS THE THING THAT WE WANT TO DO IN THE FUTURE. AND AT THE SAME TIME, WE ARE ABLE TO DO A LOT OF WORK TODAY BY PARTNERING UP WITH AZURE. FOR EXAMPLE, WE ARE ABOUT TO RELEASE A QUANTUM INSPIRED PLATFORM TO BE ABLE TO ADDRESS A WHOLE BUNCH OF PROBLEMS THAT WE ARE HE GOING TO HEAR ABOUT NEXT. THE WAY THAT I LIKE TO THINK ABOUT IT IS, IF YOU ARE LOOKING TO BE ABLE TO UNDERSTAND WHAT WE ARE GOING TO BE DOING WITH QUANTUM COMPUTERS IN THE NEAR FUTURE OR IF YOU ARE INTERESTED IN BEING ABLE TO HARNESS ALL OF THE THINKING THAT GOES INTO BUILDING THE DEVICES TO SOLVE REALLY DIFFICULT PROBLEMS TODAY, YOU CAN DO ALL OF THAT THROUGH MICROSOFT AND THROUGH MICROSOFT AZURE. AND TO HEAR A LITTLE BIT MORE ABOUT THE QUANTUM INSPIRED THINGS AT MICROSOFT, WELCOME TO THE STAGE, OUR FRIEND AND COLLEAGUE DR. HELMUT KATZGRABER. HE WILL TAKE THE LONG WAY AROUND TO COME UP HERE. AND ENTERTAIN YOU THE REST OF THE PROGRAM.>>THANK YOU, ANDREW.>>THANKS A LOT. [ APPLAUSE ]>>THANK YOU ALL FOR STILL STAYING HERE IN THIS LATE TIME OF THE DAY. SO WHAT I LIKE TO DO IS SHIFT GEARS A LITTLE BIT, TALK ABOUT WHAT WE CALL QUANTUM INSPIRED OPTIMIZATION. AS YOU SEE, IT IS OPTIMIZATION WITH A QUANTUM TWIST TO IT. NOW BEFORE I TELL YOU A LITTLE BIT MORE ABOUT QUANTUM INSPIRED OPT MY NATION. I WOULD LIKE TO SAY MORE ABOUT QUANTUM HARDWARE AND PULL SOME MISCONCEPTIONS OUT OF THE ROOM. THEIR SO-CALLED DIGITAL AND ANALOGUE QUANTUM COMPUTING DEVICES AND VERY DIFFERENT IN THEIR NATURE. DIGITAL DEVICES TEND TO NOWADAYS TYPICALLY HAVE A SMALL INJURY CUBIC COUNT. PROGRAMMABLE, FOR EXAMPLE, BY A Q SHARP. IN SOME SENSE, THE SAME WAY YOU WOULD PROGRAM A REGULAR COMPUTER. PROGRAMMING LANGUAGE. IMPLEMENT A PROBLEM. SOLVE IT ON THE DEVICE. EITHER OUT OF IONS, SUPER CONDUCTING FLEX CUBICS OR LIKE IN MICROSOFT CASE, TOPO LOGICAL CUBICS. IN PROJECTS, YOU HAVE QUANTUM YIELDING MACHINES. SPECIAL PURPOSE DEVICES. THINK OF THEM AS A MACHINE THAT DECIDE TO DO ONE PARTICULAR TASS, IN THIS CASE, OPTIMIZATION. THEY ARE ANALOGUE. MEANING THAT YOU KIND OF GET TO LOOK AT THE PROBLEM. THERE IS ALWAYS ERRORS ATTACHED TO THIS. AND THEY ARE MADE OUT OF SUPER CONDUCTING FLEX CUBICS. WHAT I WANT TO TALK ABOUT TODAY, BASICALLY A THIRD PARADIGM THAT IS NOT QUITE QUANTUM BUT USE SOME OF THE ADVANTAGES OF QUANTUM ON CURRENT HARDWARE. THIS IS WHAT WE LIKE TO CALL QUANTUM INSPIRED. AND WE RUN ON STANDARD SEAMLESS HARDWARE OR ACCELERATED ON FPJS. THE IDEA HERE TO EMULATE QUANTUM PROCESS ON CURRENT HARDWARE AND SOLVE CURRENT INDUSTRY CHALLENGES MORE EFFICIENTLY. NOW THE FOCUS OF THIS TALK WILL BE OPTIMIZATION. AND PARTICULAR OPTIMIZATION. WHY? THE ANSWER IS SIMPLE, OPTIMIZATION PLAY AS VERY CRUCIAL ROLE IN EVERYBODY’S LIFE. IF YOU WANT TO SHIP SOMETHING ACROSS THE WORLD, IF YOU WANT TO DO A SCHEDULING PROBLEM, IF YOU WANT TO DO SEND A ROCKET IN SPACE, YOU NEED TO SOLVE VERY HARD OPTIMIZATION PROBLEMS. THERE IS A LOT OF VALUE TO THIS. AND A LOT OF CHALLENGES. QUANTUM OPTIMIZATION MAINLY FOCUSED ON BINARY PROCESS. TAKE ZERO ONE, OR PLUS MEENOUS ONE. THE REASON IT IS VERY SIMPLE. A HUGE INDUSTRIAL VALUE. I WILL SHOW YOU A BROAD SELECTION OF THIS PROBLEM IN THE NEXT TWO SLIDES. THEY HAVE A SIMPLE MATHEMATICAL FORMALIZATION. BECAUSE OF THIS, DON’T WORRY, NOT SHOW YOU ANY EQUATIONS, THEY ARE VERY WELL SUITED FOR QUANTUM OPTIMIZERS. I AM GOING TO TRY TO GIVE YOU A BIT OF