Search results

I don't think there's any problem with asking for evidence to support claims. Not everyone has the same information, and sourcing claims puts everyone on the same page. It isn't fair or reasonable to expect people to automatically accept as fact anything they read on a forum. That's obviously...

So turn signal detection only works/worked in Germany and you can only find mention of it in German owner manuals? How is it common knowledge then? Maybe it is to German drivers, but most TMC members are from outside Germany. Mobileye’s English-language documentation doesn’t make it easy to...

Right now it feels to me like supervised learning is hard to beat for perception, but long-term it would be ideal to get rid of supervised learning because human data labelling bottlenecks progress. If learning of both perception and action can occur without human labelling or any supervisory...

Pre-AlexNet I’m skeptical how accurately a system could classify turn signals. The owner’s manual for the 2012 Audi A8 doesn’t mention anything about being able to detect turn signals. Instead, it says on page 89 that it will match a lead vehicle’s speed until that vehicle is clear: “When...

Hard to find any mention of production cars detecting and responding to turn signals this early on. Can you provide a source? What’s Mobileye’s detection error rate, and what’s Tesla’s?

Are you talking about a feature that is in production and used by customers today, or a prototype that has never been deployed commercially? Is there any non-Tesla production vehicle on the road today, driven by regular customers, that currently detects turn signals and responds by slowing down...

Someone posted a video to Reddit purporting to show Autopilot recognizing another car’s turn signal and slowing down to let the car change lanes: A video proof AP now give right of the way to cars with merging signals after 2019.5.15 : teslamotors

This is apparently new in the 2019.8.3 firmware: Autosteer Stop Light Warning Your car may warn you in some cases if it detects that you are about to run a red light while Autosteer is in use. This is not a substitute for an attentive driver and will not stop the car. Photo by u/IV-Crush...

Another video. (Thanks @diplomat33)

Paywalled, but in the preview you can see that Waymo riders made a complaint through the mobile app for 40% of rides: With Waymo Robotaxis, Customer Satisfaction Is Far From Guaranteed The reporter shared an example on Twitter: Amir Efrati on Twitter (I recommend following Amir Efrati by the...

It’s comically slow... but better safe than sorry. :) Hope to see steady improvement as they keep developing and testing it.

Apologies if these were already posted somewhere. First videos I’ve seen of Enhanced Summon. (Thanks to @TheTeslaShow for sharing.) To quote @TheTeslaShow: “Seems only useful for your own house and or empty late night lots or super low traffic office lots. But with iteration and more use...

Navigant’s report costs $4,000, so unless somebody pays that, we won’t know exactly what this ranking is based on. It’s unclear to me ATM whether Navigant has any AI/robotics expertise in-house or interviewed AI/robotics experts for this report. If not, then why should Navigant’s ranking matter...

Okay, this is a helpful clarification. I think this gets back to the recurring debate on this forum about demos — whether we should treat a demo as meaningful evidence of real technological progress, or whether we should doubt that a demo tells us anything meaningful about the commercial...

On reinforcement learning: It seems to me that the essential difference between applying RL to driving versus other activities like Dota is the challenge of representing the driving task in simulation. How do you make sure what the agent learns is driving, and not a video game that closely...

Some common mistakes in analysis (a few of which I’ve made): Conflating performance in a demo with a system’s reliability over millions of miles. Taking California DMV disengagements data at face value (a mistake I made too). Making apples-to-oranges comparisons between secretive prototype...

Autonomous driving can be split into two kinds of task: Perception: object detection, depth mapping, semantic segmentation of driveable roadway, etc. In short, computer vision tasks that are done with supervised learning. (When people say “deep learning”, they’re typically referring to deep...

I really encourage people to watch the video in the OP. The contrast between end-to-end learning, the classical architecture, and mid-to-mid learning of path planning is super helpful for understanding the autonomous vehicle problem space. Waymo, Tesla, and Mobileye all seem to be developing...

Falcon Heavy was, what, 5 years late? A lot of Elon's SpaceX predictions have been 1-5 years late. (Source: Bloomberg.) I think some people interpet Elon's comments in an extremely uncharitable way. No one can reliably predict when these things will get done. It's not just Tesla and SpaceX —...

My feeling is that if the solution to driving policy and path planning turns out to be purely reinforcement learning (RL) in simulation, then Waymo stands the best chance of solving autonomy first. Alphabet has DeepMind and Google Brain — a deep bench of RL researchers who are arguably the best...

Isn’t Alphabet (Google, DeepMind, Waymo) presumably better at deep RL than anyone else in the world? Including Mobileye? Do you think Mobileye’s putative head start is enough to overcome Alphabet’s advantage in deep RL? Or do you think Alphabet just has no advantage over Mobileye in deep RL...

Great answers @heltok thanks so much! I particularly see the value in having HD maps if the sensors fail and the car needs to make an emergency stop.

Where is the source of ground truth redundancy in HD maps if they are compiled using the same neural network that does perception in real time? Presumably it would be from statistics: if the last 1,000 times cars have passed through an intersection they have detected a stop sign, and now a car...

Bladerskb, is your belief that Mobileye will solve full autonomy with a little bit of imitation learning (under 10 million miles of manual driving data) and mostly reinforcement learning in simulation? If that is your belief, do you think Mobileye will solve full autonomy before Waymo? If you...

