This is something that happened to me just the other day. I was crossing a one way street in midtown Manhattan. The street is one way, and has two parking lanes, one on each side of the street, and a travel lane in the middle. The street was packed with stopped cars, and I was crossing it midblock between a stopped truck and a stopped car in front of it. I started crossing, and as I approached the end of the first parking lane I made eye contact with the truck driver before I walked in front of the truck. Just as I was almost approaching end end of the truck, I caught a reflection in the a parked car of a bicyclist riding between the truck and the parked car, and I abruptly stopped to let him through. If I didn't stop, we would surely collide - he had no way to see me, and I had no way to see him, our view was blocked by the truck. But being human, we have such senses that even a glimpse of reflection is enough for our brain to react and avoid collision. I don't think the Lidar system can detect objects based on reflections in other objects. Heck, in the Tesla the Lidar couldn't even detect a tractor trailer that was in plain view.

 

 

It's not bias, it's common sense.

 

 

OK, I see it is pointless in talking to you about this at this stage of its development, you obviously do not understand this technology and where it is going, so I will leave you with a thought. If you were to tell the average person in the late 1960's that everyone one day would be walking around computers in their pockets, I am pretty sure they would have sent you to the loony bin.

This is 5 MB storage capacity from the late 1950's


Samsungs 512GB SSD measuring in at 20x16x1.5mm

 

 

Absolutely, and I am sure most of you have read about the Technological Singularity. If not, probably something people should read up on.

 

It isn't actually about whether this tech is not feasible, what it is about is this: is it ready for prime time right now? You can't really conflate a Manhattan traffic situation with Tempe, AZ. They are not the same thing. In order to understand the tech you have to understand the human/machine interactions and how humans rely on, interact with and react to the tech. These are early days yet. All this has to be sorted out. Until then, you risk this exact type of incident.

You can take examples from the aviation industry where numerous errors occured between human beings and technology. Read up on how the pilots of a 777 crashed their airliner into a sea wall in SFO in 2013 on a clear day, when in fact the tech could have saved them if they had relied on it. Asiana 214.

In other cases you find that the pilots of an airliner relied so heavily on the automation/avionics, that they forgot basic airmanship. AF 447.

 

You guys are comparing apples to oranges and not offering any constructive arguments. Meanwhile a lot of leading ML specialists are stating that in many areas it has reached territory of diminishing returns.

 

 

Really? I would love to hear which specialist is saying what you say and about what part of ML that is reaching diminishing returns.

 

I think that would be a subject for a separate thread, but there is tons of info on it. I'm sure you know how to search.

 

I can only come up with the huge potential in regards to ML and some researchers at Google are saying it is even more profound than the discovery of electricity. So, I don't really know where this line of thinking that you are saying is coming from. The main issue with deep learning being abstraction and reasoning will be solved, there is not denying that and that they still have a long way to go as the Singularity is still predicted to be 30 years away. But, make no mistake general A.I. is coming , it is just a matter of when.

 

https://www.wired.com/story/self-dri...rs-challenges/

AFTER PEAK HYPE, SELF-DRIVING CARS ENTER THE TROUGH OF DISILLUSIONMENT

Great read from wired.