A FLAVOR HOW THIS QUANTUM OPTIMIZATION WORKS. BASICALLY, BY GOING BACK A FEW THOUSAND YEARS TO THE ALY IT HIC ERROR. PEOPLE TAKE CHUNK OF METAL, SLOW IT DOWN SLOWLY, THIS PROCESS. NOW BACK IN THE EARLY 80s, TIME AT INTEL, CAME UP WITH A CLEVER IDEA. WHY NOT EMULATE THIS PROCESS ON A COMPUTER. CALLED THIS PROCEDURE SIMULATED AMELING. BASICALLY LOOKING AT CHIP LAYOUT FORCE THE MICRO PROCESSORS AT THE TIME. THE IDEA IS SIMPLE. YOU BASICALLY START WITH A CONFIGURATION THROUGH OPTIMIZATION. IT CAN BE ANYTHING. OKAY. THE IMAGINATION IS YOUR LIMIT. AND YOU ATTACH A TEMPERATURE, THINK OF IT REALLY AS A TEMPERATURE IN THE SENSE OF WHAT BOILS WATER. SO YOU START FROM THE VERY HIGH TEMPERATURE. WHERE YOUR VARIABLES FOR YOUR SYSTEM, THINK AGAIN, THE WATER IS BOILING. AND THEN YOU REDUCE THE TEMPERATURE ACCORDING TO A SCHEDULE. UNTIL YOU REACH SOME TARGET VALUE. AND IF YOU DO THIS SLOW ENOUGH, YOU MIGHT FIND THE OPTIMUM OR THE SOLUTION TO THE PROBLEM. OKAY. IN THE CASE OF WATER AGAIN, IF YOU COOL IT SLOW ENOUGH, YOU MIGHT GET A WONDERFULLY PERFECT ICE CUBE WITHOUT ANY IMPERFECTS IN IT. QUANTUM AMELING, COUNTERPART TO SIMULATE, ALL YOU HAVE TO DO IS REPLACE THE QUANTUM FLUCUATES, REPLACE THE THERMAL FLUCUATES, QUANTUM FLUCTUATIONS, IF I NOW PRESS CLICK ON THIS, THIS ALGORITHM WILL CHANGE EVER SO SLIGHTLY. WHAT IS TEMPERATURE BEFORE, IS NOW TRANSFER. AND BASICALLY INSTEAD OF DOING [INDISCERNIBLE], SO YOU RUN SIMULATED ON A REGULAR COMPUTER, YOU DO QUANTUM UPDATES. IF YOU WANT TO DO THIS IN HARDWARE, WHAT YOU DO IS YOU USE SO-CALLED SUPER CONDUCTING FLEX CUBICS, SUPER CONDUCTING METAL RINGS. AS YOU KNOW, IF YOU HAVE A CURRENT, WHO DOES REMEMBER HOW THE LAW WORKS? NOBODY TOOK PHYSICS HERE? A COUPLE HANDS. ACCORDING TO THE LAW, WHEN YOU HAVE A RING OR A CURRENT, GOING COUNTERCLOCKWISE, YOU HAVE A RING, COUNTERCLOCKWISE, FLEX POINTING DOWN. STATE ZERO IN ONE. YOU CAN’T HAVE SUPER POSITIONS OF ZERO AND ONE. THIS GIVES YOU YOUR FLEX CUBICS. OKAY. NOW WHAT YOU THEN DO IS, YOU STRING THIS SUPER CONDUCTING LOOPS TOGETHER BY A SOME KIND OF INTERACTION MATRIX. IN THIS IS WHAT YOU USE TO CODE YOUR PROBLEM. BINARY MATHEMATICAL VARIABLES. PRESENT THEM ON THE HARDWARE. APPLY A STRONG TRANSVERSE FIELD. YOU REVIEW THE FLUCTUATIONS SLOWLY. IF THERE IS SOME KIND OF BUMP OR BARRIER IN YOUR CROSS FUNCTION LANDSCAPE, THE QUANTUM SYSTEM MAY BE ABLE TO TUNNEL THROUGH THE BARRIER AND FIND THE OPTIMUM MORE EFFICIENTLY WITH THOROUGHLY ASSISTED ALGORITHM. GOOD. NOW THIS ONE IS GREAT. AND YOU ARE USING QUANTUM MECHANICS TO SOLVE THE PROBLEM. WHAT IS REAL FOR CURRENT QUANTUM MACHINES? THE REALITY IS A LITTLE BIT BLEAK. IN THAT THERE IS A FEW CHALLENGES TO OVERCOME. THE FIRST ONE IS, WHAT WE CALL THE EMBEDDED OVERHEAD. SUPPOSE YOU WANT TO STUDY A PROBLEM. THINK OF IT AS A VERTEX COVER PROBLEM, A MATCHING PROBLEM, WHATEVER. SOMETHING THAT LEAVES ON THIS TIPPOLOGY HERE. WHAT YOU SEE ON THE LEFT HAND SIDE. NOW YOU SEE THIS IS YOUR ACTUAL PROBLEM WITH YOUR LOGICAL VARIABLES THAT YOU WANT TO TREAT. BUT BECAUSE AS YOU SAW BEFORE, WE HAVE A HARD WIRE CHIP THAT LIVES ON THE SQUARE GRID. WE NEED TO TAKE THIS LOGICAL VARIABLE ONE AND REPLICATE IT FOUR TIMES. SUCH YOU CAN FULFILL ALL OF THE INTERACTIONS AT VARIABLE ONE HAS WITH THE NEIGHBORS. AND SO YOU NEED TO STRONGLY COUPLE THIS PHYSICAL VARIABLES HERE, TO CREATE ONE LOGICAL