4 ms·
I don't know why they don't parametrize momentum with certainty. In any confusing situation, go into ultra slow environment scanning and when confidence increas
by agumonkey 5y ago
I don't know why they don't parametrize momentum with certainty. In any confusing situation, go into ultra slow environment scanning and when confidence increases, allow for a bit more.. rinse / repeat.
- erulabs 5y agoTesla does exactly this and it gives rise to the phantom breaking problem. Still seems like a good solution for a small bot with no passengers
- KennyBlanken 5y agoNo? There are numerous clips where Telsas in "full self driving" mode pull the equivalent maneuver of a teenager going "OH SHIT I WANT TO GO THERE" and veering very violently. The phantom braking problem is likely just one of the many symptoms of Musk's insistence on relying on optical systems instead of more expensive sensors.
- outworlder 5y ago> The phantom braking problem is likely just one of the many symptoms of Musk's insistence on relying on optical systems instead of more expensive sensors. Based on what? How would 'expensive' sensors help?
- bluGill 5y agoWe know that in some situations expensive sensors can get data that optical cannot. What we don't know is if any of the above is enough extra data. What we do know is there are times when humans are bad drivers, and other times when humans continue when they shouldn't relying on luck. (Ie driving in snow storms with low visibility)
- Spooky23 5y agoConsider your human eyeball. It works really well, but you can get into trouble in certain light or visual conditions. Technology exists (radar, lidar) that doesn’t have the same limitations as visual. (They have their own issues of course)
- gzer0 5y agoExpense was part of the equation initially, however, through economies of scale, we eventually would have been able to reach a feasible price point. Cost has nothing to do with why Tesla is pursuing an optical-only system. To get rid of the dependency on the radar sensor for autopilot, we generated over 10 billion labels across two and a half million clips. To do this we had to scale our offline neural networks and our simulation engine across 1000s of GPUs, and just a little bit shy of 20,000 CPU cores. We also included over 2000 actual autopilot full self driving computers in the loop with our simulation engine. And that's the smallest compute cluster.
- stefan_ 5y agoThat must be why complaints about phantom braking have gone through the roof since the switch away from radar.
- meheleventyone 5y agoThose are very large numbers for something that doesn’t work very well.
- atleta 5y agoSo what's the point then? You said it's not expenses and then you explain how you think it caused you extra trouble/work/development effort. But what's the reason?
- chillingeffect 5y agoAll I can say without breaking agreements is that these are products, not ideal models of conceptual engineering. They're not created by people who like the world and want it to be a better place. They're created by people with lots of money who want a lot more. They've found an avenue for this by persuading other wealthy, greedy people to give them a lot of money and promising they can give them more back. They'll do this by persuading everyday people to not do things like produce, prepare, and transfer food themselves and instead pay money for these robots to do it. These robots are minimum viable products toward moving capital around, not meeting user requirements or demonstrating great ideas. Hurting a few people in the process is part of the equation. Getting anyone to care about $cool_algorithm is not part of the equation. Getting people addicted to the convenience is part of the equation. Getting things to market as blindingly fast as possible so the capital moves before feedback from the field arrives is paramount.
- jdlshore 5y agoThat's an unnecessarily cynical generalization. Sure, maybe the leaders of the companies creating these things are profit-motivated, but is that really true of the individual engineers and designers who created it?
- jeffreygoesto 5y agoNo, that is a very accurate description. The engineers willing to work on those things and suppressing deeper thoughts for the money and kick off new tech are part of the equation and the problem. A manager I had once had a postcard in his office "The engineer is the camel on whos back the merchand rides to his success." You are a lever and even provide the excuse for being one yourself.
- sizzle 5y agoBoth of what you said can be true at the same time (not mutually exclusive of each other) while OP’s assertions may still be true for certain individuals if I’m thinking logically. We are talking about what motivates humans as human behavior, which tends to be varied, nuanced, and hard to reduce to mutually exclusive categories like being only profit driven or only driven by intellectual curiosity. I think you can be both motivated by money and intellectual curiosity. If you are an engineer turned founder, you can be both? Someone correct me if I’m wrong here.
- numpad0 5y agoThat's how to get a robot half feet into a choke point, immediately get stuck for half an hour surrounded by walls and confused people, until developer on an emergency Slack call along facility managers and company CTO verifies and communicates a likely-safe state of robot and surrounding equipment to field operators and a go is given to pull the thing out of the elevator.