HOW NVIDIA PREDICTED TEN YEARS BACK THE FUTURE OF AI AND HOW ITS GPUs ARE PLAYING A KEY ROLE IN DEVELOPMENT OF DEEP LEARNING IN GAMES .

YOU MUST BE FAMILIAR WITH CPU( CENTRAL PROCESSING UNIT) . AND IF YOU ARE EITHER A GAMING ENTHUSIAST OR SOMEONE WHO KEEPS HIMSELF UPDATED WITH AI ADVANCEMENTS CHANCES ARE HIGH THAT YOU HAVE HEARD OF GPU’S. THEY HAVE EMERGED AS A BOON FOR AI DEVELOPMENT . AND SURELY THEY ARE DOING THEIR PART IN GAMING /GRAPHICS ADVANCEMENTS. WITH THE INCREASING USAGE OF DEEP LEARNING IN GAMES ,WE MIGHT SEE REALLY DIFFERENT VERSIONS OF GAMES BY THE END OF THIS DECADE . SMART AND INTELLIGENT ONES !!!!

LETS ANSWER THE BASIC QUESTION FIRST

WHAT IS A GPU

GPU STANDS FOR GRAPHICS PROCESSING UNIT .THE CORE PRINCIPLE ON WHICH GPU WORKS IS ” DIVIDE AND CONQUER ” . IN TECHNICAL TERMS WE CALL IT PARALLEL COMPUTING . THIS IS WHAT MAKES A GPU SO GOOD IN WHAT IT DOES . THIS IS IN CONTRAST TO A CPU WHICH CAN DO ONLY A HANDFUL OF CALCULATIONS AT ONCE . BUT HOW DOES A GPU DOES SO .WELL , THE ANSWER SEEMS OBVIOUS!!

NUMBER OF CORES IN A GPU

IF YOU ARE READING THIS ON A LAPTOP , OR IF YOU OWN ONE , YOU MUST HAVE HEARD TERMS LIKE DUAL CORE ,QUADCORE AND OCTACORE . SO BASICALLY A CPU CONSISTS OF A FEW CORES WHICH CAN HANDLE ACTIVITIES AND SUPPORT MULTIPROCESSING . IN CONTRAST A GPU CONSISTS OF A HUNDREDS OF CORES WHICH ALLOWS IT TO PERFORM PARALLEL COMPUTING BY BREAKING THE PROBLEM INTO SUBPROBLEMS.

SO HOW DOES THIS HELP IN TRAINING AI / DEEP LEARNING MODELS ?

I HOPE YOU ARE FAMILIAR WITH NEURAL NETWORKS. AND IF NOT ,SUGGEST READING THIS AND THEN COMING BACK TO UNDERSTAND BETTER . ITS EASY TO SEE HOW THE NUMBER OF CALCULATIONS NEEDED PER CYCLE OF BACKPROPOGATION /FEEDFORWARD INCREASES DRASTICALLY WITH THE INCREASING COMPLEXITY OF THE NETWORK. ( MORE NEURONS , HIDDEN LAYERS AND SO) . THE NUMBER OF CALCULATIONS SOON RISE TO MILLIONS . TRAINING SUCH A MODEL ON A CPU WOULD BE A TEST OF PATIENCE . GPUs , BECAUSE OF THEIR ABILITY TO BREAK UP THE PROBLEMS USES PARALLEL COMPUTING TO ITS AID WHICH SPEEDS UP THE PROCESS EXPONENTIALLY .

GPUs IN GAMING , COMPUTER VISION ,SUPERCOMPUTING

I THINK BY NOW YOU CAN REALISE HOW GPUs WOULD OUTPERFORM CPUs IN THESE APPLICATIONS . SURELY BY THE END OF NEXT DECADE THE AI WOULD EFFECT GAMES , HEALTHCARE AND MANY DOMAINS OF COMPUTER SCIENCE . CURRENTLY NVIDIA IS LEADING THE GPU PRODUCTION MARKET . NVIDIA GRAPHICS CARDS ARE FAIRLY POPULAR AMONG GAMERS . CURRENTLY THE GAMES THAT CLAIM TO USE ” AI ” ARE IN REALITY USING ALGORITHMS FOR OPTIMISING THE ENEMIES BEHAVIOUR /APPROACH . THEY WORK ON TECHNOLOGIES LIKE “FINITE STATE MACHINES” . BUT SOON WE EXPECT THINGS TO CHANGE . TILL THEN , GAME ON!!!!