VARIABLE ON THE LEFT HAND SIDE. YOU SEE IT RIGHT AWAY IN THIS SIMPLE EXAMPLE, YOU HAVE ABOUT A 50 OVERHEAD IN THE NUMBER OF VARIABLES THAT YOU NEED TO SOLVE THE PROBLEM. THIS CAN GET VERY, VERY EXPENSIVE. AS I SAID BEFORE, YOU HAVE ANALOGUE POSITION, TYPICALLY 5-6 BITS POSITION AS YOU ALL KNOW, THAT IS NOT VERY MUCH FOR A COMPUTER. AND YOU ARE LOCKED IN WITH AN ALGORITHMIC ENGINE THAT IS QUANTUM, AND STRONG TRANSVERSE FIELD, REDUCE THE TARGET VALUE. YOU ARE UNABLE TO HANDLE COMPLEX CROSS FUNCTIONS FOR THOSE WHO COME FROM OR, THINGS OF MIXED INTEJER PROBLEMS. TO DATE, COMFORTABLE IN SAYING THIS, NO REAL COMPLICATION, SOMETHING THAT MATTERS IN INDUSTRY, THE ALGORITHMS THAT I WORK ON, NOT OUTPERFORMED THE MACHINES. THIS IS WHY WE ARE SO INTERESTED IN THIS QUANTUM INSPIRED TECHNIQUES. HAVING THIS CLAIM OF SPEED UP, IS REALLY WHAT FUELED MY INTEREST IN THIS PROBLEM AND CREATED A FRUITFUL COMPETITION THAT MADE US DEVELOP NEW METHODS THAT CAN SOLVE THESE PROBLEMS THAT WERE CONSIDERED TO BE VERY, VERY HARD MUCH MORE EFFICIENTLY THAN BEFORE. MORE IMPORTANTLY, IF YOU RUN THIS ON A REGULAR CPU, WE HAVE ABSOLUTELY NO INTRINSIC LIMITATION WHICH IS GREAT NEWS. FINALLY, IF YOU TRIED QUANTUM ON A PROBLEM, WE CAN DO IT FASTER, CHEAPER AND BETTER. IF WE JUST EMULATE THE QUANTUM PROCESSES ON CLASSICAL HARDWARE. NOW I HAVE BEEN USING THIS WORD QUANTUM INSPIRED OPTIMIZATION. BUT I YET TO TELL YOU WHAT IT ACTUALLY MEANS. IT IS LOOSELY DEFINED. IF YOU LOOK IN THE DICTIONARY, INSPIRED MEANS TO INSERT INFLUENCE ON SOMETHING. OKAY. HERE WHEN I SAY VERY LOOSELY, DEFINED, YOU HAVE THE WHOLE RANGE IN THIS SPECTRUM. ON ONE SIDE, YOU ARE TRYING TO REALLY EMULATE THE QUANTUM PROCESS. FOR EXAMPLE, BY A QUANTUM OF THE CARD, OR QUANTUM ON COMPUTER. FOR EXAMPLE, IF YOU HAVE THIS ROUGH CROSS FUNCTION LANDSCAPE THAT YOU SEE HERE, YOU WANT TO SOLVE SIMULATE, THEN YOU HAVE TO HAVE A LARGE ACTIVATION ENERGY TO OVERCOME THIS BARRIER. AND SUCH A MOVE WILL BE VERY UNLIKELY IN YOUR SIMULATION. IF YOU WERE TO DO THIS WITH PURE QUANTUM MONTE CARLO, THEN YOU MIGHT POTENTIALLY TUNNEL THROUGH THE BARRIER AND FIND THE OPTIMUM FASTER. IN QUANTUM INSPIRED OPTIMIZATION WE ARE TRYING TO TAKE THE BEST WE CAN OUT OF QUANTUM AND IMPLICATE IN CLASSICAL HARDWARE AND TRY TO DEPICT THIS BY HAVING IMPROVEMENT OVER CLASSICAL MAYBE NOT THE FULL BENEFIT OVER QUANTUM. THIS IS ONE EXTREME. THE OTHER EXTREME IS ONE THAT I AM MOST INTERESTED IN. WHICH IS USE ALGORITHMS THAT ARE COMMONPLACE IN PHYSICS AND THAT WE PHYSICISTS HAVE USED OVER DECADES TO SOLVE REALLY HARD PROBLEMS IN INDUSTRY. AND TO SHOW YOU HOW WELL WE WORK, LET ME SHOW YOU A CASE STUDY. SOMETHING THAT IS BACK IN 2016. SOME OF YOU MIGHT HAVE HEARD OF A SOUTH COMPETITION, THE COMPETITION, BASICALLY, SAT IS A PROBLEM, SOLVABILITY PROBLEM WITH HUGE INDUSTRIAL VALUE. A LOT OF THE PROBLEMS IN OPTIMIZATION GET MAPPED INTO THESE FORMULAS. IF YOU WANT A PROOF OF SOMETHING IS P OR NP, THEN YOU END UP MAPPING IT A THREE SET FORMULA. IF IT MAPS, THEN IT IS HARD TO SOLVE. SO IN MAPS, THE GOAL, THE MAXIMUM NUMBER OF CLAUSES, EQUATE TO TRUE IN SUCH A BRILLIANT FORMULA. FOUR FORMULAS. THEY CAN BE NEGATED WHICH I DENOTE WITH OVERBAR. AND THEY ARE CONNECTED BY A LOGICAL ORS. THEN THE CLAUSES OF THREE EACH ARE CONNECT BID LOGICAL ANDS. AND THE QUESTION NOW IS, CAN YOU FIND ASSIGNMENT TO THESE VARIABLES