So apparently Elon was against having radar/LIDAR on the cars:
![]()
Elon Musk's crusade against LiDAR and radar is taking its toll on Israel's Arbe and Innoviz | CTech
Tesla has stopped selling vehicles with electronic radar in the US, and Musk has been refusing to use laser radar for years. Musk's vision undermines the existence of two competing Israeli laser radar suppliers, which merged last year with American SPACs, and have since seen their value plummetwww.calcalistech.com
Now he's backtracking, what a dumbass:
![]()
Tesla appears to be turning back to radar for its vehicles
Tesla plans to add a radar product to its vehicles in mid-January, according to documents posted with the Federal Communications Commission.techcrunch.com
He gotta go.a lot of teslas have radar hardware already. he only started removing it from some models in 2021 IIRC. now they are reversing that.
he went all-in with vision only AI and tried to save costs![]()
Tesla has to disable that shyt completelyThey need to do a nationwide recall.
Radar is more reliable.. mans tried it with AI and that shyt out here straight killing folka lot of teslas have radar hardware already. he only started removing it from some models in 2021 IIRC. now they are reversing that.
he went all-in with vision only AI and tried to save costs![]()
Radar is more reliable..
mans tried it with AI and that shyt out here straight killing folk
But why can’t they have the AI working in tandem with radar?.. for some things .. obstacle and distance detection
you will always need the AI, else how is the car going to interpret its environment?
you have lidar/ridar running as a collision/obstacle warning sub-system where AI executive combines that with visual modelling.
priority being avoid colliding with dangerous obstacles (including those behind/beside you).
But why can’t they have the AI working in tandem with radar?
Aircraft today have those exact systems that you are talking about..what i said is in tandem. it is just a question of which information you give priority. you don't however mix two/n sets of raw data. you act on n/sets of interpreted data as the raw data is each sub-system is very different.
at the end of the day you have to deal with conflicting information and have a decision matrix/model based on that.
in most cases for possible collisions/obstacles you are going to trust radar/lidar. but in any real world example you are going to have to deal with faulty data, faulty devices so mitigation actions will have to be layered, with normal state recoverable and extreme action (full brake) the last option.
the more i think about it the more device redundancy, vehicle-2-vehicle data sharing etc will play a role in driving accidents down past a certain threshold.
or put another way, how much can you trust a single device? in computing systems you might build in some self-diagnotics redundancy (checksums. alive-pings, predicatble data-stream) you might need the same out there in the real world i.e. obstacles of known distances (or in practice delta differences between two objects) to allow for automatic device checks.