Glad you brought this up. Computer has access to terabytes of information, much more than a human mind can possibly memorize, and yet the computer still came up with some wrong answers - not because it didn't have the information needed to provide the right answer, but because the AI couldn't properly interpret the questions and came up with wrong answers. Human players have no problem interpreting questions, but lack information to always answer them correctly.

If we apply this scenario to self driving cars, it would be equivalent to the car incorrectly interpreting a situation on the road and making a wrong decision, which can lead to an accident. If you read up on AI and machine learning, its problems and limitations, you will quickly realize how bad it can become.

 

Och, you are obviously very misinformed and or do not know much about how ML programming is done, so don't try and equate a program playing Jeopardy game with autonomous cars, it is completely different programming. And reading a few token articles does not make you an expert.

 

That article is about having driver-less tech in everyday cars. Of course it is still too expensive to have the best sensors in a production car, great work with the straw-man argument, b/c we are not talking about if the tech is ready for everyday production cars are we.

 

You are obviously missing the point of my post. If you're such an expert on machine learning, you must be aware of its shortcomings. Many actual credible experts on AI and machine learning are very skeptical regardless its capacity, especially when it comes to such complex tasks such as autonomous cars. It would be completely insane to let a self driving car to make critical decisions based on ML considering all its flaws, and I doubt much of it is even used in todays self driving cars for purposes other than data collecting for research. Actual decisions are made with direct algorithms based on real time data that is being observed by the cars sensors.

 

LOL, I am sorry I can't explain this more simply for you, but that last sentence is actually one part of MACHINE LEARNING. You saying critical decisions are not being made by ML again is showing you lack of understanding in ML and what it actually does.

The using of sensor data processing in a car, the utilization of machine learning is used to accomplish new tasks. The applications of this include evaluation of driver condition or driving scenario classification through data fusion from different external and internal sensors. Sensors like radars, cameras and lidar. Like I have been saying from the start, the application needs to gather as much data as possible to help the machine learning algorithms make lifesaving and smart decisions on past experience within the data. Autonomous vehicles programs use machine learning algorithms including convolution neural networks, deep neural networks, regression, pattern recognition, clustering and decision matrix algorithms to build an image based model for object detection, prediction and feature selection.

Adding to the above there are,

• Supervised algorithms
• Unsupervised algorithms
• Reinforcement algorithms

Supervised algorithms make use of a training data set (Lots of data) to learn and to continue to learn until the application gets to standard where errors are extremely low. The supervised algorithms can be sub-categorized into classification, regression and object detection. Unsupervised algorithms attempt to derive value from the available data. With the available data to an algorithm, it will try to develop a relation order to detect the patterns and will divide the data sets into subgroups depending on the level of similarity between them. The unsupervised algorithms can be mostly associated with rule learning and clustering. Then there are reinforcement algorithms which are between unsupervised and supervised learning. The reinforcement learning consists of time delayed and sparse labels. For each training example, in supervised learning there are target labels ,but in unsupervised learning there are no labels at all, that is a big difference in how these algorithms function.

And you do realize that decision tree generators are technically algorithms for generating algorithms.

Also, could point out the short comings of ML to me, as I am not sure we are on the same page.

I can't spell it out any more clearly than that for you Och.

 

Just another scenario that I had never encountered until living in the Phila. area. People exit a main road, at a high speed, and reenter the same road, to bypass bumper to bumper traffic and get ahead (ambulance chasing is another I never saw until here). One example? the University exit on 76W. Being able to accomplish that requires vehicles to give up the right of way, when they rightfully have it. That is a human thing imho. I would rather let a vehicle wrongfully cut across my path, than to have an accident. How does a self driving vehicle make such a decision?

I've honestly never seen this behavior before Philly. I've always thought it was poor engineering, but humans are very crafty. Like going to the wrong side of the road to beat a red light cam, or turning from an inside lane. It has to work as I don't see a flash, and people do it every day.

edit: A concrete barrier would help. I have seen PD setup dunno what you call them only to have people run them over--this is off exit 1 476 north. What I am saying is somebody in sunny SoCal, may not know all the peculiarities of every region where the vehicle needs to function. And imho it's not practical to learn through accidents.

 

I am not exactly sure of what you are saying here, but as to how autonomous cars will deal with this will depend on the situation and its surroundings.

