HOW MAJOR TECH INDUSTRY LEADER TESLA IS INVESTING HUGE IN THE AI SECTOR

ALMOST ANY SCIENCE FICTION MOVIE RELEASED IN THE LAST 2 DECADES HAS ITS PLOT REVOLVING AROUND ARTIFICIAL INTELLIGENCE. IN THE NEXT 2 DECADES WE MIGHT SEE THESE PLOTS TURNING INTO REALITY, MOSTLY IN GOOD SENSE. ALL CREDITS GO TO THE HUGE TECH GIANTS THAT ARE INVESTING HUGE INTO RESEARCH ON AI ,NEW NEURAL NETWORK MODELS , GPUs ,DATA ANALYSIS AND PRETTY MUCH EVERYTHING THAT RELATES TO AI.

BE IT TESLA ,GOOGLE ,MICROSOFT OR GAMING INDUSTRY GIANT NVIDIA ,THE LAST DECADE HAS SEEN A BOOM IN AI RELATED TECHNOLOGY BEING USED AND DEVELOPED. TESLA, RUN BY ELON MUSK, ONE OF THE MOST DYNAMIC PERSONALITIES ON EARTH TODAY, HAS REDEFINED AUTOMATION IN AUTOMOBILE INDUSTRY. ALL THANKS TO HIS EFFORTS AND NEURAL NETWORKS!! LET’S DISCUSS TESLA’S BIGGEST AUTOMATION ACHIEVEMENT:

TESLA AUTOPILOT

TESLA AUTOPILOT , WHICH COMES EMBEDDED WITH THEIR CARS IS A SOFTWARE WHICH INTENDS TO MAKE THE DRIVING EXPERIENCE OF THEIR CUSTOMERS EASY. THE SOFTWARE IS A DEEP LEARNING MODEL (A NEURAL NETWORK) .

TESLA TRAINED A MODEL CONSISTING OF 48 NEURAL NETWORKS WHICH TOOK ALMOST 70,000 GPU HOURS TO PREDICT 1000 DISTINCT TENSOR OUTPUTS IN EACH TIME STEP. YOU CAN CHECK OUT THIS BLOG FOR A BETTER UNDERSTANDING OF NEURAL NETWORKS: AS COMPLEX AS IT SOUNDS ,ONE NEEDS TO UNDERSTAND THAT “AUTOPILOT” DOESN’T MEAN THAT IT WILL MAKE YOUR CAR COMPLETELY AUTONOMOUS AND YOU CAN LAY ON THE BACKSEAT OF YOUR CAR AND WATCH NETFLIX. THE SOLE PURPOSE OF TESLA AUTOPILOT IS TO MAKE DRIVING SAFER AND LESS STRESSFUL .EVEN THOUGH IT HAS MADE MANY ACTIONS AUTOMATED ,THE DRIVER HAS TO STAY VIGILANT. LET’S SEE WHAT THIS DEEP LEARNING HAS TO OFFER

TRAFFIC AWAY CRUISE CONTROL

THIS PACKAGE HELPS MATCH THE SPEED OF YOUR CAR TO THAT OF THE SURROUNDING TRAFFIC. IT DOES SO BY TAKING INPUTS FROM ITS 8 SURROUND CAMERAS WHICH PROVIDE 360 DEGREE VISIBILITY .THE ICING ON THE CAKE ARE 12 UPDATED ULTRASONIC SENSORS WHICH ALLOW DETECTION OF BOTH HARD AND SOFT OBJECTS AT NEARLY TWICE THE DISTANCE FROM THE FORMER TECH.

AUTO STEER

THIS PACKAGE USES TRAFFIC AWAY CRUISE CONTROL (TACC) AND ALLOWS ASSISTS IN STEERING WITHIN A CLEARLY MARKED LANE. REMEMBER ,ITS AN AI MODEL TO “ASSIST’ THE DRIVER .THE DRIVER IS EXPECTED TO STAY VIGILANT AND IS ADVISED NOT TO TAKE HIS HANDS OFF THE STEERING WHEEL. HOW DOES IT DO SO? THIS PACKAGE USES A SENSOR WHICH MEASURES THE TORQUE THAT YOU APPLY TO THE STEERING WHEEL , AND IF IT FALLS BELOW THE REQUIRED VALUE FOR THE TURN TO TAKE PLACE ,IT GIVES VISUAL AND AUDIO ALERTS TO THE DRIVER .

AUTO LANE CHANGE

THIS AI PACKAGE ASSISTS THE DRIVER IN TRANSITIONING TO AN ADJACENT LANE . ONE SHOULD REMEMBER THAT THIS FEATURE CAN BE USED ONLY ON HIGHWAYS , THE VIGILANCE OF THE DRIVER IS ADVISED .

AUTOPARK ,SUMMON ,SMART SUMMON

WHETHER YOU ARE A NEWBIE DRIVER OR AN EXPERIENCED ONE ,PARKING IS SOMETHING WHERE EVERYONE STRUGGLES . “AUTOPARK” HELPS TO PARK YOUR CAR WITH A SINGLE TOUCH ,PARALLEL OR PERPENDICULAR WAYS .