SUCH A THAT YOU CAN MAXIMIZE THE NUMBER OF CLAUSES THAT ARE TRUE. WHEN YOU HAVE JUST FOUR VARIABLES AND TWO CLAUSES, IT IS PRETTY EASY. SIT DOWN WITH PAPER AND PENCIL. WHEN YOU HAVE ABOUT A HUNDRED OF THOSE, YOU ARE GOING TO BE DOING MATH FOR THE REST OF THE AGE OF THE UNIVERSE. NOW THERE IS A COMPETITION THAT HAPPENS EVERY YEAR BY THE ASSOCIATION WHERE PEOPLE HAVE BEEN SENDING IN ALGO RISMS ONE YEAR, WHY NOT SEND PHYSICS ALGORITHM TO THE COMPETITION. WE DID. WE ENDED UP WINNING ON THE FIRST MISSION. THIS THING HERE, WE CALL BEAURAL-IS, NOTHING BUT A INDUSTRIAL PROBLEM. NOT ONLY THAT, SECOND TEAM, FOLKS THAT ARE IN MICROSOFT, STEVEN JORDAN AND BRAD, WHO SUBMITTED [INDISCERNIBLE]. THEY LANDED FIFTH PLACE. OR DIDN’T MAKE IT TO THE FINAL ROUND. LIFE IS TOUGH SOMETIMES. NOW WHEN SHOULD YOU USE QUANTUM INSPIRED OPTIMIZATION OR QIO? WELL, THERE ARE CERTAIN USE CASES. SUPPOSE A GIVEN OPTIMIZATION PROBLEM IN FRONT OF YOU. ONE OF THEM IS, IF YOU WANT TO HAVE A FIXED SOLUTION QUALITY YOU WANT TO FIND A SOLUTION FASTER. OKAY. THIS IS DIFFICULT PROBLEM WHEN YOU SAY I WANT TO MAKE SURE THAT I HAVE AT LEAST SO MUCH QUALITY IN MY SOLUTION BUT I WANT YOU TO TELL ME THIS SOONER. AND IF YOU SAY I’M HAPPY ENOUGH, 90 OF THE PROBLEM SOLVED BUT ONLY GIVE YOU ONE SECOND TO DO THIS. FOR A FIXED SOLUTION TIME, FIND A BETTER SOLUTION. THIS IS VERY IMPORTANT IN THINGS LIKE UNIT COMMITMENT PROBLEMS. WHEN TRYING TO OPTIMIZE A POWER GRID, PRE-COMPUTE THE POTENTIAL LOAD EVERY SO MANY HOURS MUCH OF COURSE, IF YOU HAVE A FIXED AMOUNT OF TIME YOU WANT TO FIND THE BEST SOLUTION POSSIBLE. AND THEN FINALLY, YOU WANT TO BE ABLE TO SOLVE MORE COMPLEX VERSIONS OF THE PROBLEM WITH FIXED EFFORT IN THE SENSE THAT WHAT IF THERE IS CONSTRAINTS, IF, I’M LOOKING AT A ROUTING PROBLEM, AND NOT ONE TO FACTOR IN THAT THERE ARE MORE TRUCKS PARKED IN OVER HERE THAN OVER THERE. WHAT ARE THE CHARACTERISTICS THAT OPTIMIZATION PROBLEM SHOULD HAVE FOR QUANTUM INSPIRED OBVIOUSLY IF IT LOOKS LIKE THE ONE UP THERE, IT IS NOT GOOD NEWS, ANY GRADE WILL FIND YOU THE OPTIMUM RIGHT AWAY. YOU WANT SOMETHING GNARLY, LIKE THIS LANDSCAPE. SOMETHING WITH A LOT OF BUMPS, SOMETHING LOCAL SEARCH ALGORITHM WILL GET STUCK RIGHT AWAY. YOU ALSO WANT THE LARGE NUMBER OF VARIABLES. SOMETHING WITH 20 VARIABLES YOU CAN EXACT NUMMERRATE ALL EXACT SOLUTIONS. IT DEFEATS THE PURPOSE. FINALLY YOU WANT THE CROSS FUNCTION TO BE EVALUATED QUICKLY. WHY? BECAUSE THE EVALUATION OF THE CROSS FUNCTION LIES AT CORE OF THE ALGORITHM. NOW I WANT TO EMPHASIZE THAT THESE METHODS, IN OTHER WORDS, YOU DON’T GET A GUARANTEED SOLUTION. HAVING SAID THAT, OFF 1 IMPROVEMENT CAN BE HUGE IF YOU SCALE UP TO INDUSTRIAL SCALES. A HUGE COLLECTION OF PROBLEMS THAT CAN BENEFIT FROM THIS QUANTUM INSPIRED MATH. STARTING WITH CIRCUIT FULL DIAGNOSIS TO VERIFICATION AND VALIDATION, WHICH ARE VERY IMPORTANT IF YOU WANT TO, FOR EXAMPLE, BUILD COMPONENTS OR BUILD A JET THAT FLIES. TO CHEMICAL OIL AND GAS, MATERIAL DISCOVERY, I’LL SHOW YOU SOME RESULTS OF THAT LATER. UNIT COMMITMENT AS I SAID, POWER GRIDS, SIGNAL PROCESS, MACHINE LEARNING TO A LESSER EXPENT TENT BECAUSE THE LANDSCAPES IN THIS PROBLEM TEND TO BE MORE CON VEC, THEY ARE EASIER. DROP SHOP SCHEDULING AND TOOL OPTIMIZATION. IF YOU HAVE A PLAN TO BUILD A VEHICLE, HOW TO DISTRIBUTE THE MACHINES ON TO DIFFERENT TEAM FORCE DIFFERENT WORKERS