Bet TSLA
Active Member
As is usual, almost all of this radar / no radar chitchat is speculation based on ignorance. Whether radar makes sense as part of Tesla's sensorium is something we don't and cannot know.
Back to first principles. Adding a new sense costs money, design time, supply chain complexity, manufacturing complexity, software time and complexity, training time and complexity, energy for operation and due to weight, and it adds new failure modes and maintenance costs. A new sense can improve safety, performance, behavior, avoid damage, and make the vehicle usable in more conditions.
Only Tesla has the data to even begin to guess what the trade-offs are. Only Tesla has the data to determine what senses would provide the best fit with current and future capabilities. Tesla is clearly not going to create the safest car possible no matter what the expense. It's entirely possible that radar's marginal utility at this point is simply no longer a win over its many negative aspects.
I didn't see anybody mention the critical superpowers the vehicles already possess: patience and tireless attention. Most critically, this means that a fully autonomous Tesla will easily and naturally never drive in a way it considers too unsafe. If it can't see the road well enough it will execute a safe alternative to proceeding. Because this is comparatively easy to do, Tesla can choose to failsafe in a variety of sufficiently rare situations without compromising utility. So even if removing radar eliminates a capability, only Tesla can tell whether it matters much.
All we can do is guess. We don't have any of the data required to tell pretty much anything about the optimum sensorium. So we're limited to agreeing with Elon that it has always been obvious that it must be possible to drive using vision alone, because people do so. I would really like to not see endless speculation that may as well be preceded by "But if Tesla engineers are really stupid, then...."
Back to first principles. Adding a new sense costs money, design time, supply chain complexity, manufacturing complexity, software time and complexity, training time and complexity, energy for operation and due to weight, and it adds new failure modes and maintenance costs. A new sense can improve safety, performance, behavior, avoid damage, and make the vehicle usable in more conditions.
Only Tesla has the data to even begin to guess what the trade-offs are. Only Tesla has the data to determine what senses would provide the best fit with current and future capabilities. Tesla is clearly not going to create the safest car possible no matter what the expense. It's entirely possible that radar's marginal utility at this point is simply no longer a win over its many negative aspects.
I didn't see anybody mention the critical superpowers the vehicles already possess: patience and tireless attention. Most critically, this means that a fully autonomous Tesla will easily and naturally never drive in a way it considers too unsafe. If it can't see the road well enough it will execute a safe alternative to proceeding. Because this is comparatively easy to do, Tesla can choose to failsafe in a variety of sufficiently rare situations without compromising utility. So even if removing radar eliminates a capability, only Tesla can tell whether it matters much.
All we can do is guess. We don't have any of the data required to tell pretty much anything about the optimum sensorium. So we're limited to agreeing with Elon that it has always been obvious that it must be possible to drive using vision alone, because people do so. I would really like to not see endless speculation that may as well be preceded by "But if Tesla engineers are really stupid, then...."