SUMMON /SMART SUMMON HELPS TO MOVE YOUR CAR IN A TIGHT SPACE USING A MOBILE APP OR A KEY , IT HELPS NAVIGATE IN A MORE COMPLEX ENVIRONMENT.

BUT OUR CONCERN IS NOT ONLY LIMITED TO UNDERSTAND WHAT DOES IT DO, RATHER TO UNDERSTAND HOW DOES IT DO THAT . SO WE SEE AN AI BASED SOFTWARE ACCOMPLISHES THE TASK OF TRAINING ITSELF TO DRIVE, SOMETHING WHICH EVEN HUMAN BRAIN MAKES MISTAKES IN. ITS A WELL KNOWN FACT THAT THE PERFORMANCE OF A NEURAL NETWORK DEPENDS ON THE AMOUNT OF DATA AVAILABLE TO TRAIN THE MODEL( CLICK TO KNOW MORE). SO TECHNICALLY DATA IS THE KEY INGREDIENT IN TRAINING MODELS. TESLA GETS ITS DATA FROM ITS FLEET OF NEARLY 5,00,000 VEHICLES ON THE ROAD EMBEDDED WITH SENSORS AND HARD WARES. TESLA’S CARS ARE DRIVING ABOUT 1.5 MILLION MILES DAILY ! AND ALSO THE NUMBER OF CARS IS INCREASING BY NEARLY 5000 CARS PER WEEK. SO OVERALL THIS GENERATES TONS OF DATA TO BEGIN WITH.

IS ALL DATA THE SAME?

YOU CAN DIVIDE THE OBJECTS PRESENT ON THE ROADS INTO VARIOUS CATEGORIES ,SOME ARE RARE, FOR EXAMPLE ENCOUNTERING A DEER ,SOME ARE TOO COMMON ,LIKE JUST ANOTHER CAR, THE NEURAL NETWORK SHOULD ADDRESS THIS. USING THE COMPUTER VISION TESLA’S AI CAPTURES PICS OF RARE APPEARANCES AND WILL LATER UPLOAD IT OVER WIFI. AND CREDITS TO THE LARGE DATA SET YOU CAN EXPECT TO HAVE MANY OCCURRENCES OF THE “RARE” OBJECT . NOW FOR OBJECTS THAT ARE COMMONLY OCCURRING TESLA’S AI : IF YOU HAVE TRAINED NEURAL NETWORKS YOU MUST HAVE COME ACROSS THREE FRAMEWORKS TO DO THE SAME . PYTORCH , TENSORFLOW AND KERAS . TENSORFLOW IS MORE FAMOUS AND IS WIDELY USED BY THE COMMUNITY OF ENGINEERS . BUT FOR NEW RESEARCHES AND DEVELOPMENT ,PYTORCH IS THE IDEAL CHOICE . IT ALLOWS RESEARCHERS TO CODE THE NETWORK FROM SCRATCH IN CONTRAST TO USING SOME PRE DEFINED LIBRARIES . WATCH THIS VIDEO TO HAVE A BETTER UNDERSTANDING ABOUT THE MODEL FRAMEWORK .

THE FINAL WORD

ALMOST EVERY INDUSTRY IS TRYING TO INCORPORATE AND UTILIZE THE ADVANCEMENTS OF THE ARTIFICIAL INTELLIGENCE TECHNOLOGIES . WHAT SEPARATES THESE KIND OF AUTOPILOT AUTONOMOUS CARS FROM DESIGNING AN AI BASED VIDEO GAME IS THAT A MISTAKE MIGHT NOT BE AS SIMPLE AS A “GAME OVER ” BEING DISPLAYED ON THE SCREEN . PEOPLE ARE SITTING IN THE CAR . ALSO TESLA AUTOPILOT GUIDELINES CLEARLY SPECIFIES THAT THE DRIVER MUST STAY VIGILENT AND THAT THE AUTOPILOT DOES NOT BY ITSLELF GIVES THE DRIVER THE ABILITY TO CLOSE HIS EYES AND RELAX .

RECENTLY AN ACCIDENT WAS REPORTED WHEN A TESLA ON AUTO PILOT RAN STRAIGHT INTO A TRUCK LAYING ON THE ROAD. THERE WERE MANY SPECULATIONS REGARDING THE REASON BEHIND IT . A POSSIBLE FEW THINGS ,RATHER QUESTIONS WERE RAISED:

1. WHAT DOES TESLA AUTOPILOT DOES WHEN IT LOOKS AT A SUPER RARE INSTANCE ? SURELY A TRUCK LAYING ON A ROAD IS NOT A SIGHT YOU SEE EVERYDAY .
2. DID THE AUTOPILOT “UNSEE” THE TRUCK BECAUSE IT WAS WHITE ? WELL NO BECAUSE TESLA USES A RADAR SYSTEM WHICH HAS NOTHING TO DO WITH COLOURS .
3. DID TESLA AUTOPILOT CONFUSE IT WITH SOMETHING ELSE THAT WAS A REGULAR INSTANCE ? LIKE AN OVERHEAD BANNER OR BRIDGE? BECAUSE THE CAR DID NOT SLOW DOWN AT ALL BEFORE HITTING THE TRUCK !!!