I'm still amazed that Teslas team isn't using a map... I know maps get outdated and are sometimes wrong, but having inaccurate knowledge of what's around the corner is far far more helpful than not having any clue whats around the corner. The smart solution would be to consider a map a probabilistic thing, which neural networks are really good at handling.
I'm still amazed Tesla has decided not to use lidar and instead just stick with cheap cameras. Better sensors are there, they're available, they're cheap and they can probably "see" better than plain old cameras... it doesn't make too much sense not to use them IMHO. But then again, I am not coding NNs for Tesla...
Andrej Karpathy talks about how Tesla's NNs are structured and trained [video]
181–190 of 249 posts
Re: Andrej Karpathy talks about how Tesla's NNs are structured and trained [video]
#182Earlier quoted context omitted.
Elon Musk poached him, and for that was kicked off the OpenAI board.
I've never heard that that was the reason for Elon leaving the OpenAI board. The official announcement said "As Tesla continues to become more focused on AI, this will eliminate a potential future conflict for Elon." Do you have a source?
Re: Andrej Karpathy talks about how Tesla's NNs are structured and trained [video]
#183Earlier quoted context omitted.
> mind boggling to think people trust this crap It's also mind boggling to think we currently trust organic tissue to do this crap, some of which is bathed in psychoactive chemicals. And yet we do, and as a result, horrendous catastrophes occur every minute of every day. > It's all fancy neural networks until someone dies No, that can't be the standard, not when people are dying right now in the current regime. Unles…
No, it needs to be a lower rate since it will kill at random. Today’s rate includes drunk drivers, people on their phone and other “unsafe” drivers. If you are an attentive driver your chance of death would actually go up if the overall death rate was the same.
Re: Andrej Karpathy talks about how Tesla's NNs are structured and trained [video]
#184Earlier quoted context omitted.
Yes it's not perfected yet, which is why it requires human supervision for now. Having it operate in the wild as it is now (again, under human supervision) will help it become less horrifying, which I think you would agree is what we want.
I don't agree, as I don't particularly value/want a future with self-driving vehicles. I also think the way Tesla is going about this is utterly idiotic and reckless. It upsets me that I'm sharing roads with vehicles having these dysfunctional immature systems. Point this crap at video game engines and don't let it anywhere near real people until it can drive millions of virtual miles in something like GTA without hi…
If bad human drivers killed only themselves, that would be one thing, plenty of Darwin awards to go around. But they don't. I do value reducing traffic fatalities to near zero. I think that would be fantastic.
The GTA idea is a good one. I think they've done that, many times over, but I could be wrong.
They way Tesla is doing it is interesting. Have human supervision, so even though the system may behave as you say like a drunk driver who lost their glasses, at least the human is there keeping it in check.
Humans aren't perfect, but I've found (and statistics have shown) that the system plus a human is better than a human alone. That alone should make you less upset. These cars are less reckless than cars driven by humans.
Human supervision is just one element of the very smart way Tesla is going about this. The other element is incremental changes. A lot can be said about that but I won't now. Teslas are getting small and big safety improvements that roll out over time as full self driving gets closer and closer. This should make you feel better.
If you see a Tesla in autopilot, it won't be tailgating. If you see people tailgating in a Tesla, you can be sure autopilot is not turned on. Even with the follow distance set to the lowest, the distance kept is larger than what is done by most humans. This should also make you less upset.
When driving in stop and go traffic with Teslas with autopilot on, if the owner sets the follow distance to say the middle range, you'll find you are often able to merge in in front of them, because they allow space for that (and for safety). This should help too.
Over time, the system is getting better and better. Eventually, it will all be ok!
Re: Andrej Karpathy talks about how Tesla's NNs are structured and trained [video]
#185Earlier quoted context omitted.
No, it needs to be a lower rate since it will kill at random. Today’s rate includes drunk drivers, people on their phone and other “unsafe” drivers. If you are an attentive driver your chance of death would actually go up if the overall death rate was the same.
if you are an attentive driver you can get killed anytime by these other drivers
The perception will be different however, since in the public eye the drunk driver “had it coming” and the only real loss is the other driver/passengers/pedestrians. If a self-driving car kills its own occupants and the occupants of another car then there are more “innocents” than in the drunk-drive scenario.
Re: Andrej Karpathy talks about how Tesla's NNs are structured and trained [video]
#186Earlier quoted context omitted.
What does Netflix do with NNs?
Their own page says, "our recommendation algorithms ... learning characteristics that make content successful ... optimize the production of original movies and TV shows ... optimize video and audio encoding, adaptive bitrate selection, and our in-house Content Delivery Network ... and advertising". https://research.netflix.com/research-area/machine-learning Here's another article on the subject: https://becominghuma…
Characteristics that make content successful.. here again, it's mostly decent quality content plus marketing. I doubt the scripts are reviewed by NNs.
I have no idea about the network and delivery stuff, but I guess a well designed network can take care of most of it.
My strong impression is that, similar to other well known cases (Uber, WeWork) Netflix is a mostly traditional company (a media company) that very strongly wants to be seen as a tech company.
Re: Andrej Karpathy talks about how Tesla's NNs are structured and trained [video]
#187I'm still amazed that Teslas team isn't using a map... I know maps get outdated and are sometimes wrong, but having inaccurate knowledge of what's around the corner is far far more helpful than not having any clue whats around the corner. The smart solution would be to consider a map a probabilistic thing, which neural networks are really good at handling.
I'm still amazed Tesla has decided not to use lidar and instead just stick with cheap cameras. Better sensors are there, they're available, they're cheap and they can probably "see" better than plain old cameras... it doesn't make too much sense not to use them IMHO. But then again, I am not coding NNs for Tesla...
LIDAR's are cheap, but not cheap enough yet to not seriously affect the bottom line if you put them into every car. It also will kill the resale price of cars without it, which in turn hurts the companies image and stock price.
Re: Andrej Karpathy talks about how Tesla's NNs are structured and trained [video]
#188Earlier quoted context omitted.
Any guess what the compensation is like for these positions ?
Less than Uber, but more than Waymo, who is only offering ~$20k-ish stock packages like a late start-up who expects to 10x. Depends on how you value Tesla stock, though. See levels.fyi
Re: Andrej Karpathy talks about how Tesla's NNs are structured and trained [video]
#189I'm still amazed that Teslas team isn't using a map... I know maps get outdated and are sometimes wrong, but having inaccurate knowledge of what's around the corner is far far more helpful than not having any clue whats around the corner. The smart solution would be to consider a map a probabilistic thing, which neural networks are really good at handling.
I'm still amazed Tesla has decided not to use lidar and instead just stick with cheap cameras. Better sensors are there, they're available, they're cheap and they can probably "see" better than plain old cameras... it doesn't make too much sense not to use them IMHO. But then again, I am not coding NNs for Tesla...
https://cleantechnica.com/2016/07/29/tesla-google-disagree-l...
Re: Andrej Karpathy talks about how Tesla's NNs are structured and trained [video]
#190Fun fact for all of you: Some time ago (around ~10 years) this guy (the presenter) was internet famous for being a Rubik cube speed solver and making tutorials and videos about that: https://www.youtube.com/watch?v=609nhVzg-5Q