TO MAKE THIS HAPPEN FASTER. AND FINALLY, OF COURSE, TRAFFIC CAR SHARING AND LOGISTICS. WE’LL SHOW YOU LATER, WE ALSO HAVE FAR MORE EXOTIC APPLICATIONS SUCH AS OPTIMIZING SEQUENCES. NOW IF YOU WANT TO COMPARE QUANTUM VERSUS QUANTUM INSPIRED OPTIM IZATION, THEN FIRST WE HAVE TO SET THE PLAYING FIELD, WHAT DO WE MEAN BY DISRUPTIVE TECHNOLOGY? IT SHOULD BE FASTER THAN ANY ALGORITHM OUT THERE RIGHT NOW. OR IT SHOULD HAVE BETTER PERFORMANCE ON SOME SPECIFIC TASKS. SO WHAT IS FASTER? AGAIN, HERE, THERE ARE TWO CASES OF FASTER. LET’S ASSUME THAT YOU ARE LOOKING AT A PROBLEM THAT HAS A SPECIFIC NUMBER OF VARIABLES. AND YOU KEEP INCREASING THE NUMBER OF VARIABLES AND YOU LOOK AT THE TIME IT TAKES YOU TO SOLVE THE PROBLEM. TYPICALLY THE THINGS SCALE. THEY ARE BAD NEWS FOR ANY KIND OF ALGORITHM. SO WE’LL GIVE YOU ROUGHLY A STRICT LINE. FASTER, WHICH IS THE IDEAL CASE, CAN MEAN THAT YOU GET A BETTER SCALING. IN OTHER WORDS, THE SLOPE OF THE CURVE IS SLIGHTLY MORE FLAT. BUT FASTER CAN ALSO MEAN A HUGE UPSET. IF THIS IS A BILLION, THEN IT IS DEFINITELY WORTH IT. EVEN THOUGH THE SAME SCALING EXISTS. WHAT IS BETTER PERFORMANCE, WELL, FOR EXAMPLE, DOES IT GIVE YOU MORE VARIETY POOL OF SOLUTIONS? WILL IT HELP YOU IN ADDING ADDITIONAL CONSTRAINTS? THE MOST IMPORTANT THING THAT I WANT YOU TO TAKE AWAY HERE IS THAT WHEN YOU SEE ANY CLAIMS OF SPEED-UP IN THE NEWS, YOU HAVE TO BE VERY CAREFUL. BECAUSE IT STRONGLY DEPENDS ON THE BENCHMARK YOU USE. A SIMPLE EXAMPLE AS A FOLLOWING. YOU WANT TO RAISE A FORMAL OF ONE VEHICLE AGAINST A RALLY CAR. FORMAL ONE WEEK, FASTER. IF YOU ARE DRIVING ON DIRT, IT IS PRETTY CLEAR WHICH ONE IS GOING TO WIN. OKAY. ALWAYS BALANCE THESE THINGS OUT. LET ME SHOW YOU AN EXAMPLE IN 2016, INTERESTING RESULTS USING THE QUANTUM AMELIER. IT IS A TIME TO SOLVE THE PROBLEM. FUNCTION OF VARIABLES. USING THE MACHINE LABELED HERE WITH QA. SIMULATING THE MACHINE ON CLASSICAL MACHINES LABELED WITH QUANTUM MONTE CARLO AGAINST SIMULATED, THE THERMAL COUNTERPART THAT I SHOWED YOU BEFORE. WHEN YOU LOOK AT THIS DATA, YOU CAN SEE RIGHT AWAY THAT THE QUANTUM APPROACH IS SCALE BETTER, THE SLOPE IS FLATTER. AND NOT ONLY THAT, BUT THERE IS A HUGE OFFSET OF EIGHT ORDERS OF MAGE MAGNITUDE HERE. WHEN YOU LOOK AT THIS, IT SOUNDS REALLY IMPRESSIVE FOR QUANTUM OPTIMIZATION. SO AGAIN, WE SAID, SOMEBODY PLEASE HOLD MY BEER. LET’S SEE WHAT OUR ALGORITHMS CAN DO. I DON’T WANT YOU TO TRY TO FIGURE OUT WHAT IS WHAT. JUST LOOK AT THE SPAGHETTI ON THE SCREEN. AGAIN HERE, YOU HAVE QUANTUM MONTE CARLO, SIMULATED, AND THE DEVICE AND A BUNCH OF OTHER ALGORITHMS. MORE IMPORTANTLY, THE PURPLE LINE DOWN HERE, THAT SCALES REALLY WELL, AND COMPARABLE IN SPEED, IT IS BURIALOUS. THE ALGORITHM THAT WON THE COMPETITION. AND I’M GOING TO DO NOW, MAKE THIS MORE PALATABLE FOR YOU, COMPUTE THE SLOPE FORCE THE LARGEST SYSTEM SIZES. IN OTHER WORDS, PERFORM A SCALING ANALYSIS AND JUST SHOW YOU THE SLOPES FOR DIFFERENT ALGORITHMS. OF COURSE, SMALLER MEANS BETTER. YOU CAN SEE RIGHT AWAY, THE QUANTUM APPROACH IS HERE ARE VERY GOOD. BUT THEY ARE NO MATCH AGAINST OUR ALGORITHMS. NOTICE ALSO THAT I KIND OF COLORED IN THIS FIGURE. ON THE LEFT HAND SIDE, HAVE YOU WHAT WE CALL SEQUENTIAL ALGORITHMS. THESE ARE SIMULATING AND POPULATION, AND MONTY CARLO. YOU HAVE A CONTROL PARAMETER. A STRONG TEMPERATURE, HIGH TEMPERATURE, AND YOU SEQUENTIALLY