Very cool! In addition to TomTom and HERE, I know there are a bunch of startups working on HD maps: e.g. DeepMap, Carmera, Mapper, Civil Maps, Mapbox, Nomoko, and lvl5. I would guess a company like Tesla would probably want to develop its own HD mapping solution in-house to eventually leverage...

That is a good point, although automated mapping from production cars isn’t the only way to acquire an HD map of U.S. freeways that includes barriers. Do you have findings from your own hacking you would like to share? What is the source of your knowledge about Tesla’s maps?

Hmm... That’s interesting. Do you mind elaborating?

I’m assuming these are 3D maps (or 3D models) that include lane lines and traffic barriers/road dividers? Watched a few minutes starting at 22:00. Didn’t hear where Amnon says all trajectories driven by all production cars are uploaded. Do you have an exact timecode, down to the second? I...

An interesting idea from a Mobileye talk on reinforcement learning is an options graph: “Semantic Abtraction: we decompose the diving policy function into semantically meaningful components using an options graph” “Decomposing the problem into this graph helps us to learn every component of...

@heltok posted this awesome video in another thread (great find heltok!). It’s a talk by a robotics professor at Carnegie Mellon who worked on autonomous cars at Uber ATG. To me, the most interesting theme of his talk was the contrast between two self-driving car architectures. First, there’s...

Wait, so is Mobileye and/or BMW collecting state-action pairs from production cars, or do we not know what data is being collected from production cars? None of the talks, papers, or patents I have seen so far specify what data is collected from production cars. "A lot of data" doesn't...

Crucially, Karpathy’s claim is not a prediction about something in the future. It’s a statement about the current performance of the new, as yet undeployed neural networks. I presume that these NNs were tested on Tesla’s test dataset, and showed better accuracy than the NNs currently running in...

Fair. I believe Karpathy is telling the truth, and if I'm wrong then I won't trust him in the future. It is easy to believe what he's saying because it doesn't seem to be a controversial idea in deep supervised learning that a ~10x more computationally intensive neural network would perform better.

Meant to say: conjecture about things that don't exist today - - - - - - - - - - - Let me try to summarize the discussion so far. I put forward an idea that I'll call the imitation learning thesis. The imitation learning thesis says: If/when Tesla solves perception (which will require, at a...

I mean this is just a general assertion. Not something that can be specifically addressed with evidence or reasoning. ¯\_(ツ)_/¯ I like it better when people make specific claims I can respond to. If we were to take this to its logical conclusion, wouldn't we conclude that fully autonomous...

Thank you! That’s the main point! I believe verygreen has said there were completely random triggers at one time. These would collect uninteresting data. Maybe jimmy_d could comment on this, but AFAIK you don’t want the training dataset to be completely statistically representative of real...

Awesome! This is really important info! This would seem to indicate that Tesla collected (in some unknown number of snapshots) the requisite data to do imitation learning.

I found an old Tesla job posting for “Simulation Engineer, Autopilot”. (archive.org | archive.is) The job description says a few things to make it clear this is a full self-driving simulator: “As an Autopilot Simulation Engineer, you will contribute to the development of a fully-autonomous...

Thanks! Any camera data? Video or still images? If the neural network’s mid-level representation (i.e. what’s represented in your videos by bounding boxes, “green carpet”, etc.) were uploaded with snapshots, would you be able to tell?

Okay, very intriguing, thanks. Just to be clear, what's included with a snapshot? I know you've detailed this elsewhere, but I forget. It includes camera and radar data too, right?

HD map data is different than state-action pairs. HD maps only include fixed features of the environment (roads, signs, lights, lane lines), not road users (cars, bikes, pedestrians). HD map data also doesn't include driver input (steering, braking, accelerating). It would be incredibly...

I think navigation and driving are different problems. I believe the navigation and route is handled by a traditional GPS navigation system like Google Maps or TomTom or whatever. I believe supervised imitation learning only comes in at the level of discrete driving tasks, like taking a right...

This is interesting. Have you noticed any patterns in when Tesla uploads steering+pedal data, or what other data it uploads alongside it?

Tesla hadn't even started working on a simulator? That's certainly not what Amir said. He wrote: "Under Mr. Bowers, the simulation and maps teams are both still in their infancy, said a person familiar with the situation. But simulation already has proven to be valuable in helping the Autopilot...

I don't read everything people post, and I don't respond to everything I read. The more respectful you are, the more likely I am to respond. The more disrespectful, the less likely. Human labelling is by far the most expensive part of the equation for the training of a perception neural...

If anyone has any technical criticisms of my conjecture, I’m all ears... (But ad hominem remarks and accusations of bad faith are not welcome)

Getting back to the topic of the OP, I think one of the most fascinating pieces of information to come out about autonomy lately is that Waymo is using imitation learning. Imitation learning, for those who don't know, means a neural network learns to map certain kinds of actions to certain...

So, is your prediction that no HW2 cars will get a free upgrade to the HW3 computer even if the owner purchased Full Self-Driving? Or that only HW2.5 cars will get it, not HW2.0? This is mis-transcribed. These call transcripts are always littered with errors. What Elon actually said (around...

Many disengagements — perhaps even the vast majority — don’t have to be reported. I don’t even look at the disengagement figures anymore. They might be interesting from a public safety point of view (which is what the DMV cares about), but they don’t seem interesting from a technology progress...

Holy cow! This is awesome. What about using radar to measure distance to nearby vehicles? What’s the best result you’ve seen on that?