Raw LLM Responses

I'm a computer scientist in the UK, with a specialty in Artificial Intelligence. My view on this (and im currently in the process of making a documentary on Autonomous Vehicles, feel free to check out the progress so far on my profile) , the documentaries try to remain impartial and factual, however that doesn't mean I dont have my own opinion on the research and thoughts based on the technology. A robot in computer science is often defined as an unit that is aware of its environment and able to interact with it / navigate around it whilst providing some useful functionality A robot vacuum cleaner for instance is a unit that is the vacuum cleaner and able to navigate around a room and know were it is, in relation to the dimensions of the room with a useful functionality to vacuum a floor and in some models even mop it up When you look at these type of vacuum cleaners there are VSLAM (Visual Simultaneous Localization and Mapping) and LiDAR (Light Detection And Ranging) and this is true to self driving cars were by Tesla and Wayve use cameras (VSLAM) and Waymo, Cruise and TuSimple have used LiDAR, the best of cause are the ones that use a combination of both along side other sensors for redundancy. This is why Waymo and Cruise achieved Level 4 Autonomous driving whilst Tesla remains at Level 2 possibly Level 3. The same problems you see in robotic vacuum cleaners are present in robotic cars that use VSLAM Poor performance in low lighting conditions (night time, light flares, obscure point of references) and whilst robotic vacuum cleaners may look at your ceiling for a point of reference or detect objects infront of it, with a AV then the point of reference may be the road markings, and a country road or low lit road makes it difficult even for human drivers let alone a camera LiDAR isnt without flaws, it too can be fooled with metalic objects and transparent or translucent objects, this is why a combination of sensors are necessary so that were one misses the object the other kicks in and states so Mark Rober did a video recently showing how to fool an AV, crashing into a fake wall depicting an image of a road, LiDAR often stops, because the Laser beam bounces off an opaque object to notify it that something is there, but equally LiDAR cannot read a road sign as its only job is to detect solid opaque objects and how far that object is from its-self. Camera as the sole backup may not necessarely work in 100% of cases either, take for instance a transparent object, both LiDAR and Cameras would fail to recognise there is something between the robot and its destination, this is why other technologies such as RaDAR can assist However people equally must accept that Self Driving cars is not intented to make road traffic accidents a historical thing, it is intented to reduce the volume of serious collisions, which it is doing these stories shown to the world in WSJ are of cause worst case scenarios, and one could say well accidents have and do continue to take place on commercial airliners and private planes yet we still use them to travel and even say Air travel is the safest form of travel, so id be enclined to disagree with Missy Cummings at the end of the video, I do agree that LiDAR is expensive. Problems with Tesla, is too many use it as self driving level 4-5 vehicles when it is not, and I agree with Missy Cummings in that people will have a false sense of security in that they trust the technology too much, very similar to how airline pilots trust their computer readings even when sensors as provding misleading data leading to a crash.