REDUCE TO TARGET VALUE. SPECIFIC TYPE OF ALGORITHM. THEN YOU HAVE THE TAILORED ALGORITHM, THAT IS CHEATING. I A KNOW AHEAD OF TIME WHAT THE PROBLEM IS SUPPOSED TO SOLVE. EXPLOIT SOME SIGNATURE OF THE PROBLEM. THEN YOU HAVE GENERIC ONES LIKE MONTE CARLO AND BASED ON TEMPERING MONTE CARLO THAT IS CLEARLY COMPETITIVE AND OUTPERFORM QUANTUM OPTIMIZATION. SO WE THOUGHT THAT WE WOULD CAPITALIZE HERE AT MICROSOFT. AND DECIDED TO LOOK INTO THIS MORE CLOSELY. WHAT IS OUR APPROACH? FIRST AND FOREMOST, DEVELOP COMBINE AND IMPROVE ALGORITHMS. THIS IS VERY IMPORTANT. I’LL GET BACK TO THIS AGAIN. WE WANT TO BUILD A FLEXIBLE STACK WITH MODULAR ALGORITHMIC DIVISION THAT IS MASSIVELY PARALLEL. YOU SEE, NOT EVERY ALGORITHM IS MEANT FOR EVERY PROBLEM. YOU WANT TO BE ABLE TO SWAP THAT IN AND OUT. DEPENDING WHAT YOU ARE LOOKING AT. ADVANTAGES ARE, THAT THE BIGGEST SPEED-UPS TO DATE, COME FROM BETTER ALGORITHMS. YOU SEE, IF YOU WANT TO DOUBLE THE SPEED YOU HAVE TO DOUBLE THE SIZE OF THE MACHINE. IF YOU COME UP WITH A BETTER ALGORITHM, THAT SCALES BETTER, YOU DON’T HAVE TO DOUBLE THE MACHINE. DEVICES ARE DIGITAL, MEANING THAT WE HAVE A LARGER APPLICATION DOMAIN AND BECAUSE WE CAN DO WALL TO WALL CONNECTIVITY, WITH LOCAL INTERACTIONS, EXPLAIN LATER, DIGITAL POSITION, NO NEED FOR MAPPING OR EMBEDDED. WE DEVELOP THE CODES AND CPU’S. WE RUN THEM NOW IN FPJ’S FOR SPEED AND HOPEFULLY ONE DAY RUN THEM ON QUANTUM ACCELERATED HARDWARE. LET ME SHOW YOU HOW WELL THIS METHOD WORKS FOR REAL WORLD PROBLEM. IN 2017, [INDISCERNIBLE] PUBLISHED FOLLOWING STUDY WHERE QUANTUM WAS USED TO OPTIMIZE THE TRAFFIC OF 480 VEHICLES IN BEIJING HEADING TO THE AIRPORT. THE HEAT HERE REPRESENTS THE TRAFFIC. NOW AS YOU SEE WITH TRANSITION, TRADITIONAL METHODS TAKES ABOUT 800 SECONDS TO SOLVE THE PROBLEM. IF YOU SAW IT ON THE QUANTUM ANEL IER, YOU ARE LOOKING AT 20 SECONDS. OUR APPROACH, 20. FPJ, 0. 0004 SECONDS. NOW THIS POINT, I WOULD LIKE TO JUST SHIFT GEARS A LITTLE. AND SHOW YOU NOW HOW TO CAPITALIZE ON THIS TO SOLVE REAL WORLD PROBLEMS TODAY. AND HERE I WANT TO HIGHLIGHT FOUR OF THE CORE STRENGTHS THAT WE HAVE. ONE IS ALGORITHMIC EXPERTISE. LARGE SCALE SCALABLE HARDWARE. MODULAR SOFTWARE. AND MORE IMPORTANTLY, DOMAIN EXPERTISE. HAVING PEOPLE THAT CAN THINK ABOUT THE PROBLEMS AN CAST THEM IN A FORM THAT MAKES THEM EASIER TO SOLVE IS AS IMPORTANT AS HAVING GOOD HARDWARE IF NOT MORE. THE FIRST ONE I WANT TO TALK ABOUT, HIGHLIGHTS OUR ALGORITHMIC EXPERTISE IS OPTIMIZING SEATTLE’S TRAFFIC. I THINK THAT EVERYBODY HAD AN ISSUE, UNLESS YOU HAPPEN TO BE STAYING AT THE SHERA IT ON, GETTING TO THIS BUILDING TODAY. AVOID IT ALL TIMES. SO THE OBJECTIVE WAS TO OPTIMIZE RUSH HOUR TRAFFIC USING THE QUANTUM INSPIRED OPTIM GLIERKS TECHNIQUES. WHAT IS CURRENT STARTER, IF WE TAKE BING’S BEST ACCOMMODATIONS AND LOOK AT REALTIME TRAFFIC DATA, FIND 50 OR LESS CONGESTION. AND WE CAN REDUCE TRAVEL TIMES BY ABOUT 8. NOW REMEMBER THE CASE STUDY I SHOWED YOU BEFORE, WAS ON 480 VEHICLES. WE THOUGHT LET’S KICK IT UP A NOTCH. FOR THIS, WE HAD FRANCES AND ARETHA, SITTING HERE IN THE FIRST, SECOND ROW, BEING KIND OF SHY AND NOT DOING THIS. AND DOING A PHENOMENAL JOB AND IMPLEMENTING THIS WITH REND THIS FOR 5, 000 VEHICLES. WELL, HERE YOU HAVE A LIVE DEMO ON THIS. AND YOU BASICALLY SELECT THE NUMBER OF VEHICLES, WE LOOK AT THE 5, 000 VEHICLES OUT OF 5, 