 

It seems to me that in a dense, urban traffic situation, the way to get ahead is to do the unexpected. So unexpected that until I moved to Phila, I have never seen anyone drive this way.

Exit a bumper to bumper expressway because the exit lane is clear, at 40 mph, then cut across 3 lanes of traffic to the same onramp, to bypass x # of vehicles. A human knows this is coming up, because it happens daily, and almost everyone wants to avoid an accident, so they let the person do this. If it didn't work, people wouldn't do it. Maybe this saves the person doing it 30 sec to 2 min. of being stuck in traffic, so for the majority, not worth it.

Can an autonomous vehicle learn all these human things, prior to being deployed? Maybe the answer is yes. But I think the limiting factor will be the programmers. They don't know enough to program the vehicles as such.

Can a robot beat a human at basketball, ice hockey, etc.? If so, society is going to change quite a bit. imho for driverless vehicles to be successful, they need to be proactive and reactive. I think the proactive piece is likely missing, at least today. If it weren't, there wouldn't even be 1 fatality. We can't justify it by saying, no human could have not done the same. It doesn't pass muster.

 

Why would this behavior have any impact on an autonomous vehicle? Are you saying the vehicle should learn this and follow suit? I would expect it to act the same as most responsible drivers and not get emotional and ignore it.

 

You're assuming that the data collected from the sensors is used in machine learning and affects decision making during future tasks, but I doubt very much this is how it actually works in current autonomous cars. I am sure that the main algorithms are static, coded only react to real life data, and not affected by any data that was collected in the past. This is not machine learning.

I am also sure that this data is also stored in a separate module, to be later used in machine learning for research purposes in a simulation, to observe how the car would react to situations on the road based on what it learned, but this tech is way to fresh to allow it to affect decision making of the actual car on public roads. If you're such an expert on machine learning you should know that it is data hungry, and it is not transparent how that data is being processed, and why the machine comes up with certain decisions. This tech is never anywhere near 100% and already hitting a brick wall in much simpler tasks that do not the amount of scenarios of real world driving. It's hit a brick wall in image, speech and writing recognition already. Think about every time a speech recognition system fails to recognize speech correctly - for a self driving car that equates to failing to asses a road situation correctly and making a wrong decision, and that can be a life or death scenario. Do you think any company is reckless enough to allow it? Not to mention that even when those systems come up with the right decisions, it is not transparent why the reached the decisions, and upon review it is often revealed that the machine happened to came with the right answer, but didn't realize enough or even the right parameters.

 

I would think that with the autonomous vehicle, the name of the game is to avoid delay, and avoid accidents. Humans are able to avoid accidents by making unreasonable evasive maneuvers, in a daily commute. Not because they like doing so, but because having an accident means damage, and delay. It's like a goaltender in a shootout. He's not necessarily responding to what the shooter is doing, he knows the shooter's habits, ability, what he has done in the past, and will often totally ignore a deke., etc. A goaltender is using a combination of what's happening, and what has happened in the past, individualized to the shooters ability and what he may or may not do at this instant. Sensors cannot provide such info.

Again, I go back to the examples of Roosevelt Blvd. and Queens Blvd. I highly doubt there is a single person on this forum that would knowingly mow down a jaywalker. Yet on these roads, we know many people are killed every year, and we also know they may not have crossed with the signal. We anticipate this happening as we drive those roads. Based on AZ, then this would be the perfect place to test. If someone is killed, statistics support that many are killed anyway with human drivers. This may even mask how ineffective the autonomous vehicles are.

edit: Can an autonomous vehicle make eye contact? Do they have firm handshakes?

 

The name of the game should have less to do with saving time and more to do with safety. Does Waze instruct you to use the ramps to save a few minutes? I don't think so.

 

So what are the autonomous vehicles' capabilities then, when it is driving along at 30 mph, and two vehicles cross its path unlawfully? I would love to see an autonomous vehicle navigate the GW Bridge during rush hour. If I were an autonomous vehicle mfg., I would do it, and upload it to YouTube. Getting cutoff unlawfully and unexpectedly is actually part of the normal traffic flow. Having a pedestrian dart out of nowhere on Queens Blvd. and Roosevelt Blvd. are as well. Again, no excuse for any fatalities.

 

I'm convinced Waymo will be renamed Skynet and become the origin of our machine overlords. I'll eventually start training my kids to combat these machines once I get around to it.