000 RANDOM LOCATIONS IN SEATTLE. WE KEEP EACH VEHICLE TEN DIFFERENT OPTIONS TO TRAVEL. AND NOW YOU SEE THIS IS NASTY TRAFFIC. YOU SEE AT THE BOTTOM HOW THE CONGESTION SCORE KEEPS GOING DOWN. AFTER 2, 000 STEPS, OR OPTIMIZER, IF YOU COMPARE THE RESULTS, I DON’T THINK THAT I NEED TO CONVINCE YOU THAT THINGS ARE MUCH BETTER ON THE RIGHT HAND SIDE THAN THE LEFT HAND SIDE. AGAIN NOT MY JOB. THESE TWO DID THE BRUNT OF THE WORK. THEY ARE AWESOME. GOOD. NOW THE NEXT EXAMPLE I WANT TO SHOW YOU, IT IS ACTUALLY WORK TOGETHER WITH OUR PARTNER ONE CUBIC. AND HERE I WANT TO SHOW YOU HOW TO TACKLE HEART OF THE PROBLEMS IN CHEMISTRY. WHAT WE DID HERE ON THE SOFTWARE SIDE IS USE MICROSOFT QUANTUM INSPIRED TOOL BOX. AND ON THE HARDWARE, THIS IS WHAT I WANT TO HIGHLIGHT HERE. WE HAVE A CONFIGURED FPJ CLOUD IN A MASSIVE SCALE. THIS IS CATAPULT VERSION TWO. WHAT IS IT BASICALLY SEVERAL, A LOT OF THE SERVER BLADES IN AZURE HAS BUILT-IN FPJ’S. NOW THIS BY ITSELF IS NO NEWS. WHAT IS AMAZING ABOUT IT IS AS YOU CAN SEE THE FPJ CONNECTED. NOT ONLY THAT, YOU HAVE A FABRIC THAT CONNECTS FP JOVMENT — FPJ CONNECTED. THAT MEANS WE CAN SOLVE THINGS THAT MOST PEOPLE WOULD SAY TODAY, THERE IS NO WAY OF SOLVING THEM. SO TO JUST SHOW YOU WHAT WE CAN POTENTIALLY DO, I JUST LIKE TO SHOW YOU AN EXAMPLE OF WORK THAT WAS DONE [INDISCERNIBLE] ONE CUBIC, GRAPH BASE MOLECULAR NARRATIVE. THIS IS KEY, IF YOU WANT TO MAKE SURE THAT YOU DON’T WAIT TOO LONG FOR PAINT TO WAIT. HOW MUCH DO THESE MOLECULES AGREE? WHAT YOU DO IS, YOU MAP THE MOLECULES TO A GRAPH. THEN YOU CREATE WHAT IS KNOWN AS THE CONFLICT GRAPH. THIS IS THE HAIR BALL THAT YOU SEE ON THE LEFT. YOU CAN SEE IT IS VERY CONFLICTIVE IN ITS NATURE. YOU SAW THIS SO-CALLED CO-K PLEX PROBLEM. WHICH YOU SEE HERE BY COLORING IN CERTAIN OF THE DOTS. AND WE CAN DO THIS ABOUT 500 TIMES FASTER ON THE FPJ WITHOUT THE PARALYZATION. THEN YOU GO BACK AND YOU DECORATE YOUR GRAPHS. AND THEN YOU DO SCORING AND MAP IT BACK ON THE MOLECULES AND YOU CAN SEE WHICH MOLECULES AGREE AND DISAGREE. WHY IS THIS IMPORTANT? NOT JUST IMPORTANT IF YOU WANT TO SEE PAINT DRY FASTER, BUT THINK ABOUT IT. IF YOU COMPARE THE MOLECULES AND ENCODING, THEY ARE VERY, VERY SIMILAR. YOU WOULD LIKE TO HAVE A COATING THAT IS STRONGER BUT DOESN’T HAVE THE SIDE EFFECTS. THE ONLY WAY TO FIND THIS MOLECULES IS BY COMPARING THEM. AGAIN ACCESS TO LARGER MOLECULES. THE NEXT CASE STUDY WHERE ONCE PILOT THE MODULAR SOFTWARE WITH A PARTNER. ON THE RIGHT HAND SIDE, YOU ARE WATCHING A VIDEO OF ONE OF OTI’S TRANSPARENT ORGANIC SCREENS THAT YOU CAN USE FOR HEAD OF THIS PLACE. THIS IS REALLY COOL. THEY ARE SPECIALIZED IN MAKING THESE SCREENS. AND CREATE NEW ORGANIC MOLECULES FOR THIS PLAY TECHNOLOGY. SO THEIR APPROACH, COMBINES MACHINE LEARNING, SIMULATION, OPTIMIZATION STEP THAT I’M AFTER. WITH PROPERTY TESTING AND MATERIAL AND VALIDATION. IF THEY DON’T FIND THE MATERIAL IN THE PROPERTIES THEY JUST ITERATE WHAT THEY FOUND THROUGH THEIR DATABASE AND KEEP GOING UNTIL THEY HAVE WHAT THEY WANT. I WANT TO FOCUS HERE ON THIS. THEY CAN HANDLE MATERIALS WITH 250 MOLECULES. THEY HAVE THE MATERIAL DISCOVERED ON THE QUANTUM. WE SAID, LET’S HAVE A LOOK AT THIS. WHAT IS THE DIFFERENCE OF WHAT WE DO TO WHAT IS TYPICALLY DONE IN QUANTUM OPTIMIZATION? QUANTUM OPTIMIZATION, YOU ARE LOCKED IN HANDLE TWO LOCAL [INDISCERNIBLE]. THESE ARE PROBLEMS WHERE EACH VARIABLE, THE BLUE DOT HERE CAN ONLY INTERACT WITH ONE OR THE OTHER NEIGHBORS. THINK OF IT AS A PARTY WHERE EVERYBODY IS ONLY ALLOWED TO HOLD HANDS WITH ONE PERSON BUT THERE IS NO KUMBAYA WHERE EVERYBODY HOLDS ONE POINT. THE NATIVE PROBLEM ON THE LEFT HAND SIDE, YOU CAN SEE, HIGHER ORDER TERMS. THREE PEOPLE HOLDING HANDS. FOUR PEOPLE HOLDING HANDS AND SO FORTH. AND INSTEAD OF HAVING TO LOOK AT THE PROBLEM, WE CAN LOOK AT THE K LOCAL PROBLEM. NOW FOR A SAKE OF TIME, LET ME SHOW YOU WHAT HAPPENS. IF YOU LOOK AT THE K LOCAL PROBLEM WITH 48 VARIABLES, THIS TURNS INTO TWO LOCAL PROBLEM WITH 3, 000 VARIABLES. IF YOU LOOK AT THE ONE WITH 95, YOU GET SOMETHING WITH OVER 10, 000 VARIABLES. NEED TO SAY SOLVING SOMETHING WITH 10, 000 VARIABLES IS VERY, VERY HARD. AS A MATTER OF FACT, WHEN WE SOLVE IN 9 SECONDS WOULD OTHERWISE TAKE MORE THAN 24 HOURS. AND THIS THING HERE THAT WE SOLVE IN ROUGHLY UNDER FOUR MINUTES, NOBODY HAS BEEN ABLE TO SOLVE BEFORE. OKAY. SO WE CAN SOLVE CURRENTLY ENTRACTABLE PROBLEMS. JUST LOOKING AT THEM. IN THEIR NATIVE FORM. NOW THE FINAL THING THAT I WANT TO SHOW YOU, WHICH IS VERY EXCITING IS WORK WITH CASE WESTERN RESERVE UNIVERSITY. WHERE WE ARE ADVANCE MRI RESEARCH. THE OBJECTIVE IS SIMPLE OPTIMIZE MRI SCANS. MAKING SHORT AND BETTER. THE IDEA HERE IS VERY SIMPLE. MRI YOU HAVE ELECTRO MAGNETIC PULSES THAT EXCITE THE MOLECULES IN YOUR BODY AND HAVE CALLS THAT PICK UP THAT SIGNAL AND DETERMINE, IS IT FAT, BONE, MUSCLE AND SO ON. SO YOU HAVE A SPECIFIC PULSE SEQUENCE THAT WILL GO IN A SPECIFIC WAY BE ABLE TO DISCERN, TICKLE THIS, YES, TICKLE, TO DISCERN THE DIFFERENT TYPES OF TISSUE, OKAY. HERE THE IDEA IS NOT TO CHANGE THE SOFTWARE, NOT TO CHANGE THE HARDWARE, JUST PROGRAM THE MACHINE DIFFERENTLY AND SEE IF WE CAN DO BETTER. WE OPTIMIZE THE PULSE FUSIONS WITH THE MONTE CARLO, DO SCANS, PUT SOMEBODY IN THE TUBE. WE HOPE THIS PERSON DOESN’T GET FRIED. HASN’T HAPPENED YET. WHEN WE GET BACK, WE FEED IN. AND WE TRUST KEEP GOING LIKE THIS. LET ME SHOW YOU SOME RESULTS. WHAT YOU SEE HERE IS TWO SCANS. ON THE LEFT INSIDE, YOU HAVE THE QIO INSPIRED SCAN. ON THE RIGHT HAND SIDE, POWER STATE OF THE ART. IF ANYONE CAN TELL ME HERE THAT THEY ARE DIFFERENT, I WON’T BELIEVE IT. THEY ARE VERY MUCH COMPARABLE. HOWEVER THE STUFF ON THE LEFT IS A PULL SEQUENCE THAT NOTHING ANYBODY HAS BEEN ABLE TO ENVISION BEFORE. THAT IS HUGE. I CAN UNLOCK SHORTER SCAN TIMES, BETTER QUALITY IMAGINING, YOU NAME IT. AND YOU MIGHT SAY WHY ARE WE DOING THIS? WHY EMULATING QUANTUM PROCESSES ON CLASSICAL HARDWARE? WHY? NUMBER ONE BECAUSE WE WANT TO DELIVER QUANTUM SOLUTIONS TODAY. NUMBER TWO, WE CAN JUST CROSS OUT INSPIRED. SOMETHING HAPPENED WITH POWERPOINT HERE. OH. SO WHAT IF WE HAVE QUANTUM HARDWARE? THE ALGORITHMS THAT WE USE IN PHYSICS HAVE TYPICALLY A RANDOM WALK OF THE CORE. IF WE REPLAY THE CLASSICAL RANDOM WALK BY A QUANTUM WALK AND IMPLEMENT QUANTUM HARDWARE, WE GET SPEED UP OVER CLASSICAL HARDWARE. IN OTHER WORDS, ANY ALGORITHMS WE RUN TODAY, WE CAN RUN TOMORROW FASTER ON QUAN YUM HARDWARE. THIS I LIKE TO THANK YOU. AS I SAID, ALWAYS THINK ALGORITHMS FIRST. DOWNLOAD Q SHARP. INVENT NEW QUANTUM ALGORITHMS. THAT IS WHERE THE BIG IMPROVEMENTS WILL BE. ENGAGE IN THE QDK. ENGAGE IN THE QUANTUM NETWORK. LEARN MORE ABOUT QUANTUM. IF YOU HAVE QUESTION, SHOOT ME A MESSAGE. THANK YOU. [